Function and Structure: In Honor of Susumu Kuno (Pragmatics and Beyond New Series)

FUNCTION AND STRUCTURE

Pragmatics & Beyond New Series Editor: Andreas H. Jucker (Justus Liebig University, Giessen) Associate Editors: Jacob L. Mey (Odense University) Herman Parret (Belgian National Science Foundation, Universities of Louvain and Antwerp) Jef Verschueren (Belgian National Science Foundation, University of Antwerp) Editorial Address: Justus Liebig University Giessen, English Department Otto-Behaghel-Strasse 10, D-35394 Giessen, Germany e-mail: [email protected] Editorial Board: Shoshana Blum-Kulka (Hebrew University of Jerusalem) Chris Butler (University College of Ripon and York) Jean Caron (Université de Poitiers); Robyn Carston (University College London) Bruce Fraser (Boston University); John Heritage (University of California at Los Angeles) David Holdcroft (University of Leeds); Sachiko Ide (Japan Women’s University) Catherine Kerbrat-Orecchioni (University of Lyon 2) Claudia de Lemos (University of Campinas, Brasil); Marina Sbisà (University of Trieste) Emanuel Schegloff (University of California at Los Angeles) Paul O. Takahara (Kobe City University of Foreign Studies) Sandra Thompson (University of California at Santa Barbara) Teun A. Van Dijk (University of Amsterdam); Richard Watts (University of Bern)

59 Akio Kamio and Ken-Ichi Takami (eds) Function and Structure

FUNCTION AND STRUCTURE IN HONOR OF SUSUMU KUNO

Edited by

AKIO KAMIO Dokkyo University

KEN-ICHI TAKAMI Tokyo Metropolitan University

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences — Permanence of Paper for Printed Library Materials, ANSI Z39.48-1984.

Library of Congress Cataloging-in-Publication Data Function and structure, in honor of Susumu Kuno / edited by Akio Kamio, Ken-ichi Takami. p. cm. -- (Pragmatics & beyond, ISSN 0922-842X ; new ser. 59) Includes bibliographical references and index. 1. Functionalism (Linguistics) 2. Grammar, Comparative and general. 3. Structural linguistics. I. Kamio, Akio, 1942- . II. Takami, Ken-ichi, 1952- . III. Kuno, Susumu, 1933. IV. Series. P147.K86F86 1998 410’.1’8--dc21 98-34602 ISBN 90 272 5073 1 (Eur.) / 1 55619 822 1 (US) (alk. paper) CIP © 1999 - John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. • P.O.Box 75577 • 1070 AN Amsterdam • The Netherlands John Benjamins North America • P.O.Box 27519 • Philadelphia PA 19118-0519 • USA

TABLE OF CONTENTS

Professor Susumu Kuno (photograph by Mary Violette)

v

TABLE OF CONTENTS

vii

Table of Contents Preface Akio Kamio and Ken-ichi Takami

ix

I. FUNCTIONAL SYNTAX

1

A comparison of postposed subjects in English and Italian Gregory Ward

3

A functional constraint on extraposition from NP Ken-ichi Takami

23

The speech act empathy hierarchy and Russian possessives Olga T. Yokoyama

57

Aspects of hypothetical meaning in Japanese conditionals Wesley M. Jacobsen

83

A context-based account of English passives with indefinite subjects Aiko Utsugi

123

Observations on three anaphoric expressions in Japanese Etsuko Kaburaki Tomoda

137

Qualification and point of view Jacqueline Guillemin-Flescher

159

II. OTHER FUNCTIONAL ASPECTS OF LANGUAGE

191

Towards a theory of desirability in conditional reasoning Noriko Akatsuka

193

Relative tense and absolute tense in adverbial adjuncts in Japanese Kiri Lee

215

Specific NP in scope Becky Kennedy

251

viii

TABLE OF CONTENTS

Some referential properties of English it and that Akio Kamio and Margaret Thomas

289

The role of empathy in sentence production: A functional analysis of aphasic and normal elicited narratives in Japanese and English Lise Menn, Akio Kamio, Makoto Hayashi, Ikuyo Fujita, Sumiko Sasanuma, and Larry Boles

317

Personal pronoun shift in Japanese: A case study in lexical change and point of view John Whitman

357

Subject Index Name Index

387 395

PREFACE

ix

Preface For nearly three decades, Susumu Kuno has been one of the leading figures in American functional linguistics. American functionalism is not monolithic, as everyone concerned with it well recognizes. Kuno’s functionalism has several components, the combination of which makes his position unique and important in contemporary linguistic research. The guiding principle informing his work is his belief that language has remarkably complicated functions in human communication and that the structure of language must be based on these fundamental functions. The title of this Festschrift “Function and Structure”, which was first proposed by John Whitman, one of his students, directly reflects this belief. Many of the papers included in this volume also directly or indirectly reflect it, as readers may confirm for themselves. The second significant component is his attention to the formal aspect of linguistic structure. Most functionalists are critical of formal linguistics and do not much concern themselves with it. Some are even antagonistic to formal analysis and disregard it entirely. Kuno, however, has always kept up with the development of formal linguistics, in particular, formal syntax within generative grammar. While it is true that his work involving formal syntax is largely critical of current research in generative grammar, his criticism is based on a deep understanding of it. His position has therefore produced rare and invaluable contributions to the development of linguistic theory. On the one hand, he has repeatedly pointed out the inadequacy of research based on the narrowly formalist view of many formal syntacticians, while, on the other he has urged these syntacticians to take into consideration the communicative functions of natural language. The third major component of Kuno’s work is his deep and wide-ranging research in Japanese linguistics. Japanese is his native language, and his keen and very insightful intuition has led to enlightening and epoch-making books and a huge number of informative papers on Japanese which have deeply affected the trend of linguistic research in both the United States and Japan. It should also be noted that many of his ideas about the analysis of other

x

PREFACE

languages come from his own detailed observations of Japanese. The fourth important component may be his familiarity with the computer processing of language. In fact, before he launched into research in general linguistics, he was already an authority in machine translation. His interest and work in this field continue to this day. His deep concern with this area of study seems closely tied to his fundamental functionalist view of language and his critical attitude toward the overly strict structuralist view. Thus, Kuno is a general linguist with a highly productive and original research record, an influential Japanese linguist, and a well-recognized researcher in the field of computational linguistics. This amazingly broad and deep research history makes him a marvelous and unique linguist. As everyone who knows him is aware, his mind works very quickly and clearly. In a moment he can find a counterexample that destroys an entire theory. He then can come up with an alternative idea that is functionally oriented and structurally superior to the ideas of those whose theory his counterarguments have just devastated. Further, as everyone who knows him will testify, he is extremely generous with his ideas. He gives them freely to whoever seeks his advice and suggestions. Moreover, he is always generous with his time and gives it generously to those who need his guidance, despite the fact that it is he himself who always needs time most. This Festschrift is dedicated to him by those who have benefited most from his generous and friendly, but at the same time scholarly and rigorous guidance and teaching, as an expression of our profound gratitude and admiration. We hope that the quality and scope of the papers in this volume will satisfy his unusually high standards for scholarly work. If so, then the papers will certainly deserve the attention of a wider audience in the community of functional linguists around the world. Finally, we are deeply grateful to those who have cooperated with us in this project. In particular, our thanks go to Kazuhiro Ichikawa, whose unsparing effort made the generation of the final text possible. Akio Kamio and Ken-ichi Takami

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

I. FUNCTIONAL SYNTAX

1

2

GREGORY WARD

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

3

A Comparison of Postposed Subjects in English and Italian* Gregory Ward Northwestern University

1.

Introduction

As is well known, different languages have different ways of marking given and new information in discourse. One option available in many languages is non-canonical word order; that is, certain languages provide their speakers with various marked syntactic constructions that can be used to indicate the information status of discourse entities. In this paper, I examine the discourse properties of one set of such constructions — subject postposing — in English and Italian. For the purposes of this study, a subject is ‘postposed’ if the logical subject appears post-verbally, typically (but not always) sentencefinally, leaving preverbal position either empty or occupied by an expletive element (Birner and Ward 1996). For convenience, I shall use the term POSTPOSING to refer to such constructions despite the suggestion of movement associated with the term; no commitment to any particular syntactic analysis should be inferred.1 The discussion will focus on four constructions: two from English — existential there-sentences and presentational there-sentences — and two from Italian — existential ci-sentences and subject postposing. These con* An earlier and much abridged version of this paper was presented at the 1996 LSA Annual Meeting in San Diego. I thank Betty Birner and Louise McNally for many helpful comments on an earlier draft of this paper. For their help in providing countless native speaker judgments on Italian, I thank Salvatore Attardo, John Bartlett, Manuela Pinto, Massimo Poesio, and Raffaella Zanuttini. For supplying some of the naturally-occurring data used in this study, I thank Beth Levin, Richard Sproat, and David Yarowsky. This research was supported by NIDCD grant R01DC01240.

4

GREGORY WARD

structions are illustrated in (1a) - (1d), respectively: (1)

a. There’s a problem with our analysis. [BB to GW in conversation] b. Daniel told me that shortly after Grumman arrived at Wideview Chalet there arrived also a man named Sleema. [Upfield 1987:246] c. C’è un segreto istruttorio. there’s a secret inquest ‘There’s a secret inquest.’ [ANSA 7/91] d. Era salita tua sorella sull’autobus. boarded your sister on the bus ‘Your sister got on (the bus).’ [adapted from Saccon 1993:169, ex.104]

An examination of naturally-occurring data reveals that all four sentencetypes share a common discourse constraint: Each requires the NP in postposed position to represent an entity that is unfamiliar in some sense — either to the discourse or (the speaker believes) to the hearer. However, the particular type of unfamiliarity to which the postposed NP is sensitive will be shown to differ among the constructions in question. Following Prince (1992), I shall refer to entities that are new to the discourse as DISCOURSE-NEW and those that are new to the hearer as HEARER-NEW; likewise, entities that have been evoked in the prior discourse will be called DISCOURSE-OLD, and those that are assumed to be known to the hearer HEARER-OLD. This distinction captures the fact that what is new to the discourse needn’t be new to the hearer (cf. Firbas 1996, Chafe 1976, Lambrecht 1994); that is, an entity may be hearer-old, yet discourse-new. However, presumably entities that are discourse-old will, by virtue of having been evoked in the prior discourse, be assumed to be hearerold as well.2 The distinction between discourse-familiarity and hearer-familiarity thus results in four possible information statuses, of which only three normally occur in naturally occurring discourse: Hearer-old, discourse-old — Information which has previously been evoked in the current discourse, and which the speaker therefore believes is known to the hearer.

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

5

Hearer-old, discourse-new — Information which has not been evoked in the current discourse, but which the speaker nonetheless believes is known to the hearer. Hearer-new, discourse-new — Information which has not been evoked in the current discourse, and which the speaker does not believe to be known to the hearer. Hearer-new, discourse-old — Theoretically, information which has been evoked in the current discourse, but which the speaker nonetheless believes is not known to the hearer. As Prince (1992) notes, this type of information typically does not occur in natural discourse. To illustrate the various possibilities, consider the following (discourse initial) utterance: (2)

A friend of mine at Stanford told me that he saw Chelsea Clinton working out in the gym yesterday.

Here, Chelsea Clinton represents information that is discourse-new but hearerold, referring to an entity which has not been evoked in the prior discourse but which can be assumed to be known to the hearer; a friend of mine at Stanford represents information that is both discourse-new and hearer-new, having not been previously evoked and also being (presumably, given the indefinite) unknown to the hearer; and he represents information that is discourse-old and (therefore necessarily) hearer-old, having been explicitly evoked in the previous clause (as a friend of mine). (See also Lambrecht 1994.)

2.

English There-Sentences

As noted in Ward and Birner (1995), previous work on pragmatics of English there-sentences has generally focused on there-sentences with main verb be. However, as argued by Aissen (1975), Larson (1988), Rochemont and Culicover (1990), McNally (1992), inter alia, there are in fact two structurally distinct types of English there-sentences: EXISTENTIAL there, restricted to main verb be, and PRESENTATIONAL there, restricted to verbs other than be.3 Notwithstanding any structural differences that may exist between them, the two types of there-sentences have been shown to be pragmatically distinct with respect to the information status of the postverbal NP (henceforth

6

GREGORY WARD

PVNP). As shown in Ward and Birner (1996), the PVNP of existential theresentences is sensitive to HEARER-STATUS, while that of presentational there-sentences is sensitive to DISCOURSE-STATUS. The discourse constraints of these two types of there-sentences will be discussed in turn, beginning with existential there. 2.1 English Existential There-Sentences As noted by Prince (1988, 1992) and, more recently, by Ward and Birner (1995), English existential there-sentences are sensitive to hearer-status. That is, the postverbal NP of existential there-sentences is constrained to represent entities that the speaker believes are not already familiar to the hearer. Consider the examples of existential there-sentences with hearer-new PVNPs in (3): (3)

a. “There’s a warm relationship, a great respect and trust” between [United Air Lines]’s chairman, Stephen M. Wolf, and Sir Colin Marshall, British Air’s chief executive officer, according to a person familiar with both sides. [Wall Street Journal, 8/23/89] b. What can happen is a hangup such as Rocky Smith ran into, as the independent hauler was traversing Chicago with a load of machinery that just had to get to a factory by morning. “There was this truck in front of me carrying giant steel coils, and potholes all over the place,” he remembers. “This guy swerves all of a sudden to avoid a big hole.” He hit it anyway. [Wall Street Journal, 8/30/89]

In both of these examples, the PVNP is hearer-new in that it represents an entity that is being introduced to the hearer/reader for the very first time.4 If the PVNPs of existential there-sentences are indeed sensitive to hearerstatus (rather than discourse-status), we would expect that attempts to situate NPs representing hearer-old referents in PVNP position would produce infelicity. This is in fact the case: Hearer-old referents are impermissible as PVNPs in existential there-sentences regardless of the discourse-status of those referents. First, consider examples of hearer-old, discourse-new PVNPs, as in (4): (4)

a. I have some news you’re going to find very interesting. #There was on the panel your good friend Jim Alterman. b. President Clinton appeared at the podium accompanied by

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

7

three senators and the Speaker of the House. #There was behind him the Vice President. The PVNPs in these examples represent entities that are presumably familiar to the hearer, yet new to the discourse; in both cases, the existential theresentence is unacceptable. Now consider there-sentences with discourse-old PVNPs, as in (5): (5)

a. A: Hey, have you heard from Jim Alterman lately? I haven’t seen him for years. B: Yes, actually. #There was on the panel today Jim Alterman. b. President Clinton appeared at the podium accompanied by three senators and the Vice President. #There was behind him the Vice President. [=Ward and Birner 1996, ex. 14b]

Here, the PVNPs represent entities that are discourse-old, which are necessarily also hearer-old: If an entity has been evoked in the discourse, it follows that the hearer can be presumed to be familiar with it. And, as predicted, such examples are infelicitous. Thus, whenever an NP represents a hearer-old entity, it is disallowed in the postverbal position of an existential there-sentence. 2.2 English Presentational There-Sentences As noted above, a number of previous syntactic analyses of English theresentences have argued that so-called ‘presentational’ there-sentences are structurally distinct from existential there-sentences. In Ward and Birner (1996), we argue that presentational there-sentences are also functionally distinct. Based on an analysis of a corpus of over 400 naturally-occurring tokens, we show that, unlike existential there-sentences, presentational theresentences are sensitive to the discourse-status of the PVNP. In the vast majority of cases, the referent of the PVNP in a presentational there-sentence is both hearer-new and discourse-new, as in (6): (6)

And so as voters tomorrow begin the process of replacing Mr. Wright, forced from the speaker’s chair and the House by charges of ethical violations, there remains a political vacuum in the stockyards, barrios, high-tech workshops and defense plants of Tarrant County. [AP Newswire 1989]

8

GREGORY WARD

In the news story from which this token is taken, the PVNP is the first reference to the political vacuum in question and can therefore safely be assumed to represent a new entity to the readership. However, the PVNP of presentational there-sentences may also represent a referent that, while new to the discourse, is presumably already familiar to the hearer/reader. Consider the examples in (7): (7)

a. There only lacked the moon; but a growing pallor in the sky suggested the moon might soon be coming. [adapted from Erdmann 1976:138] b. Suddenly there ran out of the woods the man we had seen at the picnic.5 [Aissen 1975, ex. 12] c. Famous men came — engineers, scientists, industrialists; and eventually, in their turn, there came Jimmy the Screwsman and Napoleon Bonaparte […]. [Upfield 1950:2]

In these examples, the referent of the PVNP is one that is familiar to the hearer, yet new to the discourse. In (7a) the moon has not been evoked previously in the discourse, yet it is clearly hearer-old. Similarly, in (7b) and (7c), the PVNPs represent individuals that can be assumed to be already familiar to the hearer/reader. Thus, while both types of there-sentences allow hearer-new, discourse-new PVNPs, they do so for different reasons: Existential there-sentences, being sensitive to hearer-status, require hearer-new PVNPs, while presentational there-sentences, being sensitive to discoursestatus, require discourse-new PVNPs. As predicted, presentational there-sentences — like existential theresentences — disallow PVNPs representing discourse-old entities. Consider the examples in (8): (8)

a. A: Hey have you heard from Jim Alterman lately? I haven’t seen him for years. B: Yes, actually. #There appeared before the committee today Jim Alterman. b. President Clinton appeared at the podium accompanied by three senators and the Vice President. #There stood behind him the Vice President.

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

9

The PVNPs in these examples represent discourse-old (and therefore also hearer-old) entities in context, and hence are infelicitous in postverbal position. Note that both of the presentational there-sentences in (8) would be acceptable without prior mention of the PVNP’s referent — i.e., with the PVNP representing an entity that is hearer-old but discourse-new. Thus, unlike the PVNPs of existential there-sentences, those of presentational there-sentences are sensitive to the discourse-status of their referents; that is, reference to previously evoked entities is disallowed in this position.

3.

Italian Presentational Ci-Sentences

Like English (and many other languages), Italian also has a presentational construction (‘c’è presentativo’), consisting of an expletive or dummy surface subject, some form of the verb essere (‘be’), and a postverbal logical subject. The expletive element is the same lexical item as the locative anaphor ci (‘there’), from which it is derived. However, as with the expletive there of the English existential and presentational constructions, no locative meaning is associated with ci in its presentational use. Consider the Italian ci-sentence in (9): (9)

C’è un ponticello dove ogni anno, la notte del 2 there’s a bridge where each year, the night of the 2 aprile appare un fantasma. april appears a ghost. ‘There’s a bridge where, each year on the night of April 2nd, a ghost appears.’ [ANSA, 5/91]

Since this is a discourse-initial utterance with no prior location specified, the ci in (9) cannot be interpreted anaphorically and must therefore be interpreted as the expletive of a ci-sentence. Pragmatically, the PVNP of Italian ci-sentence, like that of English there-sentences, is constrained to represent entities that are new to the discourse (Berruto 1986). In (9), for example, the PVNP — un ponticello (‘a bridge’) — represents an entity that is new both to the discourse and presumably to the hearer/reader as well. However, postverbal position in ci-sentences, like that of English presentational there-sentences, admits hearer-old

10

GREGORY WARD

information, as long as that information is discourse-new. Consider the examples in (10):6 (10) a. Se c’è la mafia in Italia, è perché c’è la if there’s the Mafia in Italy, is because there’s the Democrazia Cristiana. Democracy Christian ‘If there’s a Mafia in Italy, it’s because of the Christian Democrat Party.’ [ANSA, 9/92] b. Oggi, c’è il sole. today, there’s the sun ‘Today, the sun is out.’ c. Oggi c’è il Papa e ci sono le baracche. today there’s the Pope and there are the shacks ‘Today, the Pope is here and there are shacks.’ [ANSA, 10/92] In each of these examples, the referent of the PVNP is new to the discourse, though it can safely be assumed to be already familiar to the reader. On the other hand, when the PVNP represents an entity that is discourseold, as in (11), infelicity results: (11) A: Ho parlato con la Giulia oggi. Sta molto bene. I have spoken with the Julia today. is very well ‘I talked to Julia today. She’s doing well.’ B: Mi fa piacere. #A proposito, sai che to-me makes happy. by the way, you know that c’era la Giulia alla festa di Paolo ieri there was the Julia at the party of Paul yesterday sera? evening ‘That’s good. By the way, did you know that Julia was at Paul’s party last night? Here, Guilia has been evoked within the same discourse and, as a discourseold entity, cannot be represented by a PVNP in an Italian ci-sentence.

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

11

Thus, Italian ci-sentences require that the PVNP represent a discoursenew entity, regardless of hearer-status. This is exactly the constraint found for English presentational there-sentences, though not for English existential there-sentences. Indeed, translating the ci-sentences in (10) with the corresponding PVNP in an English existential there-sentences results in infelicity, as illustrated in (12): (12) a. #If there’s the Mafia in Italy, it’s because there’s the Christian Democrat party. b. #Today, there’s the sun. c. #Today, there’s the Pope and there are the shacks. Note that it is the information status and not the definiteness of the Italian PVNP that renders these English translations infelicitous; when a definite PVNP can be interpreted as hearer-new in Italian, both the ci-sentence and its English equivalent are perfectly appropriate. Consider the examples in (13): (13) a. C’era quel buffone di cui ti ho parlato there was that fool of whom to you (I) spoke alla festa ieri. at the party yesterday. ‘There was that fool I talked to you about at the party yesterday.’ b. C’erano le solite obiezioni da parte dei docenti there were the usual objections from the faculty alla riunione. at the meeting. ‘There were the usual objections from the faculty at the meeting.’ c. Secondo Agnoletto, c’è il rischio che si tenda according to Agnoletto, there’s the risk that one tends a limitare il diritto dei bambini sieropositivi a to limit the right of the children seropositive to frequentare la scuola insieme agli altri. frequent the school together with the others ‘According to Agnoletto, there’s the risk that you might restrict the right of HIV-positive children to go to school with other children.’ [ANSA, 5/91]

12

GREGORY WARD

In fact, the same types of definite hearer-new PVNPs are found in Italian cisentences and English there-sentences (see Ward and Birner 1995 for a discussion of definite PVNPs in English existential there-sentences). However, while they both tolerate definite hearer-new PVNPs, English existential there-sentences and Italian ci-sentences are nonetheless sensitive to different discourse constraints. Although the two constructions are formally quite similar (both in the presence of the expletive and in the presence of main-verb be), we have shown that English existential there-sentences are sensitive to hearer-status, while Italian ci-sentences are sensitive to discoursestatus. In fact, pragmatically speaking, ci-sentences have much more in common with English presentational there-sentences, which likewise require discourse-new PVNPs. Just as presentational there-sentences disallow discourse-old information in postverbal position, so too do ci-sentences. In this way, the Italian ci construction corresponds pragmatically not to the English existential construction — its syntactic analog — but rather to the English presentational construction.

4.

Italian Subject Postposing

In addition to ci-sentences there is another construction in Italian that shares the pragmatic constraints found for English presentational there-sentences, namely subject postposing (Burzio 1986, Belletti 1988, Calabrese 1992, Saccon 1993, Lambrecht 1994, Pinto 1994).7 In this construction — restricted to verbs other than essere (‘be’) — the subject appears in postverbal position with nothing in preverbal position, as illustrated in (14): (14) È arrivato stamattina una lettera dall’America. arrived this morning a letter from the America ‘A letter from America arrived this morning.’ Here, the indefinite PVNP represents an entity that is both new to the discourse and presumably to the hearer as well. However, as previous studies have pointed out (e.g. Calabrese 1992, Saccon 1993, Pinto 1994, inter alia), what is relevant for felicitous subject postposing is the discourse-status of the PVNP. The postverbal position of subject postposing is restricted to those NPs representing entities new to the discourse — or, equivalently, NPs that are ‘non-anaphoric’ (Calabrese 1992), ‘non-presupposed’ (Saccon 1993), or ‘-d-linked’ (Pinto 1994). Consider the examples of discourse-new, hearer-old

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

13

PVNPs in (15): (15) a. È morto Mario. died Mario ‘Mario died.’ [Calabrese 1992, ex. 9] b. Era salita tua sorella sull’autobus. boarded your sister on the bus ‘Your sister got on (the bus).’ [adapted from Saccon 1993:169, ex. 104] In each of these examples, the PVNP represents an entity with which the hearer is presumably familiar; what licenses the subject postposing is the fact that it is new to the discourse. When an NP representing a discourse-old entity appears in postverbal position, infelicity results. Consider Saccon’s (1993: 169) example in (16) (slightly modified, but with her original judgments): (16) [Context: A and B are sitting together on a bus] A: Mi sono addormentato, ma mi è sembrato di (I) fell asleep, but to me seemed averti sentito parlare con qualcuno. to have you heard speak with someone ‘I fell asleep, but I thought I heard you talking to someone.’ B: Si, infatti sono saliti tua sorella e tuo fratello. yes, in fact boarded your sister and your brother ‘Yes, in fact your sister and brother got on [the bus].’ A: È dove sono andati ora? and where (they) are gone now ‘And where have they gone now?’ a. B: *È scesa tua sorella a far spese in centro. got off your sister to do shopping in center ‘Your sister got off to go shopping downtown.’ b. B: Tua sorella è scesa a far spese in centro. your sister got off to do shopping in center ‘Your sister got off to go shopping downtown.’ In the subject postposing in B’s initial utterance in (16), the PVNP tua sorella e tuo fratello (‘your sister and brother’) represents a hearer-old, discourse-new entity (consisting of a set of two individuals) and, as such, it is fully felicitous

14

GREGORY WARD

as a postverbal subject. The unacceptability of B’s reply in (16a), on the other hand, can be attributed to the discourse-old status of the PVNP in context. Note that the corresponding canonical-word-order utterance in (16b) is felicitous in the same context. We may conclude, then, that both ci-sentences and postposed subjects in Italian, like presentational there-sentences in English, are sensitive to the discourse-status of the postposed subject.

5.

Italian Inversion

When subject postposing occurs with concomitant preposing, the result is an inverted word order that shares many properties with its English analog, locative inversion. As argued by Birner (1992, 1994), the felicity of English inversion, which she refers to as an ‘argument reversing construction’, depends on the relative discourse-familiarity of the preposed and postposed constituents. Specifically, the preposed constituent in an inversion may not represent less familiar information in the discourse than does the postposed constituent. Birner found many examples where both the preposed and postposed constiturents represented previously evoked information, but in that case it was consistently the more recently evoked (and thus arguably more familiar) information that appeared in preposed position. Following Birner (1996), it can be hypothesized that the felicity of argument-reversing constructions in general may be sensitive to the relative — as opposed to absolute — information status of the reversed constituents. If this is so, we would predict that Italian inversion, like invention in English, is sensitive to relative rather absolute discourse-status, a prediction that is supported by the data. First, consider (17): (17) C’è un nuovo albergo a Verona. In questo albergo there’s a new hotel in Verona. in this hotel lavora Chelsea Clinton/un greco. works Chelsea Clinton/a Greek man ‘There’s a new hotel in Verona. In this hotel, Chelsea Clinton/a Greek man works.’ In the inversion in (17), the preverbal position is occupied by information that is discourse-old (this hotel), while the postverbal position may be felicitously occupied by information that is either hearer-new (a Greek man) or hearer-old

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

15

(Chelsea Clinton), as long as it is also discourse-new. On the other hand, when the preverbal constituent represents discourse-new information, and the postverbal constituent represents discourse-old information, as in (18), infelicity results: (18) A: Come va tuo zio? how goes your uncle? ‘How’s your uncle?’ B: Bene, grazie. #In un’albergo lavora mio zio. well, thanks. in a hotel works my uncle. ‘Good, thanks. My uncle works in a hotel.’ Here, the postverbal mio zio represents discourse-old information while the preverbal constituent represents discourse-new information (a hotel), and the resulting inversion is infelicitous. Thus, as with inversion in English, the postposed constituent of an Italian inversion may not represent information that is more familiar than that represented by the preposed constituent. However, that is not to say that the postposed constituent of an Italian inversion may NEVER represent discourse-old information; on the contrary, postverbal position may be occupied by a discourse-old NP just in case there is another entity more recently evoked within the same discourse that is realized in preverbal position. Consider the following example of an inversion with a PVNP representing a discourse-old entity: (19) Il presidente doveva fare un discorso alla Camera e aveva bisogno del supporto di tutto il governo. Dubitava però dell’appoggio del vicepresidente, e fino all’ultimo minuto tutti i suoi collaboratori più stretti avevano dubitato che il vicepresidente si sarebbe presentato. Quando la seduta cominciò, i collaboratori del presidente tirarono un sospiro di sollievo. Dietro di lui sorrideva sornione il vicepresidente. The President had to give a speech to the Parliament and needed the support of the entire government. However, he had doubts about the support of the Vice President, and up to the last minute all of his closest advisors doubted that the Vice President would show up. When the meeting began, the President’s advisors breathed a collective sigh of relief. Behind him the Vice President was smiling coyly. (lit. ‘Behind him was smiling coyly the Vice President’.)

16

GREGORY WARD

In (19), the postposed NP il vicepresidente represents a familiar entity, having been evoked in the prior discourse. However, the NP of the preposed PP— lui (‘him’) — represents an even more familiar entity (the President), in the sense of having been more recently evoked than the entity represented by the postposed constituent, and the inversion is consequently felicitous. When the preposed NP does not represent discourse-old information that is more recently evoked, infelicity results, as seen in (20): (20) Il presidente doveva fare un discorso alla Camera e aveva bisogno del supporto di tutto il governo. Dubitava però dell’appogio del vicepresidente, e fino all’ultimo minuto tutti i suoi collaboratori più stretti avevano dubitato che il vicepresidente si sarebbe presentato. Quando la seduta cominciò, i collaboratori tirarono un sospiro di sollievo. #Dietro di lui sorrideva sornione il vicepresidente. The President had to give a speech to the Parliament and needed the support of the entire government. However, he had doubts about the support of the Vice President, and up to the last minute all of his closest advisors doubted that the Vice President would show up. When the meeting began, the advisors breathed a collective sigh of relief. Behind him the Vice President was smiling coyly. (lit. ‘Behind him was smiling coyly the Vice President’.) Here, the preposed lui is LESS recently evoked, i.e. evoked earlier in the discourse, than the postposed vicepresidente and the result is infelicitous. In this way, Italian inversion patterns exactly like inversion in English. That is, the felicity of each of these argument-reversing constructions depends on the relative discourse-familiarity of the preposed and postposed constituents. In the case of Italian subject postposing, on the otherhand, there is no reversal of arguments, and the constraint on postverbal position is absolute rather than relative; the same is true for Italian ci-sentences. The postposed subjects of both ci-sentences and subject postposing, then, pattern just like those of English presentational there-sentences: They must be new to the discourse in absolute, rather than relative, terms. We can summarize our comparison of English and Italian postposing and argument-reversing constructions as shown in Table 1.

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

17

Table 1. Comparison of Subject-Postposing Constructions in English and Italian be/essere

non-be/essere

Italian ci-sentences

English presentational there-sentences

Discourse-New PVNP

Italian subject postposing Hearer-New PVNP

English existential there-sentences

Relatively New PVNP

English inversion, Italian inversion

Here, we see that both English and Italian employ different sentence-types for postposing depending on the matrix verb. In Italian, use of essere (‘be’) requires presentational ci; use of verbs other than essere requires subject postposing. Both constructions, however, are sensitive to the same constraint on the PVNP. In English, on the other hand, expletive there is possible with both be and non-be verbs; however, the PVNPs in these two cases are sensitive to different discourse constraints.

6.

Conclusion

In this paper I have considered the pragmatic constraints on the use of a number of constructions in which the logical subject appears in postverbal position, leaving the preverbal position empty or occupied by an expletive element. All such constructions were found to require that the postposed NP represent an entity that is unfamiliar in some sense although the type of unfamiliarity required differs among the various sentence-types. English presentational there-sentences and Italian ci-sentences and subject postposing were shown to be sensitive to the discourse-status of the postposed constituent. English existential there-sentences, on the other hand, were shown to be sensitive to the hearer-status of this constituent. When subject postposing in Italian is accompanied by preposing, the constraint on discourse-status becomes relative: The entity represented by the preposed constituent must be more familiar within the discourse than that represented by the postposed constituent. In contrast, English there-sentences were shown to be sensitive to absolute information status with or without the presence of preposing: The PVNP of existential there-sentences must be

18

GREGORY WARD

hearer-new while the PVNP of presentational there-sentences need only be discourse-new. These findings demonstrate that it is possible for two formally equivalent constructions in different languages — e.g., Italian ci-sentences and English existential there-sentences — to be subject to different pragmatic constraints. The PVNP of an English existential there-sentence must represent hearernew information, while that of an Italian ci-sentence must represent discourse-new information. Second, two formally distinct constructions within a single language may be subject to precisely the same pragmatic constraint, as with Italian ci-sentences and subject postposing. Finally, formally distinct constructions in two different languages — e.g., presentational there-sentences in English and subject postposing in Italian — may share the same pragmatic constraint, in this case that the PVNP must represent discoursenew information. From this, we can conclude that formally equivalent constructions from different languages can be subject to distinct discourse constraints, and that distinct constructions both across and within languages can nonetheless share discourse constraints.

Notes 1.

Such cases of subject postposing are to be distinguished from right-dislocation of the subject, in which the marked postverbal placement of the logical subject is accompanied by the appearance of a coreferential pronoun in syntactic subject position (Birner and Ward 1996). Thus, the crucial feature distinguishing between postposings and dislocations, as defined here, is the presence or absence of a coreferential pronoun occupying the marked constituent’s canonical position.

2.

As Birner (1994) observes, the discourse-old/discourse-new distinction is clearly related to, though not isomophic with, the theme-rheme distinction discussed by Firbas 1996, inter alia, where the theme is said to convey information that is known or can be gathered from the preceding sentence. Although the same intuitions motivating the theory of communicative dynamism motivate the discourse-old/discourse-new distinction, the latter seems to provide a more concrete way of accounting for the data.

3.

There is a great deal of controversy regarding the correct characterization of verbs that can felicitously occur with presentational there (as well as so-called locative inversion). For example, it has been claimed that such verbs are exclusively unaccusative (e.g., Levin 1983), exclusively presentational (e.g., Bresnan 1994), or exclusively verbs of existence or appearance (e.g., Penhallurick 1984). Birner (1995), however, argues convincingly that the correct formulation is in terms of information structure, not semantic class.

4.

See below for a discussion of the felicity of (morphologically) definite PVNPs in English there-sentences.

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

19

5.

Assume here a context in which the referent of the PVNP is hearer-old, discourse-new.

6.

Excluded from consideration are ci-sentences with preposed locative PPs as they could be analyzed as locative sentences with clitic left-dislocation, e.g., Dietro di lui, c’era un gattino (‘Behind him there was a kitten’). I thank Louise McNally for bringing this possibility to my attention.

7.

Previous studies use the term ‘subject inversion’ for this construction; however, that term is misleading in that it is used to refer to pure subject postposing (with no concomitant preposing) as well as argument reversal (involving preposing and postposing). As we shall see, the two are subject to distinct discourse constraints and, arguably, constitute distinct constructions; thus, distinct terms will be used to refer to them. It should also be noted that there is a superficially similar construction in Italian in which the logical subject also appears postverbally. This construction, variously termed right-dislocation or ‘emargination’, requires an intonational boundary between the VP and the postposed subject. As Calabrese (1992) and Saccon (1993) point out, emargination is both formally and functionally distinct from subject inversion, in which the matrix clause and postposed subject occur within a single intonational phrase.

References Aissen, Judith 1975 “Presentational-there insertion: A cyclic root transformation”. CLS 11: 1-14. Belletti, Adriana 1988 “The case of unaccusatives”. Linguistic Inquiry 19: 1-34. Berruto, Gaetano 1986 “Un tratto sintattico dell’italiano parlato: Il c’è presentativo”. In K. Lichem, E. Mara, and S. Knaller (eds), 61-73. Birner, Betty 1992 The discourse function of inversion in English. Dissertation, Northwestern University. Birner, Betty 1994 “Information status and word order: An analysis of English inversion”. Language 70: 233-259. Birner, Betty 1995 “Pragmatic constraints on the verb in English inversion”. Lingua 97: 233256. Birner, Betty 1996 “Passivization and inversion: A functional correspondence”. Paper presented at the 1996 LSA Annual Meeting, San Diego. Birner, Betty and Gregory Ward 1996 “A crosslinguistic study of postposing in discourse”. Language and Speech: Special Issue on Discourse, Syntax, and Information 39: 111140.

20

GREGORY WARD

Bok-Bennema, Reineke and Crit Cremers (eds) 1994 Linguistics in the Netherlands. Amsterdam/Philadelphia: John Benjamins. Bresnan, Joan 1994 “Locative inversion and the architecture of universal grammar”. Language 70: 72-131. Burzio, Luigi 1986 Italian syntax. Dordrecht: Reidel. Calabrese, Andrea 1992 “Some remarks on focus and logical structures in Italian”. Harvard Working Papers in Linguistics 1: 91-127. Chafe, Wallace 1976 “Givenness, contrastiveness, definiteness, subjects, topics, and point of view”. In C. Li (ed), 25-55. Erdmann, Peter 1976 There sentences in English. Munich: Tuduv. Firbas, Jan 1966 “Non-thematic subjects in contemporary English”. Travaux Linguistiques de Prague 2: 39-56. Goldberg, Adele (ed) 1996 Conceptual structure, discourse and language. Center for the Study of Language and Information: Cambridge University Press. Lambrecht, Knud 1994 Information structure and sentence form. Cambridge: Cambridge University Press. Larson, Richard 1988 “Light predicate raising”. Lexicon Project Working Papers 27. MIT: Center for Cognitive Science, 1-104. Levin, Lori 1983 Operations on lexical forms: Unaccusative rules in Germanic languages. Dissertation, MIT. Li, Charles (ed) 1976 Subject and topic. New York: Academic Press. Lichem, Klaus, Edith Mara, and Susanne Knaller (eds) 1986 Parallela 2: Aspetti della sintassi dell’italiano contemporaneo. Tübingen: G. Narr. McNally, Louise 1992 An interpretation for the English existential construction. Dissertation, University of California, Santa Cruz. Penhallurick, John 1984 “Full-verb inversion in English”. Australian Journal of Linguistics 4: 3356. Pinto, Manuela 1994 “Subjects in Italian: Distribution and interpretation”. In R. Bok-Bennema and C. Cremers (eds), 175-186.

A COMPARISON OF POSTPOSED SUBJECTS IN ENGLISH AND ITALIAN

21

Prince, Ellen F. 1988 “The discourse functions of Yiddish expletive es + subject postposing”. Papers in Pragmatics 2: 176-194. Prince, Ellen F. 1992 “The ZPG letter: Subjects, definiteness, and information-status”. In S. Thompson and W. Mann (eds), 295-325. Rochemont, Michael and Peter Culicover 1990 English focus constructions and the theory of grammar. Cambridge: Cambridge University Press. Saccon, Gabriella 1993 Post-verbal subjects. Dissertation, Harvard University. Thompson, Sandra and William Mann (eds) 1992 Discourse description: Diverse analyses of a fundraising text. Amsterdam/Philadelphia: John Benjamins. Upfield, Arthur W. 1946 The devil’s steps. New York: Collier Books reprint, 1987. Upfield, Arther W. 1950 The Bachelors of broken hill. New York: Collier Books reprint, 1984. Ward, Gregory and Betty Birner 1995 “Definiteness and the English existential”. Language 71: 722-742. Ward, Gregory and Betty Birner 1996 “On the discourse function of rightward movement in English”. In A. Goldberg (ed), 463-479.

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

23

A Functional Constraint on Extraposition from NP* Ken-ichi Takami Tokyo Metropolitan University

1.

Introduction

This paper discusses the phenomenon of the so-called “Extraposition from NP,” as illustrated in (1a, b) below: (1)

a. A review of Chomsky’s book appeared. b. A review appeared of Chomsky’s book.

It has generally been assumed in the literature that (1b) is derived from (1a) by extraposing the PP of Chomsky’s book, part of the subject NP a review of Chomsky’s book in (1a), to sentence-final position.1 (1b) is a perfectly acceptable sentence, but this does not mean that all sentences involving Extraposition from NP result in acceptability. On the contrary, there are in fact many unacceptable sentences, as will be shown later in this paper. In the following section, I consider under what conditions Extraposition from NP is permissible, and attempt to offer a functional constraint controlling the acceptability vis-à-vis the unacceptability of this phenomenon. In Section 3, I briefly review alternative semantic constraints on Extraposition

* I am deeply indebted to Susumu Kuno for his numerous insightful and informative comments on the content of this paper. I am also indebted to Akio Kamio, Heizo Nakajima, Karen Courtenay, Ben Fortson and two anonymous reviewers for their invaluable and constructive comments on both content and exposition. I would like to thank the following people, who were generous with their native speaker intuitions: Bruce Davison, Michael Lloyd, John Maher, Sara Backer, Dianne Jonas, Dan Coleman, Cheryl Zoll, Beth Chappel. I also benefited from discussion with Mitsumasa Zushi on some of the example sentences in the paper.

24

KEN-ICHI TAKAMI

from NP, put forward by Guéron (1980) and Rochemont and Culicover (1990). I then point out that there are some major difficulties with their analyses. Section 4 is concerned with sentences involving not only Extraposition from NP but also predication relation. Observe the following: (2)

a. John left the party [that Mary gave for him] yesterday [extremely angry and upset]. b. John left the party yesterday [that Mary gave for him] [extremely angry and upset].

Sentence (2b) is derived from (2a) by extraposing the relative clause that Mary gave for him, part of the object NP in (2a), to the right of the adverb yesterday. Note that the predicative AP extremely angry and upset is subjectoriented. Sentences involving Extraposition from NP and predication relation do not always result in acceptability, as will be shown later on, and I demonstrate that the functional constraint I propose in Section 2 for Extraposition from NP is applicable to the account of this sentence pattern as well.

2.

A Functional Explanation

2.1 More/Less Important Information Observe first the following pair of sentences: (3)

a. A man came yesterday with blue eyes. b. * A man came by taxi with blue eyes.

In (3a), the PP with blue eyes, part of the subject NP a man with blue eyes, can be extraposed to sentence-final position, crossing over a temporal adverb like yesterday. In contrast, in (3b) Extraposition of the PP crossing over an adverbial denoting means like by taxi is not allowed. A similar contrast is observed in the following pair of sentences: (4)

a. John drove a car in London with a sunroof. b. *John drove a car carefully with a sunroof.

In (4a) the PP with a sunroof, part of the object NP a car with a sunroof, can be extraposed over a locative adverbial like in London, while in (4b) it cannot be

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

25

extraposed over a manner adverb like carefully. Sentences (3) and (4) show that the difference in adverbial expressions plays a part in deciding the acceptability of Extraposition from NP. Furthermore, temporal and locative adverbials behave differently from other adverbials such as those expressing means and manner; the former can be easily preposed to sentence-initial position as thematic adverbials, while the latter cannot:2 (5)

a. b.

Yesterday, a man with blue eyes came. (cf. 3a) In London, John drove a car with a sunroof. (cf. 4a)

(6)

a. ?*By taxi, a man with blue eyes came. (cf. 3b) b. ?*Carefully, John drove a car with a sunroof. (cf. 4b)

Since sentence-initial position is generally assumed to be a topic position of the sentence, the difference in acceptability between (5) and (6) shows, as has been argued in the literature (e.g., Kuno 1975), that temporal and locative adverbials can convey thematic or “unimportant” information, whereas adverbials denoting means and manner convey rhematic or “important” information. Then, we can say that while Extraposition crossing over an element carrying unimportant (given) information results in acceptability, Extraposition crossing over an element carrying important (new) information results in unacceptability. Of course, temporal and locative adverbials do convey important information when they are stressed. This should imply that the acceptable instances of Extraposition in (3a) and (4a) turn out to be unacceptable when yesterday and London are stressed.3 This prediction is borne out: (7)

a. ??/*A man came YESTERDAY with blue eyes. (cf. 3a) b. ??/*John drove a car in LONDON with a sunroof. (cf. 4a)

It should further hold that adverbial complements such as those italicized in (8a, b) will have the effect of precluding Extraposition, even though they are temporal or locative adverbials. This is because (i) they are obligatory elements in the sentences, (ii) they cannot take sentence-initial thematic position, as seen in (9a, b), and therefore (iii) they are construed as conveying more important information than the rest of the sentences. (8)

a. The meeting lasted for five hours. b. John put a book on the dining room table.

26

KEN-ICHI TAKAMI (9)

a. * For five hours, the meeting lasted. b. * On the dining room table, John put a book.4

The above prediction is corroborated by the unacceptability of the following sentences: (10) a. ??/*A meeting lasted for five hours about the strike. b. ??/*John put a book on the dining room table about linguistics. In (10a) the PP about the strike, part of the subject NP a meeting about the strike, cannot be extraposed over the adverbial complement for five hours. Similarly, in (10b) the PP about linguistics, part of the object NP a book about linguistics, cannot be extraposed over the adverbial complement on the dining room table. In addition to the preposability observed above, the differences in the discourse functions of the adverbials mentioned above have a number of other syntactic and semantic manifestations. For instance, when (11a-c) below are reformulated as interrogative sentences, as in (12a-c), the focus of question is interpreted differently: (11) a. John drove [a car with a sunroof] in London. (cf. 4a) b. John drove [a car with a sunroof] carefully. (cf. 4b) c. John put [a book about linguistics] on the dining room table. (cf. 10b) (12) a. Did John drive [a car with a sunroof] in London? b. Did John drive [a car with a sunroof] carefully? c. Did John put [a book about linguistics] on the dining room table? Sentence (12a) is ambiguous. It can be interpreted either as a question asking whether or not it was a car with a sunroof that John drove in London, or as a question asking whether or not it was in London that John drove a car with a sunroof. In the former interpretation in London, serving as the topic of the sentence, conveys less important information than the rest of the sentence, particularly than a car with a sunroof. In the latter interpretation, however, it conveys more important information than the rest of the sentence, serving as the focus of the sentence. In contrast to (12a), (12b, c) are not ambiguous at all. They are interpreted only as questions asking whether the manner of John’s driving a car with a sunroof was careful or not, and whether the place

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

27

where John put a book about linguistics was on the dining room table or not. In other words, carefully and on the dining room table, serving the foci of the sentences, necessarily convey more important information than the rest of the sentences.5 From the above observations we can surmise that Extraposition from NP is only possible if it crosses elements conveying unimportant information. It is important to note here that reordering of constituents in a sentence generally takes place in such a way as to place those that represent unimportant (given) information closer to sentence-initial position, and those that represent important (new) information closer to sentence-final position (e.g., Kuno 1979, Quirk et al. 1985, Rochemont 1986). This reordering of constituents takes place in order to be consistent with the following Flow-of-Information Principle, a principle often argued in the framework of functional sentence perspective: (13) Flow-of-Information Principle: Elements in a sentence that does not contain an emphatic stress or morphologically-marked focus elements are ordinarily arranged in the order “less important information first and more important information last”. (Kuno 1995: 222) In light of the principle (13), it can be argued that Extraposition from NP is only possible if the element extraposed to sentence-final position conveys more important information than the rest of the sentence. As a subcondition of the general principle (13), let us hypothesize the following, which I call the More/Less Important Information Condition for Extraposition from NP, or the More/Less Important Information Condition, for short: (14) More/Less Important Information Condition for Extraposition from NP: Extraposition from NP is allowed only when the element extraposed to sentence-final position is interpreted as being more important than the rest of the sentence. As far as I know, the concept of “more/less important information” was first introduced explicitly in Kuno (1978, 1979, 1980) to account for the phenomenon of discourse deletion. Takami (1988, 1992) has extended the notion of “more/less important information” to the phenomenon of preposition stranding, and argued that the acceptability and unacceptability of the phenomenon are aptly captured by the concept of “more/less important

28

KEN-ICHI TAKAMI

information” rather than that of the more familiar “new/given information” or “focus/presupposition.” Throughout the following discussion on Extraposition from NP as well, I will employ the term “more/less important information” rather than “given/ new information” or “focus/presupposition.” There are two main reasons for this: (i) “New/given information” (or “focus/presupposition”) has generally been contextually established and is to that extent “discoursal,” as is clear from the following definition: (15) Newness of Information: An element in a given sentence in a given context conveys new information if it is not recoverable from the preceding context when garbled with noise. (Kuno 1983: 33) This indicates that an element that is interpreted as new information in a certain context can be interpreted as given information in another context, and vice versa. As an illustration of this, observe the following: (16) A1: Where did John drive a car with a sunroof? A2: What kind of car did John drive in London? A3: What did John drive in London? A4: What did John do in London? A5: What happened? B: John drove a car with a sunroof in London. (=11a) When (16B) is used in response to (16A1), the adverbial in London is new information (focus), in response to (16A2) the PP with a sunroof, in response to (16A3) the NP a car with a sunroof, in response to (16A4) the VP drove a car with a sunroof, and in response to (16A5) the whole S. Therefore it would be difficult to apply the term “given/new information” to the information structure of such isolated sentences as those observed so far. On the other hand, “more/less important information” is intended to capture not only the information structure of single sentences but also that of sentences dependent on context. (This will be made clear later in this section.) (ii) As I will demonstrate later, there are also cases in which the acceptability of the sentences under consideration is marginal (?/??), being neither perfect nor totally unacceptable. These cases seem to be characterized only by relative degrees of importance, hence the use of the concept “more/less important information.” With respect to “new/given information,” on the other hand, Chafe (1976: 33) argues that “it is necessary to say that . . . it has not been

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

29

demonstrated linguistically that given vs. new is anything more than a discrete dichotomy.” Therefore, it would be difficult for these terms to convey the concept that certain elements in a sentence may be more important than the remainder.6 The More/Less Important Information Condition can further account for the often noted fact that Extraposition from the subject NP is generally blocked if the VP contains a “true” object (Guéron 1980). Observe the following: (17) a. A man talked yesterday with blond hair. b. A man gave a talk yesterday with blond hair. c. * A man hit Mary with blond hair. d. * A book delighted Mary by Charles.7 (Guéron 1980: 663) (17a) is acceptable, just as (3a) is. (17b) is also acceptable even if the verb gave takes an object, but gave a talk simply means ‘talked’ as in (17a). On the other hand, (17c, d), in which hit/delighted Mary cannot be expressed by intransitive verbs, are totally unacceptable.8 This is attributable to the fact that a VP containing an object, with neutral declarative intonation, normally conveys more important information than the subject, as witnessed in (18) by the fact that nuclear stress normally falls on Mary, not on John:9 (18) John hit Mary. The More/Less Important Information Condition can account not only for the sentences observed so far but also for the following: (19) John read a book over the summer by Chomsky. (Guéron 1980: 637) (20) a. *Even John read a book over the summer by Chomsky. b. *John didn’t read a book over the summer by Chomsky. The acceptability of (19) can be explained in the same manner as that of (3a). It is important to note here that if the subject John is accompanied by an adverb like even (similarly only or alone) as in (20a), the resulting sentence becomes unacceptable. This is attributable to the fact that adverbs like even, only and alone draw both speaker’s and hearer’s attention to the element that they modify. Consequently in (20a), the subject even John is interpreted as being more important than the extraposed phrase by Chomsky. This is illustrated in the following sentence without Extraposition, by the fact that even John is construed as carrying the most important information.

30

KEN-ICHI TAKAMI (21) Even John read a book by Chomsky over the summer.

It is also important to note that if (19) is turned into a negative sentence, like (20b), the resulting sentence also becomes unacceptable. In a negative sentence, the speaker and hearer take particular notice of a negative element, and therefore in (20b) what is at issue is whether John read or didn’t read a book by Chomsky, as is clear from the following sentence without Extraposition from NP: (22) John didn’t read a book by Chomsky over the summer. Hence, it is difficult to interpret the extraposed by Chomsky as more important than didn’t read, and sentence (20b) is ruled out by the More/Less Important Information Condition.10 The More/Less Important Information Condition can also account for the following contrasting pair of sentences: (23) a. A man spoke yesterday with blond hair.  whispered b. * A man  grumbled  yesterday with blond hair.  yelled  The PP with blond hair can be extraposed crossing over a verb like spoke (and yesterday), as seen in (23a), but it cannot be extraposed crossing over the socalled non-bridge verbs like whisper and grumble, or a verb like yell, as seen in (23b).11 Note that the verbs in (23b) are semantically more complex than the verb speak, as partly witnessed by the fact that whisper means ‘to speak very low or under the breath’. It follows from this that in (23a) the speaker’s and hearer’s concern is what kind of man spoke yesterday, whereas in (23b) their concern is rather what a man with blond hair did yesterday. Therefore, in (23a) the extraposed phrase can be interpreted as being more important than spoke, whereas in (23b) whispered/grumbled/yelled are interpreted as being more important. Hence the difference in acceptability of these sentences. The acceptability of (23a) allows the extraposed phrase with blond hair to achieve its primary function of identifying what kind of man spoke yesterday. But in (23b) the speaker and hearer pay primary attention to what a man did because of the semantic richness of the verbs. Accordingly, it becomes difficult for the extraposed phrase, less important than the verb, to identify the subject a man. An interesting point to note here is that the unacceptable (23b) can be greatly improved. Observe the following, which are distinctly better than

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

31

(23b), if not perfect: (24) a. ? A man whispered/grumbled in the middle of the meeting with blond hair. b. ? A man yelled in the stadium with blond hair. Verbs such as whisper, grumble and yell carry important information if uttered in isolation, or with an adverb like yesterday, as we have just seen above. However, a man’s whispering/grumbling in the middle of a meeting, or a man’s yelling in a stadium is commonplace in daily life. Therefore, it would be easy to imagine (24a, b) appearing in contexts where what is at issue is what kind of man whispered/grumbled in the middle of the meeting, or what kind of man yelled in the stadium. In short, in (24a, b) the extraposed phrases are interpreted as being more important than the contents that the VPs express. Hence the (near) acceptability of the sentences. The More/Less Important Information Condition can also account for a cline of acceptability as shown in the following: (25) a. b. c.

I sat beside a man in the meeting from Los Angeles. ?I sat beside a man in a meeting from Los Angeles. ??I sat beside a man in a pay-raise meeting from Los Angeles.

(26) a. I talked with a man yesterday with a mustache. b. (?)I talked with a man several days ago with a mustache. c. ?/?? I talked with a man last Thursday with a mustache. d. ??/* I talked with a man one year and four months ago with a mustache. We have observed so far that temporal and locative adverbials do not generally convey more important information than extraposed elements, hence the acceptability of (25a) and (26a). However, as these adverbials become semantically “richer,” the acceptability of the resulting sentences decreases. This shows that the adverbials crossed by the sentence-final PPs, even if they are temporal and locative, tend to convey more important information as they become semantically heavier. Therefore the cline of acceptability demonstrated in (25) and (26) can be straightforwardly captured by the More/Less Important Information Condition, which resorts to the concept of “more/less important information.”

32

KEN-ICHI TAKAMI

2.2 Verbs and Context Observe first the following pairs of sentences without Extraposition: (27) a. John read a book by Chomsky yesterday. b. John tore up a book by Chomsky yesterday. (28) a. I bought a book on Italian cooking yesterday. b. I burned a book on Italian cooking yesterday. (29) a. Einstein formulated a theory about relativity in 1949. b. Einstein attacked a theory about dynamics in 1949. Sentences (27)-(29) are all acceptable. Now observe the following pairs of sentences, in which the PPs, part of the object NPs in (27)-(29), are extraposed: (30) a. b.

John read a book yesterday by Chomsky. ?/?? John tore up a book yesterday by Chomsky.

(31) a. b.

I bought a book yesterday on Italian cooking. ?/?? I burned a book yesterday on Italian cooking.

(32) a. Einstein formulated a theory in 1949 about relativity. b. √/?/?? Einstein attacked a theory in 1949 about dynamics. The (a) examples of (30)-(32) are acceptable. The acceptability judgments of the (b) examples fluctuate from speaker to speaker, but it is clear to all that they are distinctly worse than the (a) examples. It is intuitively felt that in the (a) examples reading a book, buying a book and formulating a theory are such usual, ordinary actions in our society that the speaker, with normal intonation, would not generally put any special emphasis on these parts of the sentences. The speaker’s attention (and also the hearer’s attention) would instead be put on the extraposed phrases, which are in consequence interpreted as being more important information. Hence, the (a) examples are predicted to be acceptable, in keeping with the More/Less Important Information Condition. The extraposed phrases can now successfully perform their functions of identifying what kind of book John read, what kind of book “I” bought and what kind of theory Einstein formulated. In the (b) examples, in contrast, tearing up a book, burning a book, and attacking a theory are so unusual and surprising under normal circumstances that the

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

33

speaker (and also the hearer) takes particular notice of these parts of the sentences. Therefore, they tend to be interpreted by many speakers as being more important than the extraposed phrases. Hence the low acceptability or unacceptability of the (b) examples. Note further that although the (b) examples of (30) and (31) are unacceptable, (32b) is nevertheless slightly better than these examples. This is attributable to the fact that in an academic context, attacking a particular theory is not so unusual and unexpected as tearing up or burning a book would be in everyday life. It is appropriate to address here an anticipated claim that the unacceptability of the (b) examples in (30)-(32) is attributable to the intrinsic properties of the verbs concerned, since verbs such as tear up and burn represent the eventual disappearance of the things that they take as the object. That is, it might be argued that use of these verbs makes the resulting sentence with Extraposition always unacceptable.12, 13 This claim, however, is soon proved to be wrong by the acceptability of the following (b) examples, in which verbs such as tear up and burn are used: (33) a. b.

?/??John tore up a book yesterday by Chomsky. (=30b) The lousy student of linguistics tore up a book yesterday by Chomsky.

(34) a. √/?/?? Einstein attacked a theory in 1949 about dynamics. (=32b) b. The critical physicist attacked a theory in 1949 about dynamics. (35) a. b.

?/??I burned a book yesterday on Italian cooking. (=31b) Recently I burned several of my books that I have no further use for. I burned a book yesterday on Italian cooking, and I burned some books today on African music.

The More/Less Important Information Condition can instead deal with the acceptability of the (b) examples above. To the extent that it is easily imaginable that a lousy student of linguistics tears up a book on linguistics, and that a critical physicist attacks a particular theory, these pieces of information are no longer unusual or surprising. Thus the speaker’s emphasis shifts to the extraposed elements, which in consequence are interpreted as being more important. Hence the acceptability of the (b) examples of (33) and (34), in keeping with the More/Less Important Information Condition. (35)

34

KEN-ICHI TAKAMI

shows that the acceptability of Extraposition from NP is contingent on context. The marginal or unacceptable (35a) turns out to be acceptable in an appropriate context, as in (35b). This is attributable to the fact that the action of burning a book, though unusual in isolation under normal circumstances, is already mentioned in the preceding discourse. Hence, in the sentence involving Extraposition, this piece of information is construed as conveying less important information than the extraposed phrase. The speaker’s and hearer’s attention focuses on what kind of book “I” burned, and in consequence the extraposed phrase is interpreted as being more important than the action done to the book. In the above examples of this subsection, I have considered the case of Extraposition from object NPs. The same appears to hold in the case of Extraposition from subject NPs, as well. Observe the following examples: (36) a. * A man died from India. (Guéron 1980: 653) b. Many soldiers were killed in the war from Germany and Japan. c. Many patients died in the hospital who had been infected with malaria. (37) Several visitors from foreign countries died in the terrible accident. A woman died from Peru and a man died from India.14 (Guéron 1980: 653) A man dying in (36a) is generally thought of as a serious matter. Therefore, this piece of information tends to be interpreted as being more important than the subject or the extraposed element, and the unacceptability results. On the other hand, (36b), an instance of PP Extraposition, and (36c), an instance of Relative Clause Extraposition, are perfectly acceptable. In these sentences, it is readily understood that soldiers are killed in a war, and that patients die in a hospital. What is at issue in (36b, c) is what nationality the soldiers were who were killed in the war, or what kind of patients died in the hospital. This shows that the extraposed elements convey more important information than the contents that the VPs represent. Hence the acceptability of (36b, c), in keeping with the More/Less Important Information Condition. The acceptability of (37) can be explained in exactly the same manner as observed in (35b). In (37) the scene of people dying is already set in the previous discourse, and in the sentence involving Extraposition, such information is no longer deemed to be unexpected. Therefore it is considered less important

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

35

information than the extraposed elements, and acceptability results.15 2.3 Double Object and Dative Constructions Observe the following sentences of the dative construction; Extraposition of PP has taken place in (39) out of NP in (38): (38) a. I gave [a book about African music] to Mary yesterday. b. I gave candies to [a child with big sad eyes] yesterday. (39) a. I gave a book to Mary yesterday [about African music]. b. I gave candies to a child yesterday [with big sad eyes]. Extraposition of PP is allowed either from the direct object, as shown in (39a), or from the indirect (dative) object, as shown in (39b). Now compare (39a, b) with the following double object construction involving Extraposition from NP: (40) a. I gave Mary a puppy yesterday [with cute floppy ears]. b. * I gave a man an interesting book yesterday [with green eyes]. While Extraposition from the direct object is allowed, as shown in (40a), that from the indirect object is disallowed, as shown in (40b). How can the contrast between (39) and (40) be accounted for? To solve this problem, it is first necessary to consider the derivation of the dative and double object constructions. It is commonly assumed in the literature that the dative construction is the underlying or basic form and that the double object construction is transformationally derived from the dative construction (see Culicover 1976, among others, for arguments for this assumption). With this in mind, observe the following pairs of sentences: (41) a. b.

John gave the book to a girl. John gave a book to the girl.

(42) a. John gave the girl a book. b. ?? John gave a girl the book. Sentences (41) are examples of the dative construction and (42) are of the double object construction. Each of the objects is marked either with the definite article (the) or with the indefinite article (a). This difference indicates that an object marked with the definite article conveys less important infor-

36

KEN-ICHI TAKAMI

mation since its referent is interpreted as known to the hearer, whereas an object marked with the indefinite article conveys more important information since its referent is generally interpreted as unknown to the hearer. Therefore we can represent the flow of information observed in sentences (41) and (42) as in (43) and (44), respectively: (43) a. b. (44) a.

John gave the book less important John gave a book more important

John gave the girl less important b. ?? John gave a girl more important

to a girl. more important to the girl. less important a book. more important the book. less important

(43a) and (44a) observe the Flow-of-Information Principle (13), i. e., “less important information first and more important information last”. Hence (43a) and (44a) are predicted to be acceptable. In contrast, (43b) and (44b) violate this principle, but in spite of this, (43b) is acceptable and (44b) is unacceptable. Kuno (1978, 1979, 1987, class lecture 1988) attributes this difference to the following principle: (45) Markedness Principle for Discourse-Rule Violations: Sentences that involve marked (or intentional) violations of discourse principles are unacceptable. On the other hand, sentences that involve unmarked (or unintentional) violations of discourse principles go unpenalized and are acceptable. With respect to the difference between (43b) and (44b), the above principle indicates that the violation of the Flow-of-Information Principle observed in (43b) is unmarked or unintentional because the dative construction is the basic form and therefore the arrangement of each constituent is automatically decided. Thus the violation goes unpenalized and the sentence is acceptable. On the other hand, the violation of the Flow-of-Information Principle observed in (44b) is marked or intentional because the double object construction is the derived form and therefore the speaker can arrange the information structure by using the basic form, as in (43a). Nevertheless, he/she has intentionally used the derived form in (44b), which results in a violation of the

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

37

Flow-of-Information Principle. Hence the sentence is predicted to be unacceptable. Bearing the above prerequisite in mind, let us now turn to the double object and dative constructions involving Extraposition from NP. First, observe (38) and (39), repeated below: (38) a. I gave [a book about African music] to Mary yesterday. b. I gave candies to [a child with big sad eyes] yesterday. (39) a. I gave a book to Mary yesterday [about African music]. b. I gave candies to a child yesterday [with big sad eyes]. (39a) derives from (38a), and (39b) from (38b). The information structure in (38a) can be represented as in (46): (46) I gave a book about African music to Mary yesterday. more important less important In (46) a book about African music is interpreted as being more important than the proper noun Mary, whose referent can be easily identified by the hearer. Therefore (46) violates the flow of the “from less important to more important” information. However, this violation, as observed above, is unintentional; hence the acceptability of the sentence. Next, the NP a book about African music, more important than Mary, can be divided into two constituents, the NP a book and the PP about African music. Sentence (39a), derived from (47) via Extraposition of the PP about African music, indicates that the extraposed element is more important than a book. In other words, it can be said that the information conveyed by “I gave a book to Mary yesterday” is understood or presupposed between the speaker and the hearer, and that what is at issue here is what kind of book “I” gave to her. Thus the flow of information in (39a) can be represented as in the following: (47) I gave a book about African music to Mary yesterday. more important less important less important more important Extraposition The extraposition of about African music conforms to the More/Less Important Information Condition proposed in Section 2.1 because it conveys the

38

KEN-ICHI TAKAMI

most important information in the sentence. Let us now examine the flow of information in (38b), repeated below as (48): (48) I gave candies to [a child with big sad eyes] yesterday. Here we cannot say which element, candies or a child with big sad eyes, is more important because both elements are indefinite expressions. However, whichever element is more important does not matter since (48) is the underlying or basic form. Therefore, even if candies is assumed to convey more important information than a child with big sad eyes, this violation is not intentional. On the other hand, if a child with big sad eyes is assumed to convey more important information than candies, this information structure observes the Flow-of-Information Principle. Consider sentence (39b), which involves Extraposition from NP. Judging from the fact that with big sad eyes is extraposed to sentence-final position, it is reasonable to suppose that a child with big sad eyes conveys more important information than candies. The information structure can be represented as follows: (49) I gave candies to a child with big sad eyes yesterday. less important more important less important

more important

The PP with big sad eyes conveys the most important information in the sentence and can therefore be extraposed in keeping with the More/Less Important Information Condition. (39b) indicates that the information “I gave candies to a child yesterday” is understood between the speaker and the hearer and that what is at issue here is what kind of child “I” gave them to. Next let us consider (40a, b), repeated here as (50a, b): (50) a. I gave Mary a puppy yesterday [with cute floppy ears]. b. *I gave a man an interesting book yesterday [with green eyes]. The underlying sentences for (50a, b) are (51a, b), respectively: (51) a. I gave [a puppy with cute floppy ears] to Mary yesterday. b. I gave an interesting book to [a man with green eyes] yesterday.

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

39

From (51a, b) the double object construction is produced, as shown in (52a, b), respectively: (52) a. I gave Mary [a puppy with cute floppy ears] yesterday. b. I gave [a man with green eyes] an interesting book yesterday. In (52a) Mary is understood as being less important than a puppy with cute floppy ears, thereby satisfying the Flow-of-Information Principle. On comparing a puppy and with cute floppy ears, the latter can be interpreted as being more important. In (50a) the most important information in the sentence, with cute floppy ears, is extraposed and this extraposition is marked acceptable, in keeping with the More/Less Important Information Condition. Turning to (52b), it may appear difficult to decide which constituent, a man with green eyes or an interesting book, is more important. However, given the fact that the double object construction is intentionally used here so that the resulting sentence can meet the Flow-of-Information Principle, it can be concluded that an interesting book is more important than a man with green eyes, since the former is placed closer to sentence-final position. Hence we can represent the information structure of (52b) as follows: (53) I gave [a man with green eyes] [an interesting book] yesterday. less important more important It is true that between a man and with green eyes the latter is more important. But the extraposition of with green eyes violates the More/Less Important Information Condition because it is less important than an interesting book. Hence (50b) is correctly ruled out. To put this point differently, Extraposition is blocked from an NP that has already been placed closer to sentence-initial position via a certain operation, because the NP is already regarded as less important information than other elements that are placed closer to sentencefinal position. The above point is further confirmed by the fact that Heavy NP Shift cannot apply to the indirect object in the double object construction.16 Observe the following sentences: (54) a. *I gave ei an interesting book [a man with green eyes]i. (cf. 53) b. *John bought ei a car with a sunroof [his brother-in-law]i. The underlying sentences for (54a, b) are the following:

40

KEN-ICHI TAKAMI (55) a. I gave an interesting book to a man with green eyes. b. John bought a car with a sunroof for his brother-in-law.

From (55a, b) the double object construction will be derived, so that a man with green eyes and his brother-in-law can be interpreted as being less important information than an interesting book and a car with a sunroof. The information structures of the resulting sentences are the following: (56) a. I gave a man with green eyes an interesting book. less important more important b. John bought his brother-in-law a car with a sunroof. less important more important It is commonly held in the literature that an element Heavy NP Shift applies to conveys the focus (most important information) of the sentence. However, since a man with green eyes and his brother-in-law do not convey the most important information as a result of being intentionally displaced from sentence-final position, they cannot be Heavy NP-shifted. Hence the unacceptability of (54a, b). Finally, it should be noted that the acceptable (39a), repeated below as (57a), becomes worse when the less important information Mary is stressed, or is replaced by an unidentified, semantically heavier, NP: (57) a.

I gave a book to Mary yesterday [about African music]. (=39a) b. ??/*I gave a book to MARY yesterday [about African music]. c. (?) I gave a book to a friend of mine yesterday [about African music]. d. ?? I gave a book to a nice-looking girl student yesterday [about African music].

It is clear that the marginality or unacceptability of (57b-d) is straightforwardly explained by the More/Less Important Information Condition (see 25 and 26). In this subsection I have demonstrated that the More/Less Important Information Condition, coupled with the Markedness Principle for DiscourseRule Violations proposed by Kuno, can account for the phenomenon of Extraposition from NP in sentences involving dative and double object constructions.

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP 3.

41

The Semantic Analyses of Guéron (1980) and Rochemont and Culicover (1990)

This section examines how Guéron (1980) and Rochemont and Culicover (1990) account for the difference in acceptability of sentences such as those given below, and then demonstrates that their accounts run into a number of major problems: (58) a. *A man died from India. (=36a) b. Several visitors from foreign countries died in the terrible accident. A woman died from Peru and a man died from India. (=37) (59) a. *A man screamed who wasn’t wearing any clothes. b. Suddenly there was the sound of lions growling. Several women screamed. Then a man screamed who was standing at the very edge of the crowd. ((59a, b) are from Rochemont and Culicover (1990: 65)) Let us first outline Guéron’s (1980) account. Guéron argues that elements extraposed from subject, but not those from object, are subject to a semantic restriction concerning the class of verbs/predicates that tolerate them. The restriction she offers is the following, in which SX stands for phrases extraposed from subject position: (60) Guéron’s Predicate Restriction on SX: SX is subject to the restriction that the predicate must denote, essentially, the appearance of the subject in the world of the discourse. Guéron says that died in (58a), unlike verbs like appear, arrive, come and so on, does not denote the appearance of the subject, hence the unacceptability of the sentence. Likewise, screamed in (59a) is not a verb of appearance, either, hence the unacceptability of the sentence. She also says, concerning the acceptability of a discourse such as (58b), that “Ss which are unacceptable in isolation become acceptable in a context in which the verb is pragmatically emptied of all semantic content beyond that of appearance in the world of the discourse” (pp. 653-654). This is perhaps intended to mean that, for example, the verbs died and screamed used in the sentences involving Extraposition in (58b) and (59b) convey nothing but the meaning of appearance, because they have already been used in the prior contexts.

42

KEN-ICHI TAKAMI

Guéron’s account of the acceptability of (58b) and (59b), however, is quite nebulous. It is counter-intuitive to say that died and screamed in the sentences with Extraposition are “emptied of all semantic content beyond that of appearance in the world of the discourse,” because, even if they are used in the prior contexts, they do convey the meanings of ‘die’ and ‘scream’. For the sake of argument, however, let us assume that this is the case, and examine next Rochemont and Culicover’s analysis. Rochemont and Culicover (1990) maintain that phrases extraposed from subject NP must observe a semantic restriction concerning the class of verbs/ predicates, and they state, along with Guéron (1980), that there is no parallel requirement on phrases extraposed from object NP. The restriction they offer is as follows: (61) Rochemont and Culicover’s Predicate Restriction on SX: SX is subject to the restriction that the predicate must be c(ontext)-construable, whether directly or indirectly. “Directly c-construable” is defined as in (62), and “indirectly c-construable,” divided into two subcases (i.e., lexically specified indirectly c-construable and pragmatically determined indirectly c-construable), is defined as in (63a, b) (see Rochemont 1986): (62) An expression is directly c-construable if it is “under discussion,” that is, if it is directly retrievable from some ongoing discourse. (63) a. An expression is indirectly c-construable if it is a member of verbs of appearance, or of the so-called indexical expressions like the personal pronouns I, you, we, locative and temporal adverbs such as here, there, now, then, today, tomorrow, last night, and so on. [lexically specified] b. An expression is indirectly c-construable if the speaker takes it not to be noteworthy in view of his audience, who he anticipates will readily identify it as an unremarkable scenesetter for the topic of speaker’s discourse. [pragmatically determined] The definition of directly c-construable (62) and that of lexically specified indirectly c-construable (63a) are clear enough, but the definition of pragmatically determined indirectly c-construable (63b) seems to require further explanation. Observe the following examples, cited from Rochemont (1986: 57), in which upper case signals the required locations of sentence accents:

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

43

(64) a. I ran into JOHN in the HOSPITAL this morning. b. I ran into JOHN in the hospital this morning. Either of the sentences in (64a, b) may be used to initiate a new discourse topic. For instance, suppose that Speaker A and Hearer B live together, and during their habitual dinner conversation A says (64a, b) to B. Since it is assumed that no mention has been made up to this point in the conversation about the hospital, it is naturally to be expected that (64a), where hospital is stressed, may be used to initiate a new discourse topic. However, if A works in a hospital, not only (64a) but also (64b) would be appropriate. In other words, “A can legitimately take the hospital to be indirectly c-construable, under the reasonable assumption that B knows that A works in a hospital (“reasonable,” because they are housemates). Similarly, if B is aware that A has been to the hospital that day, or that A goes to the hospital regularly for treatment or the like, then A can also appropriately take this location to be a suitable, pragmatically established, indirectly c-construable scenesetter” (Rochemont 1986: 57). On the other hand, in the absence of such background information, A will naturally use (64a) rather than (64b), and the hospital is not indirectly c-construable. From this observation it should be noted that an expression that is indirectly c-construable for the pragmatically determined reason is highly dependent on the speaker’s/hearer’s shared background knowledge. Now that I have made clear what c-construable means, let us turn to observe how Rochemont and Culicover’s Predicate Restriction on SX accounts for sentences such as (58) and (59). In (58a) and (59a) the predicates died and screamed are neither directly nor indirectly c-construable, because (i) each of them is introduced into the discourse for the first time (see 62), (ii) it is neither a member of verbs of appearance nor of the so-called indexical expressions (see 63a), and (iii) these two sentences are uttered without any assumption about the speaker’s/hearer’s shared background information (see 63b). Hence, (58a) and (59a) violate the Predicate Restriction on SX, and unacceptability results. In (58b) and (59b), on the other hand, died and screamed in the sentences involving Extraposition from NP are directly cconstruable because they are directly retrievable from the prior discourses (see 62). Hence, (58b) and (59b) observe the Predicate Restriction on SX, and acceptability results. Guéron’s and Rochemont and Culicover’s analyses, though ingenious, encounter a number of problems. First, their analyses fail to explain the

44

KEN-ICHI TAKAMI

acceptability of the following sentences observed in Section 2: (36) b. Many soldiers were killed in the war from Germany and Japan. c. Many patients died in the hospital who had been infected with malaria. In addition to the fact that were killed and died in (36b, c) are not verbs of appearance, they are introduced in these sentences for the first time. Therefore it is clear that they are not emptied of their semantic content, and that they are neither directly nor indirectly c-construable. It must be concluded then that Guéron’s and Rochemont and Culicover’s frameworks fail to capture the acceptability of (36b, c) in a consistent manner. Second, Guéron’s and Rochemont and Culicover’s analyses fail to account for the fact that there is a difference in acceptability even in cases where the same non-appearance verbs are employed. Observe (23b) and (24a, b), repeated here for ease of exposition: whispered  (23) b. *A man grumbled  yesterday with blond hair.  yelled (24) a. ?A man whispered/grumbled in the middle of the meeting with blond hair. b. ?A man yelled in the stadium with blond hair. Contrary to fact, Guéron’s and Rochemont and Culicover’s accounts predict that all these sentences are unacceptable. Third, Guéron’s and Rochemont and Culicover’s analyses, as we have seen above, are restricted to Extraposition from the subject NP only, and therefore cannot apply to sentences involving Extraposition from the object NP. Observe, for instance, sentences (30) and (31), repeated below: (30) a. John read a book yesterday by Chomsky. b. ?/??John tore up a book yesterday by Chomsky. (31) a. I bought a book yesterday on Italian cooking. b. ?/??I burned a book yesterday on Italian cooking. In the above sentences, none of the verbs are verbs of appearance, and yet there is a clear difference in acceptability between the (a) and the (b) examples.

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

45

Although there are other sentences discussed in this paper that Guéron’s and Rochemont and Culicover’s analyses cannot deal with, it suffices to say that their accounts are not comprehensive enough to capture a wide range of examples involving Extraposition from NP.

4. Extraposition from NP over Secondary Predicates This section is concerned with sentences involving not only Extraposition from NP but also predication relation, and it attempts to demonstrate that the More/Less Important Information Condition, proposed in Section 2, can also explain the acceptability of the sentence pattern under consideration. I also review Rochemont and Culicover’s (1990) account of the same phenomenon, and show that my analysis is preferable to theirs, pointing out some serious counterexamples to theirs. Observe first the following sentence: (65) John left the party [that Mary gave for him] yesterday [extremely angry and upset]. (=2a) In (65) the object contains a relative clause, and the subject John is semantically related to the subject-oriented AP extremely angry and upset. The relative clause that Mary gave for him can be extraposed and, in theory, be placed either immediately right after the adverb yesterday or at the right periphery of the sentence, producing the following sentences, respectively: (66) a. John left the party yesterday [that Mary gave for him] [extremely angry and upset]. b. *John left the party yesterday [extremely angry and upset] [that Mary gave for him]. Interestingly enough, while (66a), in which the relative clause precedes the subject-oriented AP, is acceptable, (66b), in which the relative clause follows the subject-oriented AP, is unacceptable for most speakers.17 How is this contrast to be accounted for?18 An application of the More/Less Important Information Condition, proposed in Section 2, seems to capture the difference. As I have already remarked, it is generally held that reordering of constituents in a sentence takes place in such a way as to place those that represent less important

46

KEN-ICHI TAKAMI

information closer to sentence-initial position, and those that represent more important information closer to sentence-final position (see the Flow-ofInformation Principle (13)). Long and complex elements extraposed to sentence-final position represent more important information than the rest of the sentence, and Extraposition from NP applies to such elements to produce a sentence that observes the flow of “from less important to more important” information (cf. Kuno 1979, Quirk et al. 1985: 1359). What this means in connection with the sentence pattern in (66a, b) is that Extraposition from NP can cross over a secondary predicate only when the “V+NP+AP” sequence represents a concept that is conventionally or contextually known. Conversely, if the “V+NP+AP” sequence represents a concept that is conventionally or contextually unknown, Extraposition from NP can take place only to the extent that it does not cross the secondary predicate. Therefore the information structure of the construction under discussion must essentially meet the Flow-of-Information Principle, or its subcondition, the More/Less Important Information Condition. Let us now hypothesize the following condtion for the construction in question: (67) More/Less Important Information Condition on Extraposition from NP for the “V+NP+AP” Pattern: Extraposition from NP over a secondary predicate can apply only when the “V+NP+AP” sequence represents less important information than the extraposed element. Bearing the above condition in mind, let us examine (66a, b). The concept of “leaving a party angry and upset” is not commonplace in ordinary life. Therefore, one would normally infer that what is at issue in (66a, b) is not what kind of party John left extremely angry and upset, but how John left the party that Mary gave for him. In short, the secondary predicate extremely angry and upset is likely to be interpreted as being more important than the extraposed clause. Hence the More/Less Important Information Condition on Extraposition from NP for the “V+NP+AP” Pattern (henceforth, the More/ Less Important Information Condition) can distinguish between the acceptable (66a) and the unacceptable (66b). Likewise, observe the following pairs of sentences: (68) a. Max discussed the topic yesterday [that he had in mind] [completely nude]. b. *Max discussed the topic yesterday [completely nude] [that he had in mind].

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

47

(69) a. John ate an apple yesterday [with a rotten core] [completely naked]. b. *John ate an apple yesterday [completely naked] [with a rotten core]. ((69b) is adapted from Johnson (1986: 119)) The concept of “discussing/eating something without wearing any clothes” is so uncommon and extraordinary that in (68) and (69) the speaker’s/hearer’s attention focuses on how Max discussed the topic or on how John ate an apple, rather than on what kind of topic he discussed or what kind of apple he ate. Hence the above difference in acceptability is accounted for by the More/ Less Important Information Condition.19 To explain the difference in acceptability of the sentence pattern in question, Rochemont and Culicover (1990) propose the “interpretive nesting requirement” (INR), which allows for the nesting relation, but not the crossing relation, of two interpretive links (see Fodor 1978, Pesetsky 1982). The extraposition link and the predication link, both of which are assumed in their position to be interpretive links, are nested in the acceptable (66a), (68a) and (69a), but are crossing in the unacceptable (66b), (68b) and (69b), as diagrammed in (70a, b): (70) a. NP

[. . . e ] [CP . . . ] AP (=66a, 68a, 69a)

b. *NP [ . . . e] AP

[CP . . . ]

(=66b, 68b, 69b)

As far as the above examples are concerned, therefore, the INR can account for the difference in acceptability of the sentence pattern under consideration. The INR, while ingenious, faces many counterexamples, however. Observe the following: (71) a. John left the party [dead drunk] [that Mary gave for him on his

birthday]. b. Mike never leaves parties [sober] [that he enjoys].

48

KEN-ICHI TAKAMI

In (71a, b) the extraposition link crosses the predication link, in violation of the INR, but the sentences are acceptable or nearly so for most speakers. Observe further the following examples: (72) *A man talked half-nude with a mustache.

(cf. A man talked with a mustache half-nude.) (73) *The gardener watered the tulips [completely flat to the ground]

yesterday [that he had been growing carefully]. (cf. The gardener watered the tulips yesterday [that he had been growing carefully] [completely flat to the ground]. ) In (72) half-nude is a subject-oriented predicative AP, and with a mustache is extraposed from the subject a man. The predication link and the extraposition link are nested, and yet the sentence is unacceptable. In (73) completely flat to the ground is an object-oriented predicative AP, and that he had been growing carefully is extraposed from the object the tulips. The predication link and the extraposition link are again nested, but the sentence is unacceptable. Hence the INR fails to account for the unacceptability of (72) and (73).20 The More/Less Important Information Condition, however, can account for the above sentences as well. Observe, first, sentences (71a, b). The concept of “leaving a party drunk/sober,” unlike those of “leaving a party angry” and “discussing/eating something without wearing any clothes,” would be normal in everyday life. Therefore, it is likely that the “V+NP+AP” sequence in (71a, b) conveys less important information than the extraposed clauses. Hence the acceptability of the sentences. In (72), as we have already observed above, talking half-nude is so surprising and unexpected under normal circumstances that the “V + AP” sequence is likely to be interpreted as conveying more important information than the extraposed phrase. Hence the unacceptability of (72). Similarly, in (73) the concept of “watering flowers flat” is unusual, and especially draws the speaker’s/hearer’s attention, which tends to make the “V+NP+AP” sequence more important than the extraposed clause. Hence the sentence is unacceptable. Note here that (72) and (73) turn

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

49

out to be acceptable if the order of the predicative AP and the extraposed element is reversed. This is also accounted for by the More/Less Important Information Condition. The above argument that a discourse factor, not a syntactic factor, is responsible for the contrast in acceptability of the sentence pattern under consideration is further supported by the fact that sentences of the pattern in (66b), for instance, become acceptable if placed in contexts that would make the “V+NP+AP” sequence contextually known. For example, observe the following discourse: (74) I don’t know why, but Thomas often gets offended at parties he goes to and leaves them angry at the hosts. Only two nights ago, he left a party angry which was given in his honor, because. . . (cf. Kuno and Takami 1993: 133) Observe that the “leave a party angry” pattern is perfectly acceptable in (74), where the concept of “leaving parties angry” has been contextually established. In the sentence involving Extraposition, therefore, the extraposed clause is naturally interpreted as being more important than the secondary predicate. Hence the acceptability of (74) results.

5.

Concluding Remarks

I have demonstrated in this paper that Extraposition from NP is highly sensitive to the flow of information in a sentence. Accordingly, I have proposed a functional constraint called the More/Less Important Information Condition, and demonstrated that it can account for the acceptability vis-à-vis the unacceptability of a wide range of sentences involving Extraposition from NP. It has also been shown that the condition can be applied to sentences involving not only Extraposition from NP but also predication relation. The account based on the concept of “more/less important information,” it is further shown, is more comprehensive than alternative accounts, such as those proposed by Guéron (1980) and Rochemont and Culicover (1990).

50

KEN-ICHI TAKAMI

Notes 1.

There are, however, some linguists who do not consider a sentence like (1b) as being transformationally derived from a sentence like (1a). For example, Rochemont and Culicover (1990) and Culicover and Rochemont (1990) maintain that Extraposition does not involve a movement operation at all, and that “extraposed” phrases are base-generated in place. They propose a syntactic constraint called the Complement Principle to relate an “extraposed” phrase to its associated NP. See Takami (1990) for a number of problems with this principle.

2.

Sentences (6a, b) become acceptable if the indefinite articles are made definite (see Takami 1992: 19-20), if they are uttered in a richer context, or if they are stressed. Observe the following, for which I am indebted to Ben Fortson (personal communication, 1991): (i) a. By TAXI, the man with the blue eyes then proceeded to the Hilton, where he met with the board. b. CAREFULLY(,) John maneuvered the car with the sunroof out of the driveway.

3.

I am indebted to Susumu Kuno (personal communication, 1991) for this observation.

4.

Note that (9a, b) are worse than (6a, b) in acceptability. This indicates that an adverbial complement is more tightly connected to the verb of the sentence than a pure adjunct is.

5.

See Takami (1992: chapter 2, 1995: chapter 3) for further differences between adverbials denoting time and place on the one hand and those denoting means, manner, accompaniment and so on on the other.

6.

Furthermore, there are a number of different definitions of new/given information in the literature, as succinctly summarized in Prince (1981), so that the use of new/given information might cause unnecessary confusion.

7.

Johnson (1986) claims that Extraposition from the subject NP in a sentence involving a psych-verb like delight, impress, annoy and so on is acceptable or nearly so. However, all the native speakers with whom I checked this type of sentence found it totally unacceptable, as is indicated by Guéron’s example (17d). For further problems with Johnson’s analysis, see Takami and Kuno (1992).

8.

Observe the following sentence, cited from Guéron (1980: 664): (i) A book hit the newsstand by Chomsky. Guéron states that hit the newsstand in (i) is interpreted only in the figurative sense of ‘appear,’ and that if it were to be interpreted literally, the sentence would become unacceptable.

9.

It is interesting to observe that, unlike (17c, d), the following sentences are much better, in spite of the fact that the VP contains an object: (i) a. √/?A man just phoned me with a funny name. b. √/?A man just phoned me from my hometown, saying he had just arrived in town. (Takami and Kuno 1992: 164) The (near) acceptability of (ia, b) is attributable to the fact that the direct object is the personal pronoun, me, the speaker, and that the VP just phoned me denotes the appearance of the subject a man into the present discourse. Hence, the VP just phoned me does

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

51

not carry information as important as the VPs hit Mary and delighted Mary in (17c, d). For further details, see Takami and Kuno (1992). 10.

Observe that the following sentences are acceptable in spite of the fact that negation is also involved: (i) a. John read no books over the summer by Chomsky. b. John read few books over the summer by Chomsky. c. John didn’t read any books over the summer by Chomsky. In comparing (20b) with (ia-c), it becomes clear that while in (20b) the negative morpheme negates the verb, in (ia-c) it negates the noun phrase. Therefore, in the former didn’t read is the focus of the sentence, carrying more important information than the extraposed element by Chomsky, hence the sentence is unacceptable. In the latter, on the other hand, the NPs no books by Chomsky, few books by Chomsky, and so on are the foci of the sentences before Extraposition from NP has taken place, and after Extraposition only part of these NPs, namely by Chomsky, becomes the focus of the sentence, carrying more important information than the rest of the sentence. Hence, (ia-c) are predicted to be acceptable, observing the More/Less Important Information Condition. (I am indebted to Susumu Kuno (personal communication, 1991) for this observation.)

11.

The reader may recall a contrast like the following, in which the movement involved is leftward, not rightward: (i) a. What did John say that Mary stole e? b. *What did John whisper that Mary stole e? A bridge verb like say in (ia) allows an element in an embedded clause to be extracted, but a non-bridge verb like whisper in (ib) does not (see Erteschik-Shir 1973, among others). It is interesting to note here that the extraction observed in (ia) becomes impossible if the bridge verb say is accompanied by a manner adverb, as shown in (ii) (Kennedy 1989): (ii) *What did John say abruptly that Mary stole e?

12.

The verb attack in (32b), unlike the other verbs, does not seem to directly represent the eventual disappearance of the thing that it takes as the object, and therefore the anticipated claim mentioned in the text may not be applicable to this case.

13.

Chomsky (1977) and Koster (1978), among others, observe a difference in acceptability of sentences involving preposition stranding, as seen in (ia, b), and offer a similar argument: (i) a. Who did you see a picture of e? b. *Who did you destroy a picture of e? Chomsky and Koster claim that the sentence pattern in (ia, b) always results in unacceptability when it contains a verb like destroy. However, this argument is proved to be false by the acceptability of sentences such as (iia, b): (ii) a. Which actress did the lunatic destroy a picture of e? b. Which river did the flood destroy the banks of e? For further details, see Takami (1992).

14.

Karen Courtenay (personal communication, 1992) says that PP Extraposition demonstrated in (37) is weird, and that Relative Clause Extraposition, given below, sounds much better: (i) Several visitors from foreign countries died in the terrible accident. A woman died who was from Peru and a man died who was from India. Akio Kamio (personal communication, 1993) points out that the weirdness of (37) that

52

KEN-ICHI TAKAMI Karen Courtenay observes may arise from the string died from Peru, which is close in meaning to died of Peru, a nonsense.

15.

We will observe Guéron’s account of the acceptability of (37) in Section 3, and point out its problems there.

16.

I am indebted to Susumu Kuno (personal communication, 1991) for this observation.

17.

As Kuno (1974) argues, the sentence-internal bare that-clause generally results in unacceptability. Observe the following: (i) a. It is likely [that John will succeed]. b. [That John will succeed] is likely. c. *Is [that John will succeed] likely? (ii) a. I believe John to be honest. b. *I believe [that John will succeed] to be likely. In (ia) the that-clause is sentence-final and in (ib) it is sentence-initial. In contrast, in (ic) it is sentence-internal; hence unacceptability. The unacceptability of (iib) is accounted for in the same manner. In spite of this fact, note that (66a), in which the relative thatclause is sentence-internal, is acceptable, whereas (66b), in which it is sentence-final, is unacceptable.

18.

It is generally assumed in the literature (e.g., Baltin 1981, 1983, Guéron 1980, Nakajima 1989) that an element extraposed from the object NP is adjoined to VP. With respect to the syntactic position that subject-oriented predicative APs occupy, Williams (1980), Rothstein (1983) and Nakajima (1990a) argue that they are dominated by IP, while Culicover and Wilkins (1984), Roberts (1988), McNulty (1989) and Napoli (1989) argue that they are dominated by VP. If it is assumed here that they are dominated by IP, the contrast between (66a) and (66b) is straightforwardly explained by a constraint banning crossing branches; the structure for (66a) does not involve any crossing branches, while the structure for (66b) does, because extremely angry and upset is dominated by IP and the extraposed clause following the AP is dominated by VP. However, although this sort of syntactic account seems to be theoretically interesting, the phenomenon under consideration, as will be shown later, presents recalcitrant data for such a syntactic analysis (see 71a, b, for instance).

19.

A condition similar to (67) seems to be applicable to Heavy NP Shift from a sentence containing a predicative AP. Observe the following contrast: (i) a. *Max discussed ei nude [the recent Broadway musical by Sondheim]i. (Larson 1988) b. John left ei dead drunk [the party Mary gave for him on his birthday]i. The concept of ‘discussing something nude’ is, as observed in the text, not commonplace in society. Therefore, it is difficult to imagine that (ia) has been uttered in a context in which what is at issue is not whether Max was nude when he discussed the musical, but what Max discussed in the nude. On the other hand, in (ib) the concept of ‘leaving a party drunk’ is commonplace in society. Therefore, it is likely that the ‘V + AP’ sequence in the sentence conveys less important information than the postposed object. See Kuno and Takami (1993: Chapter 4) for more examples of this pattern and the explanation along this line.

20.

Nakajima (1990b) also points out a number of serious problems with the INR. For instance, he presents the following counterexamples to the INR:

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

53

(i) a. Mary gave the coffee to students [hot and black] [who would usually drink coffee cold and white]. b. John took the puppy from his friend [yelping and squirming] [who John knew never could stand animals]. In (ia, b) the secondary predicates are direct object-oriented depictive APs, and Extraposition is triggered from the prepositional object. Here, again, the extraposition link crosses the predication link, and therefore, the INR incorrectly predicts (ia, b) to be unacceptable. In fact, the sentence pattern in (ia, b) is highly productive, and similarly acceptable sentences, albeit a little clumsy in style, could be easily constructed, as in (iia, b): (ii) a. Mary sent vegetables to friends [fresh from her garden] [whom she became acquainted with in Boston]. b. John offered bottles of milk to his guests [fresh from the dairy] [who were visiting from New York].

References Baker, Carol L. and John McCarthy (eds) 1981 The logical problems of language acquisition. Cambridge, MA: MIT Press. Baltin, Mark R. 1981 “Strict bounding”. In C. L. Baker and J. McCarthy (eds), 257-295. Baltin, Mark R. 1983 “Extraposition: Bounding versus government-binding”. Linguistic Inquiry 14: 155-162. Bedell, George, Eichi Kobayashi, and Masatake Muraki (eds) 1979 Explorations in linguistics: Papers in honor of Kazuko Inoue. Tokyo: Kenkyusha. Chafe, Wallace L. 1976 “Givenness, contrastiveness, definiteness, subjects, topics, and point of view”. In C. Li (ed), 25-55. Chomsky, Noam 1977 “On wh-movement”. In P. Culicover, T. Wasow, and A. Akmajian (eds), 71-132. Cole, Peter (ed) 1981 Radical pragmatics. New York: Academic Press. Culicover, Peter 1976 Syntax. New York: Academic Press. Culicover, Peter and Wendy Wilkins 1984 Locality in linguistic theory. New York: Academic Press. Culicover, Peter and Michael Rochemont. 1990 “Extraposition and the complement principle”. Linguistic Inquiry 21: 2348. Culicover, Peter, Thomas Wasow, and Adrian Akmajian (eds) 1977 Formal syntax. New York: Academic Press.

54

KEN-ICHI TAKAMI

Erteschik-Shir, Nomi 1973 On the nature of island constraints. Ph.D. dissertation, MIT, Cambridge, MA. Fodor, Janet 1978 “Parsing strategies and constraints on transformations”. Linguistic Inquiry 9: 427-473. Guéron, Jacqueline 1980 “The syntax and semantics of PP extraposition”. Linguistic Inquiry 11: 637-678. Johnson, Kyle 1986 A case for movement. Ph.D. dissertation, MIT, Cambridge, MA. Kennedy, Becky 1989 “On bridging”. Manuscript, Harvard University. Koster, Jan 1978 “Conditions, empty nodes, and markedness”. Linguistic Inquiry 9: 551594. Kuno, Susumu 1974 “The position of relative clauses and conjunctions”. Linguistic Inquiry 5: 117-136. Kuno, Susumu 1975 “Conditions for verb phrase deletion”. Foundations of Language 13: 161175. Kuno, Susumu 1978 Danwa no bunpoo (Grammar of discourse). Tokyo: Taishukan. Kuno, Susumu 1979 “On the interaction between syntactic rules and discourse principles”. In G. Bedell, E. Kobayashi, and M. Muraki (eds), 279-304. Kuno, Susumu 1980 “Functional syntax”. In E. A. Moravcsik and J. R. Werth (eds), 117-135. Kuno, Susumu 1983 “Principles of discourse deletion”. Proceedings of the XIII International Congress of Linguistics : 30-41. Kuno, Susumu 1987 Functional syntax: Anaphora, discourse and empathy. Chicago: The University of Chicago Press. Kuno, Susumu 1995 “Null elements in parallel structures in Japanese”. In R. Mazuka and N. Nagai (eds), 209-233. Kuno, Susumu and Ken-ichi Takami 1993 Grammar and discourse principles: Functional syntax and GB theory. Chicago: The University of Chicago Press. Larson, Richard 1988 “Light predicate raising”. Manuscript, MIT. Li, Charles (ed) 1976 Subject and topic. New York: Academic Press.

A FUNCTIONAL CONSTRAINT ON EXTRAPOSITION FROM NP

55

Mazuka, Reiko and Noriko Nagai (eds) 1995 Japanese sentence parsing. Hillsdale, N. J.: Lawrence Erlbaum Associates, Inc. Moravcsik, Edith A. and Jessica R. Werth (eds) 1980 Syntax and semantics 13: Current approaches to syntax. New York: Academic Press. McNulty, Ellen 1988 The syntax of adjunct predicates. Ph.D. dissertation, University of Connecticut, Connecticut. Nakajima, Heizo 1989 “Bounding of rightward movements”. Linguistic Inquiry 20: 328-334. Nakajima, Heizo 1990a “Secondary predication”. The Linguistic Review 7: 273-309. Nakajima, Heizo 1990b “Against the interpretive nesting requirement”. Metropolitan Linguistics 10: 40-54. Napoli, Donna Jo 1989 Predication theory: A case study for indexing theory. Cambridge: Cambridge University Press. Pesetsky, David 1982 Paths and categories. Ph.D. dissertation, MIT, Cambridge, MA. Prince, Ellen 1981 “Toward a taxonomy of given-new information”. In P. Cole (ed), 223255. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A comprehensive grammar of the English language. London: Longman. Roberts, Ian 1988 “Predicative APs”. Linguistic Inquiry 19: 703-710. Rochemont, Michael 1986 Focus in generative grammar. Amsterdam: John Benjamins Publishing Company. Rochemont, Michael and Peter Culicover 1990 English focus constructions and the theory of grammar. Cambridge: Cambridge University Press. Rothstein, Susan 1983 The syntactic forms of predication. Ph.D. dissertation, MIT, Cambridge, MA. Takami, Ken-ichi 1988 “Preposition stranding: Arguments against syntactic analyses and an alternative functional explanation”. Lingua 76: 299-335. Takami, Ken-ichi 1990 “Remarks on extraposition from NP”. Linguistic Analysis 20: 192-219. Takami, Ken-ichi 1992 Preposition stranding: From syntactic to functional analyses. Berlin: Mouton de Gruyter.

56

KEN-ICHI TAKAMI

Takami, Ken-ichi 1995 Kinouteki koubun-ron ni yoru nichieigo hikaku (A comparison between English and Japanese from the standpoint of functional syntax). Tokyo: Kurosio Publishers. Takami, Ken-ichi and Susumu Kuno 1992 “Extraposition from NP and VP-internal subjects”. Harvard Working Papers in Linguistics 1: 155-173. Williams, Edwin 1980 “Predication”. Linguistic Inquiry 11: 203-238.

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

57

The Speech Act Empathy Hierarchy and Russian Possessives Olga T. Yokoyama University of California, Los Angeles

1.

Introduction

The introduction of the concept of Empathy into linguistic analysis launched a transition from a mechanistic to a discourse-oriented, and, ultimately, humanoriented approach to syntax. The by now familiar hierarchies of Semantic Case, Speech Act Empathy, Descriptor Empathy, Word Order Empathy, Surface Structure, Anaphoricity, Topicality, and Humanness have been shown to explain a host of linguistic phenomena from various languages that otherwise defy consistent syntactic explanation in any formal framework that excludes the human factor (Kuno and Kaburaki 1975/77, Kuno 1987).1 As is the case with any viable and vital theory, the actual formulation of the hierarchies has undergone changes and modifications since their first appearance in Kuno (1972). The Speech Act Empathy hierarchy, specifically, was first formulated as E(speaker) > E(hearer) > E(third person) in Kuno and Kaburaki (1975/77), and then revised as E(speaker) > E(others) in Kuno (1987). Violations to this hierarchy noted in certain contexts have been explained by the unintentional character of the violations (Kuno 1987: 213-4) and have not been considered significant enough to warrant a modification of the theory. Evidence from Russian possessives, however, shows that violations are not only much more extensive, but also principled. As I shall argue here, a more regular theoretical treatment of such violations is actually called for. Given the multiplicity of the hierarchies, the question of their interaction has, of course, already received considerable attention. To handle the problem of the interaction of multiple hierarchies affecting reflexivization, Kuno

58

OLGA T. YOKOYAMA

(1987: 159 ff.) has suggested a quantitative approach, whereby the trigger potential of a given NP for reflexivization can be computed by adding the positive and negative values of the NP with respect to relevant hierarchies. The NPs referring to the entities participating in the logical form of the sentence, i.e., the participants of the narrated event,2 are ranked according to their total values, and the higher the rank of an NP, the more acceptable the sentence is with a reflexive controlled by that NP. This calculus of multiple factors explains the gradations in acceptability judgments of sentences with a particular referential interpretation. This approach suggests a way of evaluating the trigger potential of the referential material directly contained in the sentence, namely the NPs referring to the participants of the narrated event. The participants of the speech event, e. g. the speaker “I”, are considered only when they also happen to participate in the narrated event — that is, when first or second person pronouns occur in the sentence. Granted that syntactically, it seems absurd to consider the structural effects of elements absent in the logical form of the sentence; pragmatically, however, it is surely justified to considier the possibility of the participants of the speech event affecting the form of the utterance, even in cases when no overlap between the participants of the narrated event and the participants of the speech event occurs. This is precisely the possibility that will be examined below, on the basis of data from Russian possessive pronouns. The role of Empathy in Russian reflexives was explored in Yokoyama (1975, 1978, 1980, 1991), and in Yokoyama and Klenin (1976). The analysis in these works of the distribution of reflexive and non-reflexive possessive pronouns (anaphors and pronominals, in Chomsky’s 1981 framework), and of cases of unusual control of reflexive NPs across clause boundary (“free” reflexives of various sorts), highlighted the role of the speaker’s point of view in selecting the reflexive option. The analysis, however, was never performed in terms of Speech Act Empathy or in terms of the interaction of hierarchies. In this paper, then, I reconsider the Russian possessive pronouns in the light of these discourse-theoretical issues. I will argue that what at first glance appears to be pervasive pattern of violations of the Speech Act Empathy hierarchy in Russian constitutes a regular phenomenon explainable by two possible discourse orientations speakers of Russian regularly choose from. I will suggest that these two orientations correlate with two opposite orderings in the Speech Act Empathy hierarchy, and that either ordering is equally

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

59

justified in a given discourse orientation. I will, moreover, suggest that this revised “reversible” Speech Act Empathy hierarchy regularly interferes with other Empathy hierarchies, regardless of whether or not the participants of the speech event formally figure in the logical form of the utterance. With regard to Russian reflexive possessives, I will also propose a refined formulation of their status as Empathy expressions, which I will explain by the general fact that non-reflexive possessive pronouns are Speech Act Shifters.3 In section 1 of this paper, I provide a brief introduction into the relationship between nominal referential expressions and the participants of the speech event; specifically, I consider the role of point of view in the speaker’s choice of nominal referential expressions. In sections 2 and 3, Russian reflexive possessives in sentences with first and third person subject are examined in terms of the systematic interaction between possessives and the speaker’s point of view;4 in section 4, I consider theoretical implications and conclusions.

2.

Nominal Referential Expressions as Indicators of the Speaker’s Point of View

Violations of the Speech Act Empathy hierarchy as discussed by Kuno occur mostly in parrot questions and in corrective and precorrective statements, where they are explained by the unintentional character of the violations (1987: 213-4). It is possible to view such sentences as originating in the addressee’s mind. In repeating them, the speaker assumes the addressee’s point of view, which means that the speaker empathizes with the addressee rather than with him-/herself. This has a direct bearing on the Speech Act Empathy hierarchy, as the speaker in effect reverses it from “E(speaker) > E(others)” to “E(others) > E(speaker)”. I will argue that the speaker’s ability to take the point of view of others than self, thereby violating the Speech Act Empathy hierarchy, is actually a regular phenomenon observable in many areas of language. The violation of the Speech Act Empathy hierarchy is perhaps most obvious in the choice of nominal referential expressions. When a mother asks her child “Where is daddy?”, she is referring to her husband (whom she normally calls, say, Paul) by the label her child normally uses to refer to him. While this is a common sort of experience for all of us, the choice of

60

OLGA T. YOKOYAMA

referential expression is not as trivial as it may seem at first sight. In Yokoyama (1987), I propose a discourse model that operates with the interlocutors’ knowledge sets, in which referential knowledge is a complex cognitive category governed by pragmatic rules. Unlike lexical entries for common nouns, where the specification for, e.g., “table”, is more or less the same for all speakers of the language, the “referential entries” for, say, the father and husband in question differ depending on whose “listing” we consult. In the wife’s “listing”, the referential entry may be {Paul}, while in the child’s, it is {daddy}. The wife, however, also “cross-lists” the fact that the individual listed by her as {Paul} is listed as {daddy} in her child’s “listing”; the child, on the other hand, may or may not be aware that there are other options for “cross-listing” {daddy} in the mother’s “listing”. When the mother asks her child “Where is daddy?”, the referential entry she uses for the man is not her own, but her child’s, which she happens to have “cross-listed”. She has thus chosen her child’s point of view over her own.5 We normally know several ways to refer to one and the same individual, and only some of them we consider to be “our own”, while the others represent various ways known to us that others call the same individual, included among which are, of course, those of a general or public status. The choice of a specific referential expression depends on the speaker’s point of view. Thus the mother will not ask her child “Where is Mr. Smith?”, even if the child can be expected to know that his father is Mr. Smith. The same question, on the other hand, is quite plausible when addressing the maid, or the plumber, in whose listings the referent is likely to be listed as Mr. Smith. Similarly, a student referring to a professor as “Professor O’Connor” exhibits a public point of view as compared to when s/he privately refers to the same individual as “Jill” or as “O’Connor”. The same can be said about autoreferences: referring to oneself as “I”, as “Bill Smith”, as “your humble servant”, as “we” (editorial or royal), or as “Bibi” (Bill’s baby name), all reflect points of view the speaker can take when talking about himself. These examples are non-trivial in that they crucially show the fundamental ability of the speaker to take the addressee’s or a third party’s point of view over his or her own. Once this ability and its potential for linguistic manifestation are established, it should occasion little surprise to find that the speaker’s point of view manifests itself in other linguistic phenomena as well.

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES 3.

61

Possessive Pronouns Coreferent with a First Person Subject

Before considering the problem of the first person antecedents of possessive pronouns in Russian, I will outline the basic facts about Russian reflexive possessive pronouns. Although their use is not obligatory, Russian reflexive possessive pronouns, when used, are normally coreferent with the clause subject. Notably, exceptions to clausal subject control do exist, as cases are found where the coreference of a reflexive pronoun is not with the grammatical subject, i.e. the NP in the nominative case, but with the Theme or with the “deep” subject (Chvany 1973, Klenin 1974). Significantly, however, in every configuration when a reflexive possessive pronoun is possible, a non-reflexive is — at least technically — possible as well. Timberlake (1980) presents a detailed study of the distribution of the two possessives, isolating syntactic and semantic conditions that correlate with it, but excluding Empathy. In contrast, it is primarily the role of Empathy in controlling the actual choice between reflexive and non-reflexive possessives that will concern us in this paper. It is well known that the person of the subject affects the likelihood of the use of reflexive possessives over non-reflexives: unlike first and second person subjects, which show no clear dominance of one type of possessive over the other, third person subjects in modern literary Russian tend to cooccur more frequently with reflexives. In stylistics handbooks, the correlation of person and possessive pronouns is formulated prescriptively: with first and second person subjects, most handbooks recommend reflexives over nonreflexives, while acknowledging certain emotive connotations of the nonreflexive alternatives; with third person subjects, on the other hand, virtually all handbooks prescribe the exclusive use of reflexives, despite some wellattested cases of non-reflexives in literature, and considerable fluctuation in colloquial usage. This prescription is invariably motivated by the potential referential ambiguity of the non-reflexive alternative, which in the third person could refer to any third person referent, including those outside the domain of the sentence in question; a reflexive pronoun, being coreferent with the subject, would of course disambiguate the reference, as in (1):6 (1)

Borisi zabral svojui/egoi/j stat’ju. B-n took self’s /his paper-a ‘Boris took his paper.’

62

OLGA T. YOKOYAMA

While formally the potential for ambiguity can hardly be disputed, it is worth noting that in context referential ambiguity rarely arises. In fact, potential ambiguity has traditionally been blamed for far more phenomena than it should be held responsible for. In the case of (1), consider, for example, the English translation: his is potentially no less ambiguous in the English version,7 yet the language has managed quite well without the reflexive alternative, which is allegedly indispensable in Russian to disambiguate the possessor’s reference. In fact, within Russian itself, countless examples of potential ambiguity can be found, which are regularly disambiguated in context without resorting to special markers; consider (2): (2)

Sergeji resitel’nymi sagami podosel k Petruj. Oni/*j/*k S-n with-decided steps approached to P. he kazalsja vzvolnovanym. seemed agitated ‘Sergej approached Peter decisively. He seemed agitated.’

In (2), although the third person subject he of the second sentence could potentially refer to Sergej, Peter, or any third male referent, there is little doubt that the actual co-reference is with Sergej. As in all other similar cases, it is discourse factors alone that disambiguate the reference, not lexical cues or syntax. Since ambiguity is not an issue, in the presentation that follows, I will ignore the possibility of referential ambiguity of a third person pronoun with an unspecified third person referent. What will concern us with respect to third person pronouns, as well as with first and second person pronouns, is the choice between reflexive and non-reflexive possessives coreferent with the clause subjects. Consider the following utterances (3) and (4); in (3), the possessive pronoun is an anaphor, and in (4), it is a pronominal: (3)

Ne bespokojsja, jai svoei delo znaju! not worry I-n self’s job know ‘Don’t you worry, I know my job!’ (Krylov)

(4)

Ox, i ljublju ze jai mojui Verin’ku! oh ptcl love ptcl I-n my V-a ‘Oh, do I adore my little Vera!’

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

63

The possessive pronouns in these two sentences both refer to the clause subject. To account for this, technically, one might say that in Russian surface pronominals are actually underlying anaphors. This, however, would imply that the strings svoju and moju are completely synonymous and that the choice between the two forms is essentially random and meaningless. It is true that reversing the possessives won’t change the reference or the truth value of sentences (3) and (4), yet the fact is that Krylov, a poet and playwright, chose a reflexive in (3), and that a reflexive would be virtually impossible in (4). Assuming that the two forms are completely synonymous and interchangeable would thus not gain us much by way of explaining the empirical facts that reflect native competence. Timberlake (1980) finds preference for reflexives with inanimate heads, and for personal possessives with animate heads. It could then be pointed out that the difference in the possessives depends — rather mechanistically — on animacy: the NP modified by the possessive is inanimate in (3) and animate in (4). This, however, is not tenable in this case, since even if delo ‘business” were to be replaced with e.g. sem’ju ‘family’, the preferred possessive would still be reflexive. Conversely, if Verin’ku were replaced with the inanimate lodocku ‘boat’, the possessive pronoun would remain the same moju ‘my’. So the need remains to explain the distribution of the possessives in (3) and (4) without mechanistically resorting to animacy. In fact, if the correlation with animacy is indeed significant, we want to explain why it is, since the correlation may be a result of some more general principle. Turning now to a functional explanation, we observe first that in both (3) and (4) a participant of the narrated event and a participant of the speech event, namely the speaker, overlap. The trigger potential of the first person subject can be expected to be very high in both sentences. The ostensible failure of the antecedent to trigger a reflexive in (4), as opposed to (3), therefore still remains to be explained. From numerous examples like these, we might conclude that reflexivization in Russian possessives is not affected by discourse factors like Kuno’s Empathy and its hierarchies; alternatively, we might conclude that discourse/pragmatic factors other than these hierarchies control the choice in question; and finally, we could try to re-examine the hierarchies themselves, in an attempt to account for the contrast in (3) and (4). The first option is most unlikely, as reflexives have been shown to be sensitive to point of view in more than one language and in many reflexive constructions.8 The second option is suspect for the same reason, although it is

64

OLGA T. YOKOYAMA

not excluded that more discourse factors are operative in language than have hitherto been uncovered.9 Considering the well-documented validity of Kuno’s Empathy hierarchies in data from several languages, the third option appears at the moment to be the most attractive one. I will therefore pursue this line of inquiry. I will suggest that the choice between reflexive and personal possessives coreferent with the person subjects is dependent on the ways in which the speaker views him-/herself. Let us consider examples (5) and (6), where the possessives are both coreferent with first person subjects:10 (5)

... tut napisano / cto mnenie zavisit ... suzdenie here written that opinion-n depends judgment-n zavisit ot togo kak ja otvedu / kak mol jai ego depends on that how I-n parry how ptcl I-n him udovletvoril svoimii otvetami. satisfied with-self’s answers. ‘...it’s written here that the opinion depends, the judgment depends on how I warded off, how I (he thinks) satisfied him with my answers.’ (cited in Zemskaja and Kapanadze 1978:166)

(6)

To est’ ty podozrevaes’, cto jai prisel sklonjat’ tebja i. e. you-n suspect that I-n came to-persuade you-a ostat’sja u knjazja, imeja v tom svoii vygody. No to-remain at prince’s having in that self’s benefits but drug moj, uz ne dumaes’ li ty, cto jai iz friend my ptcl not think ptcl you-n that I-n from Moskvy tebja vypisal, imeja v vidu kakuju-nibud’ Moscow you-a called having in mind some svojui vygodu? self’s profit-a ‘You mean that you suspect that I came here to talk you into staying at the prince’s place for my own good. Well, my dear, don’t you think by any chance that I have called you here from Moscow with some interest of my own in mind as well?’ (Dostoevskij)

The reflexive possessives in (5) and (6) are coreferent with the first person subjects of their clauses. Even though these sentences have first person subjects, however, there are formal indicators in both cases that the speaker utters these sentences from a point of view other than his own: in (5),

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

65

we find the evidential particle mol ‘allegedly’, and in (6), the clauses with the reflexive possessive are embedded under main clauses that explicitly refer to the addressee’s opinion (“you suspect that I”, “don’t you think that I”). The situation is logophoric, in Kuno’s (1987) sense of the term, despite the first person subjects. Moreover, the logophoric nature of the clauses, which introduces a point of view different from that of the clause subject, apparently favors reflexive possessives. Examples (5) and (6) particularly clearly show what I suggest is the main general factor distinguishing sentences with reflexive possessives from those with non-reflexives in sentences with first person subjects. Based on numerous similar examples of reflexive possessives coreferent with first person subject, I have suggested that the speaker of such utterances assumes an extrinsic point of view on him-/herself (Yokoyama 1980). The intuition that a reflexive possessive in such utterances encodes an extrinsic point of view originally belongs to Peskovskij (1914/38: 169), who suggested that if the reflexive possessive in the sentence from Boris Godunov’s aria “I poisoned self’s sister the czarina, the humble nun” were to be replaced with a non-reflexive, the czar Boris would be admitting his guilt, whereas in the aria as it stands, he is attributing his involvement in his sister’s death to public opinion, in effect denying it. In this case, there are no other indicators of logophoricity in the sentence, and the choice of the pronoun alone affects the truth value of the sentence, even acquiring a certain legal and/or historical significance.11 While such drastic connotations of pronominal choice may be rare, the more subtle implications of the choice do nevertheless affect the overall message the addressee gets when hearing these utterances. A general disengagement and distancing of the speaker from the sentential subject is the subliminal message sent by reflexive possessives coreferent with a first person subject. Significantly, the distancing need not imply another person’s point of view. The speaker can view him-/herself from a distant point of view in a number of situations, the great majority of which occur in self-analytical and/or retrospective contexts. This implicitly assumes the existence of two “I’s”. The “I” that reflexivizes the possessive pronoun in (3), (5) and (6), for example, is different from the “I” that does not do this in (4). The “I” in (3), (5) and (6) is objective and extrinsic, while the “I” in (4) is subjective, intrinsic, and intimate. Although this assumption may seem excessively metaphysical, the phenomenon of split ego is well known in psychology and is eloquently described in

66

OLGA T. YOKOYAMA

works of fiction. The Russian reflexive possessive data merely corroborate numerous other similar phenomena borne out by our cognitive capacity to view ourselves through others’ eyes. Neither Russian nor English happens to have otherwise synonymous first person pronoun forms that would reflect this distinction lexically, but recall some of the autoreferential options mentioned in section 1: “Your humble servant is here, ma’am” is one way of referring to oneself by another, addressee-oriented referential expression. Along these lines, example (7) represents a case of self-analysis, (8) is a case of viewing oneself in retrospect, and (9) describes an analytical perception of one’s own voice: (7)

(...) jai (...) i togda uze sam soznaval svoei padenie. I-n even then already myself realized self’s fall-a ‘... even then I already was aware of my fall myself.’ (Dostoevskij)

(8)

Jai perezil svoii zelan’ja, Jai razljubil svoii I-n outlived self’s desires-a I-n stopped-loving self’s mecty (...)12 dreams-a ‘I have outlived my passions, I love my dreams no longer...’ (Puskin)

(9)

Jai slysal svoei punktirnoe, trjasusceesja dyxanie (...) I-n heard self’s interrupted shaking breath-a ‘I heard my own interrupted, shaky breath’ (Zamjatin)

The detached, extrinsic point of view in (5)-(9) contrasts with a selfengaged, intrinsic point of view the speaker holds on him-/herself in corresponding utterances with non-reflexive pronouns; consider (10) and (11): (10) ... jai predajus’ moimi mectam. I-n give-myself-away my dreams ‘I indulge in my dreams.’ (Puskin) (11) Ne uterpev, jai sel zapisyvat’ ètu istoriju moixi pervyx not bearing I-n sat to-write this history-a of-my first sagov na ziznennom poprisce, (...) nikogda uze bolee ne strides on life’s road never again sjadu pisat’ mojui avtobiografiju, daze esli not will-sit to-write my autobiography-a even if

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

67

prozivu do sta let. I-will-live to 100 years ‘Not able to hold back, I sat down to write down this story of my first steps on the path of life, (...) never again will I sit down to write my autobiography, even if I live to be a hundred years old.’ (Dostoevskij) Sentence (10) constitutes a striking minimal pair with (8) above. Note that in both cases the personal possessives modify the same inanimate noun mecty ‘dreams’. The most obvious difference between (8) and (10) is their tense, but it is only one of the possible manifestations of a more general distinction that two “I’s” can have. In (8), the speaker views himself with detachment characterisitc of remembering one’s past. In (10), on the other hand, his stance is self-affirming, and he does not separate the dreaming self from the self that utters the sentence. Similarly, there are no indications in (11) of the speaker’s assuming a distant, extrinsic point of view upon himself. Instead, the speaker is direct and open, and the narration clearly reflects his personal point of view. This particular speaker, a teenage protagonist of the novel, in fact shows a clear preference for non-reflexive possessive pronouns throughout the novel, and this is indeed what can be expected from a confessional first person narrative by an immature, self-centered raw youth13. Significantly, however, even this self-centered raw youth reserves reflexives for cases like (6) and (7) (both are his narration from the same novel), as well as other contexts in which he views himself extrinsically, conveying another person’s point of view, or his own self-reflective, self-analytical view of himself14. Returning to the contrast between (3) and (4), we can see that (4) is a highly subjective, intimate, and defenselessly open exclamation in which the speaker expresses her private, self-engaged and emotionally-colored point of view. It is not entirely an accident that the sentence contains a high number of discourse particles, which typically lend a personal, subjective, and intimate ring to an utterance; and the exclamatory intonation of the utterance also suggests an intimate discourse setting. In (4), then, as in (11), the speaker speaks from her own, personal point of view. Not so in utterance (3), which is a stand-offish, extrinsically-oriented statement: the speaker, in fact, is a police officer, telling off a nosy outsider to refrain from uncalled-for advice. While not exactly conveying the addressee’s point of view, as was the case with (5)

68

OLGA T. YOKOYAMA

and (6), utterance (3) is nevertheless a statement oriented towards the addressee, of whom the speaker is keenly aware and whose very views he wants to straighten by the content of the utterance. It is thus clear that point of view is a critical factor in the choice of Russian possessive pronouns coreferent with first person subjects. It is also clear, however, that at work here are factors of far greater subtlety than the Empathy hierarchies as they have been formulated so far. How, then, can these data be explained in terms of Empathy? If empathizing with N means taking N’s point of view, then (3) and (5)-(9) suggest that the speaker, when speaking about him-/herself, can exercise the option of not empathizing with him-/herself, despite the Speech Act Empathy hierarchy. The Russian data suggest, moreover, that this option is not taken randomly but that it correlates with distanced, official, formal discourse situations, in which, as I argue elsewhere (Yokoyama 1994), “the grammar of restraint” is operative; conversely, when “the grammar of closeness” is in operation, the speaker tends to empathize with him-/herself, and the Speech Act Empathy hierarchy as we know it operates in the expected fashion15. Examples like (3) and (4) suggest, moreover, that not only can the Speech Act Empathy hierarchy be reversed, but that, reversed or not, the Speech Act Empathy hierarchy, containing an entity (i.e. the addressee) absent from the logical form of the sentence, must be taken into account if we are to explain the distribution of possessives in such sentences. Finally, the data in this section indicate that reflexive possessives coreferent with a first person subject occur in Russian precisely when the speaker does not empathize with him-/herself, i.e., in the “grammar of restraint”. Reflexivization fails, on the other hand, in intimate, private utterances like (4), (10) and (11), in which the speaker expresses his or her personal point of view and operates with the grammar of closeness. As argued in section 1 in connection with the choice of nominal referential expressions, it is hardly unusual for the speaker to violate the Speech Act Empathy hierarchy by taking another person’s point of view over his or her own. Russian possessives, as demonstrated in cases like (3) and (5)-(9), provide an additional illustration of the pervasiveness of violations of the Speech Act Empathy hierarchy. It is the frequency and the predictability of the context of such violations that suggest that the Speech Act Empathy hierarchy itself is not permanently fixed in one direction, but reverses its order in a regular dependence on two opposite but equally plausible discourse

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

69

orientations: the distant/restrained communicational mode and the intimate/ close mode. I suggest, moreover, that in a given mode, the Speech Act Empathy hierarchy must be taken into consideration regardless of the presence or absence of the participants of the speech event in the logical form of the sentence. What remains problematic, significantly, with possessives coreferent with first person subject is that, at least as far as utterances with first person subject are concerned, reflexive possessives in Russian indicate the lack of the speaker’s Empathy with self rather than the other way around. This is problematic for the elegant and attractive characterization of reflexives as a priori Empathy expressions, as described in Kuno (1986). Before we attempt to explain this unexpected behavior of Russian possessive reflexives, let us first consider possessives coreferent with third person subjects.

4.

Possessive Pronouns Coreferent with Third Person Subject

As mentioned in section 2, reflexivization by a third person subject rarely fails to apply in literary texts16. In texts duplicating colloquial speech, however, non-reflexive possessives can in fact be found, despite the norm. This is particularly true in first person narratives; (12) appears in the same novel as (6), (7) and (11): (12) Onii mogut prodolzat’ zit’ po-svoemu v samyx they-n can continue to-live their-way in most nenatural’nyx dlja nix polozenijax i v samyx ne ixnixi unnatural for them conditions & in most not their polozenijax ostavat’sja soversenno samimi soboj. states remain completely self self My tak ne umeem. we-n thus not can ‘They can continue to live in circumstances that are most unnatural for them, and in the most alien [lit. ‘not their’] kinds of situations, they can remain completely their own selves. We don’t know how to do that.’ (Dostoevskij) The non-reflexive possessive ixnix ‘their’ in (12) appears in a relatively isolated syntactic environment (Locative phrase), which diminishes (but does

70

OLGA T. YOKOYAMA

not block) its accessibility for reflexivization. It nevertheless requires an explanation, since the subject NP oni ‘they’ ranks high on several Empathy hierarchies and its trigger potential is uncontested by any other NP in the sentence17. If we look for the reasons for not using the reflexive option despite the high Empathy ranking and seek formal clues about the point of view from which (12) is delivered, the clearest indication is found not in the same sentence but in the next one. The contrastive we in the beginning of the following sentence suggests that the gist of the two sentences is the difference between the two classes, them and us, the educated westernized gentry and the uneducated simple Russian folk; the judgment, significantly, is passed by a westernized member of us upon the simple them. Another important piece of evidence comes from the phrase nenatural’nyx dlja nix ‘unnatural for them’, where there is also a potential option of a reflexive pronoun sebja instead of the non-reflexive nix. As I have argued in Yokoyama (1978) and (1980: 737-51), reflexivization by the “upstairs” third person subject across a clause boundary into an adjectival clause occurs when the speaker takes the point of view of the “upstairs” subject whose judgment the adjective represents; conversely, failure to use a reflexive indicates that the speaker retains his or her own point of view, from which the subjective adjective is true, and does not empathize with the third person subject of the main clause. Sentence (12), then, contains at least two formal indications that the speaker has not taken the third person subject’s point of view but has retained his own, selfcentered viewpoint: the speaker’s Empathy is thus with himself, not with his subject. Another example with an unexpected non-reflexive possessive corroborates this explanation18: (13) Predstavit’ sebe, cto zdes’, na Krasnoj Ploscadi, u to-imagine to-self that here on Red Square by Mavzoleja, ne my, a fasisty, v ix forme, v ix Mausoleum not we-n but facists-n in their uniform in their furazkax, s ix svastikami na rukavax ... Net, predstavit’ caps with their swastikas on sleeves no to-imagine sebe èto bylo nevozmozno! to-self this was impossible ‘To imagine that here, on Red Square, by the Mausoleum, it’s not us, but the fascists, in their uniforms, with their caps on, with their swastikas on their sleeves ... No, it was impossible to imagine it!’ (Simonov)

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

71

The prepositional phrases in (13) are “internal” to the subject NP, and as such represent environment relatively unaccessible to refelxivization (Timberlake 1980: 785). Nevertheless, reflexives do regularly occur in internal phrases like ona so svoej molodost’ju ‘she with her youth’. I suggest that the choice of non-reflexive possessive in (13) is motivated by point of view. In this sentence, the opposition is between us (Russians) and them (the fascists), who, in a frightening fantasy, are placed in the heart of Russian soil during World War II. The contrast ne my, a fasisty ‘not us but fascists’ is explicitly verbalized, and the point of view of the speaker is unambiguous from the meaning of the main clause. Note that compared to the preceding example, the subject NP fasisty ranks lower with respect to its trigger potential than the subject in (12), because the presence of the first person us in (13) diminishes the subject’s trigger potential. It may be tempting, therefore, to attribute the non-reflexive to the lower trigger potential of the subject. The acceptability of the non-reflexive in (12) and (13), however, does not show any difference in degree that would be expected because of the calculable difference in their trigger potential. When considered in conjunction with (12), moreover, the preference for non-reflexives in both cases is clearly mirrored by the point of view opposition between the speaker and the third person subject, between us and them. The failure of the difference in the subjects’ trigger potentials in (12) and (13) to explain the equal acceptability of these sentences on the one hand, and the parallel usage of non-reflexive possessives in the same discourse conditions of contrast between the speaker and the third person subject on the other, suggest that it is the speaker’s self-centered point of view, rather than trigger potential, that controls the non-reflexivization in these two sentences. In both cases, the speakers do not empathize with the subjects of their sentences, retaining instead their own, personal standpoint; reflexives in these sentences would have erased the class and national boundaries between the third person subjects and the speakers that the latter are so acutely aware of. The situation in (12) and (13) contrasts with (14), where the possessive is also found in a relatively isolated syntactic environment (prepositional phrase). In (14), however, the speaker appears as an objectively-inclined omniscient narrator; this speaker does empathize with his third person grammatical subject, his animate topic, and his semantically prominent role, as the Empathy hierarchies predict, and a reflexive possessive coreferent with the subject occurs as expected19:

72

OLGA T. YOKOYAMA (14) (...) naxodjas’ lis’ v kosvennom rodstve being only in indirect blood-relationship so svoimii literaturnymi predkami, Zamjatini ne with self’s literary ancestors Z-n not odinakovo dalek to nix. equally far from them ‘... being only indirectly related to his literary predecessors, Zamjatin maintains different degrees of distance from some of them.’ (Bondarenko)

Let us consider for the moment the role of the Speech Act Empathy hierarchy in (12) and (14). Since neither the speaker nor the addressee figures in the logical form of either sentence, it is tempting to disregard this hierarchy. Nevertheless, the choice of the possessives, as well as the clear distinction in the speaker’s orientation in these two sentences, vitiates such a dismissal. As was suggested in connection with (12) and (13), the speaker of (12) does not empathize with his third person subject they but instead empathizes with himself. Thus the speaker’s self, while not overtly in the sentence, nevertheless attracts his Empathy, and precludes the use of reflexive possessive in (12) no less effectively than it does in (13), where the third person subject does in fact compete with an overt first person pronoun. This suggests that the Speech Act Empathy hierarchy must be reckoned with even when the speaker is not part of the logical form of the sentence. Only by incorporating the speaker’s Empathy with himself can we account for the difference in the possessives in (12)-(14). Consider now an example form colloquial speech, where the speaker tends to empathize with his or her own self over a third person subject more commonly than in formal styles. Observe the contrast between (15) and (16)20: (15) Eta zenscinai, kazetsja, k eei nacal’niku posla. this woman-n apparently to her boss went ‘This woman went to her boss, I guess.’ (16) *Vyjdja na pensiju, Tat’janai Ivanovna poexala zit’ going to pension T. I.-n went to-live k eei bratu. to her brother ‘Upon retirement, T. I. went to live with her brother.’

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

73

If the use of a non-reflexive indicates the speaker’s empathizing with the speaker him-/herself instead of empathizing with a third person subject, then the non-reflexive in (15) is well accounted for, as all the cues point to a lack of Empathy on the part of the speaker towards the third person subject: the alienating referential expression the woman and the hedge kazetsja ‘apparently’ suggest that the speaker is telling us about the woman from the speaker’s self-centered point of view. The utterance is clearly a colloquial one, uttered in a setting of close interlocutor distance21, and the usage of nonreflexive is hardly exceptional in this speech mode, as long as the speaker’s Empathy is with the self, and not with the third person subject. The Speech Act Empathy hierarchy is thus operative again in this environment, even though the speaker is absent from the logical form of the utterance. Example (16), on the other hand, has a clearly literate, formal and distanced ring, borne out by the adverbial gerund clause and the formal referential expression for the subject; in this speech mode, the speaker’s retaining his or her own point of view over that of the subject is highly unlikely, and a non-reflexive possessive is therefore unacceptable. Significantly, the Speech Act Empathy hierarchy in (14) and (16) is reversed. Such a reversal, which we have seen in both sections 1 and 2 above, is, in fact, expected when it correlates with a distanced discourse orientation of the speaker, which in turn correlates with reflexive possessives. Recall, also, that in the modern literary language, particularly in omniscient narration, reflexives coreferential with a third person subject are much more frequent than non-reflexives. If reflexives indicate the speaker’s empathy with a third person subject referent despite the Speech Act Empathy hierarchy, or rather, in accord with a reversed Speech Act Empathy hierarchy compatible with a distant discourse orientation, then the high frequency of reflexives in formal language naturally follows.

5.

Conclusions

Let us summarize our conclusions so far. In section 1, the choice of nominal referential expressions was analyzed as a case of point of view shifting, explained as the speaker’s borrowing a referential entry from another person’s referential lexicon. It was pointed out that such borrowing, which is tantamount to taking the viewpoint of people

74

OLGA T. YOKOYAMA

other than the self, constitutes a regular and wide-spread violation of the Speech Act Empathy hierarchy. In section 2, three suggestions were made: (1) It was suggested on the basis of sentences with first person subject that, as a matter of principle, the speaker can either take or abandon his or her own point of view. When the speaker takes it, s/he is empathizing with him-/herself, and when s/he abandons it, s/he empathizes with others (the speaker’s other, analytical self, the addressee, or an unspecified or generalized third party). It was pointed out that those cases in which the speaker abandons his or her point of view can be explained as a regular function of the distant mode of discourse orientation, rather than reflecting another instance of a violation of the Speech Act Empathy hierarchy. (2) It was suggested, moreover, that the revised, orientation-dependent Speech Act Empathy hierarchy must be taken into consideration at all times, regardless of whether or not the sentence in question contains the participants of the speech event in its logical form. Finally, (3) the material indicated that reflexive possessives in Russian utterances with first person subject correlate with a lack of the speaker’s Empathy with self rather than the other way around; conversely, it is the non-reflexives that indicate the speaker’s Empathy with self. This serves to complicate, as was pointed out, Kuno’s elegant characterization of reflexives as a priori Empathy expressions. Section 3 examined utterances with third person subject. Here, too, three suggestions were made: (1) it was observed that the speaker can either take or abandon his or her own point of view. When the speaker takes it, s/he is empathizing with him-/herself, and when s/he abandons it, s/he empathizes with others (the third person participant of the narrated event). It was argued that the predominance of cases in which the speaker empathizes with a third person subject over him-/herself provides an additional piece of evidence that two discourse modes (a distant and a close one) are operative in the language, and that they regularly correlate with two opposite orderings of the Speech Act Empathy hierarchy. (2) Those cases in which the speaker empathizes with him-/herself over a third person subject, which are much more common in the close mode of colloquial speech, were seen as an additional piece of evidence that the Speech Act Empathy hierarchy is relevant regardless of whether or not the speaker figures in the logical form of the sentence. Finally, (3), reflexive possessives occur when the speaker’s Empathy is not with him/herself, but with others (the third person subject).

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

75

The situation with sentences containing first and third person subjects, when viewed in this way, turns out to be strikingly parallel. In both cases the speaker regularly reverses the Speech Act Empathy hierarchy when in the distant communicational mode. In both cases there is a demonstrated need to consider this new communication-mode-dependent Speech Act Empathy hierarchy, regardless of the presence or absence of the participants of the speech event in the logical form of the sentence. And finally, in both cases the speaker’s failure to empathize with his or her self correlates with reflexive possessives. It is this last point that requires further consideration, as in sentences with first person subject it presents a problem for Kuno’s characterization of reflexives as a priori Empathy expressions. In sentences with third person subject, the fact that Empathy with the subject correlates with reflexives does not require any explanation, as this is expected. But to explain the status of reflexives in sentences with first person subject in a way that is consistent with sentences with third person subject, we must revise the discourse conditions for the use of reflexive possessives. I suggest the following formulation: (17) Reflexive possessives in Russian occur under coreference with a first or third person subject only when the speaker takes the point of view of others than him-/herself22. Such a formulation, of course, does not yet constitute an explanation. Why should the speaker’s Empathy with his or her self (the “real”, intimate, intrinsically-viewed self) correlate with non-reflexive possessives, and vice versa? A possible answer to this question lies in the deictic nature of nonreflexive pronouns23. Adjectives can generally be objective (e.g. American, metal), which do not necessarily presuppose an individual’s judgment, or subjective (e.g. difficult, boring), whose truth value depends on the individual judgment24. Possessive pronouns are adjectival, and as adjectives they belong to the latter, subjective group, at least to the extent that they are Speech Act Shifters. Quite simply, what I call my book, you must not call my book, but your book. The choice between the lexical items my and your when referring to the same book thus depends on who says it, you or I. When speaker A thinks of a book as belonging to him or her, A of course says my book. Even when A adopts the addressee B’s point of view, however, A cannot possibly adopt (as the mother in section 1 could borrow her child’s referential expression daddy ) B’s way

76

OLGA T. YOKOYAMA

of referring to that book, since this would change the truth value: from B’s point of view and in B’s words, the book is yours. Under conditions of coreference with the sentential subject (or its equivalents like Experiencers etc.), Russian seems to offer a lexical25 solution to this dilemma: instead of using Speech Act Shifters moj ‘my’, ego ‘his’, ee ‘her’, nas ‘our’ etc., the speaker uses the referentially equivalent anaphoric reflexive pronoun svoj ‘self’s’ whenever the speaker departs from the point of view of his or her private intimate self and adopts an extrinsic viewpoint. This explanation is admittedly has one flaw: it does not address the fact that reflexives are available only in very limited structural conditions of coreferentiality with the subject. I leave this question open at this point. To summarize: based on Russian material, it was suggested here that a further elaboration of Kuno’s Speech Act Empathy hierarchy may be in order. The modification presupposes two discourse orientations — a self-centered, intimate one and a self-disengaged, distanced one. It is in the former, selfcentered orientation that the speaker adheres to the natural, Speech Act Empathy hierarchy dictating that he or she must empathize with his or her own self over and above anybody else’s, regardless of whether or not the speaker occurs in the logical form of the sentence. When in this mode, the speaker uses the Speech Act Shifter pronouns my, his, her etc., even when coreferent with sentential subjects and, thereby, technically reflexivizable. When in the objective, self-disengaged mode, on the other hand, the speaker can empathize with the non-self over the self. It is in this objective mode that the speaker reverses the Speech Act Empathy hierarchy (as it was proposed in Kuno 1987), and uses anaphoric reflexive pronouns in accordance with the calculated sum values of the remaining discourse hierarchies26. Even though this proposal at first glance may appear to argue against Kuno’s (1987) formulation of the Speech Act Empathy, it is actually in agreement with the more general principles of Discourse Grammar, as advanced more recently in Kuno and Takami (1993). What unites the analysis of Russian possessive pronouns presented here with all of Kuno’s work, at its most abstract level, is the significance of the speaker and the hearer in structuring all sentences regardless of their propositional content.

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

77

Notes 1.

Kuno and Takami (1993) advance compelling arguments for the prominence of even more general discourse principles in syntactic structures. The evidence presented in Kuno and Takami (1993) for the critical role of the hearer’s ability to make associations, and to reconstruct the speaker’s intent, based on the frequency of real life situations encoded in sentential structures, essentially constitutes evidence for the role of pragmatics in syntax, and as such further underscores the crucial role of the human factor in language.

2.

This term is taken from Jakobson (1957).

3.

Cf. Jakobson’s (1957) paper on shifters, the first important empirically-based theoretical study of the interaction of such performative categories as speaker and hearer with the concrete morphological and syntactic phenomena of the Russian verb — a study which both predates and surpasses the analyses based on the performative hypothesis proposed by generative semanticists in the early 1970s.

4.

Sentences with second person subject are the most complex among cases of this sort in Russian. Although Cantrall’s (1974) analysis of “free” reflexives in English implied that first and second person behave in similar ways, and in ways that differ systematically from the third person, my own data regarding second person possessives in Russian are inconclusive on this point (cf. Yokoyama (1975, 1980), and Yokoyama and Klenin (1976), vs. Yokoyama (1991)). The situation in Russian is complicated by the presence of two second person pronouns, the familiar and the formal one, which agree with singular and plural verb forms, respectively. I reserve the question of Russian second person possessives for a separate study.

5.

See Uspenskij (1970) for an extensive documentation of point of view shifts in literature and art.

6.

The case-marking of major constituents is indicated by a lower-case letter hyphenated to the noun: -n(ominative), -a(ccusative), -d(ative), etc.

7.

See, among others, McCawley (1976), which suggests that English possessive pronouns like “his” represent a conflation of two pronouns, which in our terms would correspond to personal and reflexive.

8.

See e.g. Kuno (1987: 237-40); Zribi-Hertz (1989) and Reinhart and Reuland (1991) also support this view.

9.

Baker (1995), for example, suggests that the discourse parameters controlling reflexives are “intensity”, “contrast” and “prominence”. He does not, however, provide a clear definition of these concepts, either conceptually or formally.

10.

The symbol “/” in Zemskaja’s transcripts of conversations indicates a non-finite pause, preceded by a rising intonation.

11.

Timberlake (1980: 781ff.) argues that referentiality inhibits reflexivization in Russian possessives while non-referentiality favors it, basing his claims on statistical co-occurrence. It is quite likely that the co-occurrence is caused by the more general tendency to refer to existing and presupposed entities in empathetic contexts. With regard to this particular sentence, Timberlake (1980: 784) suggests that the phrase svoju sestru ‘self’s sister’ is descriptive and therefore non-referential, and that the sentence is to be glossed ‘I

78

OLGA T. YOKOYAMA poisoned the person who is my sister’. Such interpretation, however, would imply that Boris does admit to the murder of his sister, which, is clearly not what Boris himself thinks.

12.

Rugaleva (1980b: 54ff.) argues that this example (as well as other examples of reflexive possessives) must be explained by the “semantic accent” on the subject “I” rather than on the object “desires”. Motivating this by the general fact that reflexives are coreferent with the subject, Rugaleva suggests that non-reflexive possessives, conversely, imply a focus on the noun modified by the possessive. The nuances captured by Rugaleva are in a sense consistent with Timberlake’s (1980) statistics indicating that referentiality inhibits reflexivization: referential nouns, after all, represent independent entities in the knowledge sets of the speaker and hence may be perceived as more “in focus”. One of the problems with Rugaleva’s analysis is that it is generally difficult to determine where the focus is. In a sentence like Ja — rab moix tainstvennyx, neobycajnyx snov ‘I am a slave of my mysterious, unusual dreams’ (1980b: 57), it is unclear whether the focus is on the dreams or on I, since after all I is the dreamer of these dreams. Also problematic in principle, I believe, is her denial of any relationship between reflexives and point of view, which goes counter to numerous cross-linguistic facts.

13.

An informal sample of 160 possessives in this novel contained only 23% of reflexives.

14.

More examples of these can be found in Yokoyama (1980: 199-210).

15.

In Yokoyama (1994), I argue that these two grammars co-exist in Russian; the argument draws on an analysis of extensive data from every component of linguistic structure. Every speaker, in effect, switches codes between these two grammars depending on whether or not the proposition [[I and you are close to each other]] is found in the speaker’s set of current concern. The rules of the “grammar of closeness” are “plugged in” when this proposition is present in it and those of the “grammar of restraint” when it is not.

16.

This must be due in part to the red pen of editors well-schooled in the prescriptive rules requiring reflexive possessives coreferential with third person subjects. Evidence for such editorial interference appears in the following sentence from a novel by Kocetov: Emu odnomu, predsedatelju, dostavalos’ ot semenovodki. Ona soznavala procnoe ee polozenie v kolxoze i komandovala pravleniem. ‘Only he, the chairman, had to take so much guff from the seed specialist. She was aware of her firm position in the collective farm and took charge of bossing it.’ The non-reflexive possessive occurred in several editions of this novel until it was “caught” by stylists, who cited it as an example of wrong usage in the stylistics handbook by Finkel and Bazenov (1960:343). All subsequent editions of the novel have the “corrected” reflexive possessive. For a detailed discussion of this example, see Yokoyama and Klenin (1976: 252-4).

17.

The non-reflexive possessive in (12) is in fact doubly non-normative: a third person nonreflexive possessive pronoun is not supposed to agree in case or number with the noun it modifies (as it does in this example), and the reflexive svoix should of course have been used in lieu of the non-reflexive form.

18.

I owe the reference to this example (cited in Veyrenc (1980: 290)) to Katerina Krivinkova.

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

79

19.

When the speaker empathizes with a third person subject, s/he suppresses his or her own viewpoint and can be seen as effacing him-/herself. Kuno (1987: 264) objects to attributing Russian reflexive possessives in sentences with third person subject to such selfeffacing by the speaker; this objection is justified only to the extent that the speaker of Russian does not always have to efface him-/herself when uttering sentences with third person subjects. In personal, intimate and private situations, which in a sense can be seen as most natural to human communication, the speaker does not efface him-/herself, does not empathize with the third person subject, and, consequently, does not use reflexive possessives coreferent with a third person subject. Such utterances are not at all uncommon in colloquial speech, as seen in (15) below. Otherwise, however, the self-effacing of the speaker in utterances with third person subject basically corresponds to cases when the Speech Act Empathy hierarchy is not operative. Self-effacing is thus, in effect, no more unusual than ignoring Speech Act Empathy in calculating the trigger potential in a given sentence (as is regularly done in Kuno (1987) when the sentence contains only third person referents).

20.

I do not consider here the pragmatics of another option, namely that of omitting the possessive altogether; note, however, that Krivinkova (1995) argues convincingly that the choice between a zero and a non-zero possessive in NPs denoting bodyparts is motivated by Empathy.

21.

Note the preverbal position of the goal of action in this utterance. This word order correlates with sentential stress and encodes the perception of a lack of interlocutor distance on the part of the speaker. These and other features of colloquial speech characteristic to informal relationships are discussed in detail in Yokoyama (1993a, 1993b, and 1994).

22.

The rule proposed here apparently does not apply to English, where the situation seems to be reverse: reflexivization fails to apply under coreference with (at least) a first person subject only when the speaker takes the point of view of others than him-/herself. This is corroborated by cases like (a) and (b); (a) (b)

A boy just reprimanded by his father for bad table manners responds: I didn’t bring me up! A person standing in front of a mirror looking at himself says: Hmm... I think I like me!

I believe that in (a), the boy adopts his father’s point of view, urging him to assume the responsibility of bringing “me” up; in (b), the man does not just say that he likes his personality (in which case he would use a reflexive), but instead says that he likes the looks of that person whose reflection he sees in the mirror (who happens to be “me”). In both cases, I believe, it is the speaker’s extrinsic view upon himself that motivates the pronominal. 23.

The inspiration for this explanation comes from Rugaleva (1980a); although her conclusions differ from mine, I owe to her thoughtful paper the basic idea of considering the deictic nature of pronouns in general. Note that Zribi-Hertz (1989) also attributes the basic difference in behaviour of reflexives and non-reflexives to the difference in the “endophoric” vs. deictic nature of the two classes. She does not, however, elaborate on pronominals.

80

OLGA T. YOKOYAMA

24.

Recall example (12), showing interaction of reflexivization with a subjective adjective; cf. also Yokoyama (1978).

25.

Or syntactic, if one believes in the syntactic origin of reflexives.

26.

To the extent that the explanation suggested above refers to different communication modes, it must be noted that the perception of interlocutor distance, while in principle a universal anthropological phenomenon, may well differ as to the actual length of the radius of what is considered close or distant, in a given culture or by a given individual. It remains to be seen to what extent what is true for Russian applies to other linguistic and/ or cultural communities.

References Baker, Carl L. 1995 “Contrast, discourse prominence, and intensification, with special reference to locally free reflexives in British English”. Language 71: 63-101. Cantrall, William R. 1974 Viewpoint, reflexives, and the nature of noun phrases. The Hague: Mouton. Chomsky, Noam 1981 Lectures on government and binding. Dordrecht: Foris. Chvany, Catherine V. 1973 On the syntax of be-sentences in Russian. Cambridge, MA: Slavica Publishers. Finkel’, Aleksandr M. and Nikolaj M. Bazenov 1960 Kurs sovremennogo russkogo literaturnogo jazyka. Kiev: Radjanska Skola. Jakobson, Roman 1957/1971 “Shifters, verbal categories, and the Russian verb”. In R. Jakobson, 130-147. Jakobson, Roman 1971 Selected writings II: Word and language. The Hague-Paris: Mouton. Klenin, Emily 1974 Russian reflexive pronouns and the semantic roles of noun phrases in sentences. Ph. D. Dissertation, Princeton University. Koster, Jan and Eric Reuland (eds) 1991 Long-distance anaphora. Cambridge: Cambridge University Press. Krivinkova, Katerina 1995 “Lacking empathy: zero-possessives and inalenables”. In O.T. Yokoyama (ed), 67-80. Kuno, Susumu 1972 “Pronominalization, reflexivization, and direct discourse”. Linguistic Inquiry 3: 161-195.

THE SPEECH ACT EMPATHY HIERARCHY AND RUSSIAN POSSESSIVES

81

Kuno, Susumu 1987 Functional syntax: Anaphora, discourse and empathy. Chicago: The University of Chicago Press. Kuno, Susumu and Etsuko Kaburaki 1975/1977 “Empathy and syntax”. In S. Kuno(ed), 1-73. Also Linguistic Inquiry 8: 627-672. Kuno, Susumu (ed) 1975 Harvard studies in syntax and semantics I. Cambridge, MA: Department of Linguistics, Harvard University. Kuno, Susumu (ed) 1980 Harvard studies in syntax and semantics III. Cambridge, MA: Department of Linguistics, Harvard University. Kuno, Susumu and Ken-ichi Takami 1993 Grammar and discourse principles: Functional syntax and GB theory. Chicago: The University of Chicago Press. Maguire, Robert A. and Alan Timberlake (eds) 1993 American contributions to the XIth international congress of Slavists. Columbus, OH: Slavica. Matejka, Ladislav (ed) 1976 Sound, sign, and meaning: Quinquagenary of the Prague linguistic circle. Ann Arbor, MI: Michigan Slavic Publications. McCawley, James D. 1976 “Notes on Jackendoff’s theory of anaphora”. Linguistic Inquiry 7: 319341. Peskovskij, Aleksandr M. 1914/38 Russkij sintaksis v naucnom osvescenii, 6th ed. Moskva: Gos. Ucebnopedagogiceskoe izd. Reinhart, Tanya and Eric Reuland 1991 “Anaphors and logophors: An argument structure perspective”. In J. Koster and E. Reuland (eds), 283-321. Rugaleva, Anelya 1980a “Struktura dejkticeskogo znaka: pritjazatel’nye mestoimenija v russkom jazyke”. Russian Language Journal XXXIV: 25-42. Rugaleva, Anelya 1980b “Semanticeskoe razgranicenie vozvratnyx i licnopritjazatel’nyx mestoimenij v russkom jazyke”. Russian Language Journal XXXIV: 51-71. Semeka-Pankratov, Elena (ed) 1995 Studies in poetics: Commemorative volume for Krystyna Pomorska (1928-1986). Columbus, OH: Slavica Publishers. Timberlake, Alan 1980 “Coreference conditions on Russian reflexivization”. Language 56: 777796. Uspenskij, Boris A. 1970 A poetics of composition: The structure of the artistic text and typology of a compositional form, transl. V. Zavarin and S. Wittig. Berkeley, CA: University of California Press.

82

OLGA T. YOKOYAMA

Veyrenc, Jacques 1980 Etudes sur le verbe russe. Paris: Institut d’études slaves. Waugh, Linda R. and Stephen Rudy (eds) 1991 New vistas in grammar: Invariance and variation, Current issues in linguistic theory 49. Amsterdam and Philadelphia: John Benjamins. Yokoyama, Olga T. 1975 “Personal or reflexive? A functional analysis”. In S. Kuno (ed), 75-112. Yokoyama, Olga T. 1978 “Reflexivization into adjectival reduced relative clauses in Russian”. Folia Slavica 2: 366-75. Yokoyama, Olga T. 1980 “Studies in Russian functional syntax”. In S. Kuno (ed), 451-778. Yokoyama, Olga T. 1987 Discourse and word order. Amsterdam and Philadelphia: John Benjamins. Yokoyama, Olga T. 1991 “Shifters and non-verbal categories of Russian”. In L.R. Waugh and S. Rudy (eds), 363-386. Yokoyama, Olga T. 1993a “Oppozicija svoj-cuzoj v russkom jazyke”. In R.A. Maguire and A. Timberlake (eds), 452-459. Yokoyama, Olga T. 1993b/1995 “Narrative intonation in Zoscenko”. In O.T. Yokoyama (ed), 214242. Also in E. Semeka-Pankratov (ed), 559-588. Yokoyama, Olga T. 1994 “Iconic manifestation of interlocutor distance in Russian”. Journal of Pragmatics 22: 83-102. Yokoyama, Olga T. (ed) 1993 Harvard studies in Slavic linguistics II. Cambridge, MA: Slavic Linguistics Colloquium, Harvard University. Yokoyama, Olga T. (ed) 1995 Harvard studies in Slavic linguistics III. Cambridge, MA: Slavic Linguistics Colloquium, Harvard University. Yokoyama, Olga T. and Emily Klenin 1976 “The semantics of ‘optional’ rules: Russian personal and reflexive possessives”. In L. Matejka (ed), 249-70. Zemskaja, Elena A. and L.A. Kapanadze (eds) 1978 Russkaja razgovornaja rec: teksty. Moskva: Nauka. Zribi-Hertz, Anne 1989 “Anaphor binding and narrative point of view: English reflexive pronouns in sentence and discourse”. Language 65: 695-727.

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS

83

Aspects of Hypothetical Meaning in Japanese Conditionals Wesley M. Jacobsen Harvard University

1.

Introduction

Studies of conditional constructions across a variety of languages reveal a wide range of meanings among morphological conditionals making it difficult to define precisely where conditional meaning stops and non-conditional meaning begins. Traditionally it has been assumed that for a conditional construction to count as a genuine conditional it must at least be hypothetical in meaning, where “hypothetical” is taken to refer to propositions whose truth value is, at the time of the speech event, either indeterminate or inconsistent with what is known to have actually been the case. Comrie (1986), for example, argues that while a conditional may exhibit a variety of degrees of hypothetical meaning, it never involves the factuality of either of its constituent propositions. That is, from a conditional of form “If P, then Q”, neither the truth of P nor the truth of Q can be determined solely from the conditional itself. Japanese presents somewhat of an anomaly from this perspective, in that conditional forms are frequently used to express not only situations which are certain to occur in the future, but even situations which have actually occurred in the past. This can be illustrated in differing uses of the conditional verbal inflection TARA, which alongside its “standard” use as a hypothetical conditional, illustrated in (1), can be used to mark clauses where hypothetical meaning is low, as in (2), or absent altogether, as in (3): (1)

Sindo 6 no zisin ga okiTARA kono tatemono wa hookai suru.1 “If an earthquake of magnitude 6 occurs, this building will collapse.”

84

WESLEY M. JACOBSEN (2) (3)

Uti ni kaetTARA mata denwa suru. “When I get home I’ll call again.” Uti ni kaetTARA haha kara kozutumi ga todoiteita. “When I got home a package had arrived from my mother.”

Examples (2) and (3) appear at first glance to exhibit the same kind of overlap between conditional and temporal meaning as is familiar in English from the paraphrase relationship existing between if and when(ever) in certain contexts.2 (4)

If/when(ever) I stand up quickly I feel dizzy.

Such overlap is permitted in English, though, only in contexts where some degree of hypothetical meaning is preserved: the conditional clause in examples such as (4) cannot be seen to express an event which actually occurred in the past or will occur in the future. In that sense, examples (2) and (3) represent a significant departure from conditional meaning as traditionally understood, so much so as to call into question their very membership in the class of conditional constructions.3 To dismiss non-hypothetical uses of conditional forms outright as irrelevant to conditional meaning is, however, to beg the question of what constitutes conditional meaning in the first place and to ignore any insights which might be gained into such meaning from the morphology of Japanese conditionals, which encompass both hypothetical and non-hypothetical uses. Recent studies of the non-hypothetical use of conditional forms in past contexts in Japanese, motivated by the pioneering work of Kuno (1973), have in fact revealed constraints on the meaning and use of such constructions which, when properly understood, shed light on the very cognitive mechanisms which make possible the expression of conditional meaning. Building on the groundwork laid by these studies, we will consider in this paper how conditional meaning might be conceived of independently from the parameter of hypothetical meaning. Assuming a framework which does not take hypothetical meaning as the defining feature of conditional meaning, we will then proceed to ask the question of what features, semantic and pragmatic, are responsible for giving rise to hypothetical meaning in a language such as Japanese where the morphology does not itself strictly distinguish between the presence or absence of such meaning.

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 2.

85

Temporal and Conditional Meaning

One of the long-standing debates in the philosophical and linguistic literature has been whether or not conditional meaning in natural language can be adequately described by the truth function table for material implication in propositional logic. According to the truth table for material implication, a conditional sentence is false only in case the antecedent is true and the consequent false, and is true for any other combination of truth values in the antecedent and consequent. A well-known problem with the truth-functional account is its inability to account for the intuition that there must exist some connection between the antecedent and consequent, regardless of what their truth values might be, for a conditional sentence to be considered true. Conditional sentences such as (5) where no apparent such connection exists are not intuitively considered to be true, despite the fact that the truth of each constituent clause should, according to the truth table, guarantee the truth of the sentence as a whole: (5)

If 2 plus 2 equals 4, then Robert Dole was the Republican presidential candidate in the 1996 American presidential election.

A related, but perhaps more serious, problem is the fact that under the truth-functional approach the truth value of a conditional can only be evaluated at the point in time that the truth values of the constituent clauses become known, whereas it is common in everyday life to judge the truth (correctness) of conditional sentences at a point in time before the truth values of the constituent clauses are actually known. This is the case with conditionals referring to future eventualities, which represent probably the most common use of conditional sentences in everyday language. An engineer with relevant expert knowledge of the structure of a building would, for example, be perfectly justified in stating that the sentence in (6) is false, even though he had no certain knowledge of future events allowing him to say that an earthquake of magnitude 6 will occur and that the building will not collapse, the only case in which the truth-functional approach would yield a false value for (6):4 (6)

If an earthquake of magnitude 6 occurs, this building will collapse.

What the engineer would be denying is in fact nothing more than the existence of a connection between an earthquake occurring and the collapse of the building.

86

WESLEY M. JACOBSEN

A third problem with the truth-functional analysis is its inability to account for the temporal asymmetry which is typically seen to occur between the situations expressed by the antecedent and the consequent clauses. Consider a situation where, on the day following the utterance of (6), the building collapses and then, several hours later, an earthquake of magnitude 6 occurs. Such circumstances would not intuitively be seen as confirming the truth of (6), much as the truth table would yield a true value for the conditional in that case. The intuition here is that the situation expressed in the consequent must follow in some sense from a prior situation in the antecedent, typically prior in a temporal sense. This points again to the existence of some kind of connection between the content of the antecedent and consequent clauses. Given these considerations, it is now commonly accepted that a stricter requirement must be imposed on the truth of conditional sentences than that adopted in the truth-functional approach, one which makes reference to the existence of just such a connection between the two clauses. The result is that, while conditional meaning in natural language is considered to be consistent with the truth table to the extent that a conditional is never true when the antecedent is true and consequent false, conditionals with false antecedents or true consequents are not automatically considered true as would follow from the truth table. Exactly how to formulate the connection in question is not, however, a simple matter. One approach, that adopted in Comrie (1986), is to make it a requirement that the content of the antecedent be interpretable as the cause of the content of the consequent. The notion of causation is itself a difficult one to define,5 but is intuitively helpful in explaining the unacceptability of examples such as (4) and addresses the various problems pointed out above with the truth-functional approach, particularly the ability we have to evaluate the truth of conditionals in future contexts where the truth values of the constituent clauses are still unknown. The cause-and-effect analysis is still unfortunately unable to account for the full range of use of natural conditionals, as examples are common of conditionals where the antecedent and consequent do not fall into a relationship of cause and effect in any intuitively obvious way. (7) (8) (9) (10)

If it will amuse you, I’ll tell you a joke. If the red light is on, the electricity is flowing. If you’re hungry, there’s food in the fridge. If a>b and a,b>1, then a2 > b2.

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS

87

(7) is an example from Comrie (1986) where the consequent appears to express the cause and the antecedent the effect rather than the other way around. As Comrie points out, though, the antecedent may still be seen as expressing a motivation for the action in the consequent, and thus be considered as expressing a cause of some sort. Such a tack is not readily available in (8): the only apparent causal relationship here appears to be in the direction of consequent to antecedent. In (9), there appears to be no causal relationship in either direction between being hungry and food being in the fridge. The same can be said for the constituent clauses in (10). A relationship of causation directed from the antecedent toward the consequent is therefore too stringent a requirement to place on the meaning of conditional sentences. I maintain nevertheless that there is a fundamental asymmetric relationship between antecedent and consequent present even in the examples (7)-(10) whereby the state of affairs in the consequent is dependent in some sense on a prior state of affairs in the antecedent. At its most fundamental (and, I assume, cognitively basic) level, the dependency relationship is one occurring between the events expressed by the constituent propositions as phenomena, requiring that the antecedent proposition express a state of affairs which is temporally prior to that of the consequent. Such is the case with the earthquake example cited earlier in (6). This kind of relationship between clauses is one which I will henceforth refer to as “phenomenal contingency.” The dependency relationship can also be seen to occur in conditional sentences at a more abstract level where the phenomena forming the content of the constituent propositions are not at issue, but rather higher order events or acts which refer in some way to those propositions. Such events or acts can typically be represented by higher predicates which have the effect of subordinating the constituent propositions. In such cases it is the event or act expressed by the higher predicate which is temporally prior to or temporally dependent on the state of affairs expressed in its partner clause. This includes conditionals which have been termed “epistemic,” where the dependency relationship is not between states of affairs expressed in the constituent propositions but rather between acts of knowing those states of affairs, and “speech act” conditionals, where the dependency relationship involves any of a number of acts of affirming, promising, judging to be true, informing the hearer about, etc., the proposition in the antecedent or consequent.6 (8) and (9), cited above, are examples, respectively, of an epistemic and speech act

88

WESLEY M. JACOBSEN

conditional. Note how these can be paraphrased in such a way that higher predicates subordinating the constituent propositions are made explicit: (8') (9')

If (one sees that) the red light is on, then (one knows that) the electricity is flowing. If (it is the case that) you’re hungry, (then I offer you the information that) there’s food in the fridge.

In (7) and (10), likewise, the dependency relationship occurs at a level higher than that of phenomenal occurrence of the events expressed. It is not clear that these can be neatly classified into “epistemic” or “speech act” types, but the same mechanism is clearly at work as in the case of (8) and (9): (7')

If (it is the case that) it will amuse you, I (hereby decide that) I will tell you a joke. (10') If (it is the case that) a>b and a,b>1, then (it follows that) a2 > b2. As the paraphrase above shows, in (7) it is a decision to bring about a certain state of affairs that is dependent on prior knowledge of the information in the antecedent. In (10), the process of deducing the proposition in the consequent is dependent upon the prior acceptance of the proposition in the antecedent. A precedence relationship between antecedent and consequent may therefore be seen to obtain for all the examples in (7) through (10), albeit at a level higher than that of the phenomena expressed in the constituent propositions themselves. This is in contrast to (6), where the precedence relationship is clearly at the temporal, phenomenal level, making it awkward to paraphrase this conditional in a manner parallel to (7)-(10): (6') ??If (it is the case that) an earthquake of magnitude 6 occurs, (then I assert that) this building will collapse. The dependency relationship seen in examples (7) - (10) illustrates what I will henceforth refer to as “meta-propositional contingency.” The connection between the antecedent and consequent in conditional clauses, while not necessarily causal, can thus be characterized across the board as that of an asymmetric relationship of contingency, with the antecedent expressing a state of affairs which precedes that of the consequent, either at the phenomenal or meta-propositional level, and upon which the state of affairs in the consequent is dependent. Though the precedence relationship is a literally temporal one only when manifested at the phenomenal level, even the precedence relationship observed at the meta-propositional level can be

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS

89

seen as fundamentally temporal in nature, reflecting the unidirectional flow of time in which information is processed in the real world.

3.

Meta-Propositional Conditionals in Japanese

The distinction between phenomenal and meta-propositional contingency is not reflected in the morphology of English conditionals, where both meanings are expressed by if, but turns out to play a significant role in describing the conditions of use on different conditional types in Japanese. There are four basic conditional forms in Japanese, which we will refer to according to their morphological shapes as TO, TARA, BA, and NARA conditionals.7 Each of these has uses as a marker of meta-propositional contingency, as observed in the following example of the epistemic variety, where all four forms are possible: (11) Akai denki ga tuiteiru TO/tuiteiTARA/tuiteireBA/tuiteiru NARA denryuu ga nagareteiru. “If the red light is on, the electricity is flowing.” The consequent clause in (11) can, as we observed for similar examples in English, be paraphrased as in (12) to express overtly its epistemic character. This paraphrase works with any of the four conditional forms in (11): (12) ...denryuu ga nagareteiru koto ga wakaru. “...you can tell the electricity is flowing.” The NARA conditional is nevertheless distinct from the other three in requiring that the contingency relationship be of the meta-propositional sort.8 Examples (7) to (10), offered in Section 2 as illustrations of the range of meaning encompassed by meta-propositional contingency, all find natural expression in NARA conditionals in Japanese: (13) Waratte kureru NARA zyoodan o itte ageyoo. “If you will laugh, I will tell you a joke.” (14) Onaka ga suiteiru NARA reizooko ni tabemono ga haitteiru kara. “If you’re hungry, there is food in the fridge, so ... (help yourself).” (15) a to b ga 1 yori ookiku, katu a ga b yori ookii NARA, a2 wa b2 yori ookii. “If a>b and a,b>1, then a2 > b2 .”

90

WESLEY M. JACOBSEN

Where the contingency relationship is, on the other hand, of the phenomenal sort as in (16) — i.e., where there is a temporal precedence relationship holding between the events expressed by the antecedent and consequent — the NARA conditional is highly awkward, in contrast to TO, TARA, and BA: (16) Kono hon o yomu TO/yonDARA/yomeBA/??yomu NARA kitto kandoo suru daroo. “If you read this book, you will certainly be deeply moved.” We have defined meta-propositional contingency as a precedence relationship which does not involve the time of occurrence of the constituent propositions themselves, but rather the time of certain acts or events of referring to the propositions. Understood in this way, certain idiosyncratic facts about the behavior of the NARA conditional pointed out in previous studies can be accounted for readily. Kuno (1973), for example, describes the antecedent in NARA conditionals as representing a proposition which the hearer has either asserted or could potentially have asserted and the consequent as representing the speaker’s evaluation, supposition, will, resolution, request, or order. In other words, what NARA relates is not the time of occurrence of the two constituent propositions, but the time at which the information in the antecedent clause becomes available and some subsequent act on the part of the speaker in which the proposition in the consequent is affirmed, decided upon, requested, etc., in response to that information. Kuno’s condition on NARA that the event expressed in the consequent cannot be seen as dependent on the future realization of the event expressed in the antecedent is thus a prohibition on the use of NARA at the level of phenomenal (temporal) contingency. (17) *Kyuu ni tatiagaru NARA memai ga suru. “If I stand up suddenly, I feel dizzy.” Also following naturally from the meta-propositional character of NARA is the fact, again pointed out in Kuno (1973), that the temporal order of the events in the antecedent and consequent clauses of NARA may occasionally be seen to occur in an order reverse to that of the clauses in which they appear. In (18a), the most likely interpretation is one where my leaving precedes Taroo’s coming, whereas in (18b), involving the conditional form TARA, the only interpretation possible is the reverse:

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS

91

(18) a. Taroo ga kuru NARA boku wa kaeru. “If Taroo is coming, then I’m leaving (i.e., if it is the case that Taroo is coming, I hereby decide that I will leave.)” b. Taroo ga kiTARA boku wa kaeru. “If/when Taroo comes, I’ll leave.” This follows naturally from the difference between TARA and NARA in the level at which the contingency relationship is seen to hold — between propositions as events in the case of TARA, but between acts of referring to those events in the case of NARA, whatever the actual temporal order of the events themselves may be. Several facts about the formal character of NARA constructions provide additional support for the meta-propositional analysis proposed here. First, NARA is the only conditional form which allows a contrast between past and non-past tense in the antecedent clause:9 (19) a. Syatyoo ga sanzi no bin de tootyaku suru NARA sorosoro mukae ni itta hoo ga ii. “If the president is arriving on the 3:00 flight, we’d better set out soon to meet him.” b. Syatyoo ga sanzi no bin de tootyaku sita NARA sorosoro mukae ni itta hoo ga ii. “If the president arrived on the 3:00 flight, we’d better set out soon to meet him.” As it is not the time of the event expressed in the antecedent, but rather the time at which information about that event becomes available, that is critical to the precedence relationship established by NARA, whether the event in question itself is past or non-past is irrelevant to this precedence relationship. Even more importantly, NARA clauses are intuitively felt to be, and can be paraphrased by, nominalized versions of the antecedent clause where the complementizer NO occurs between the clause-final predicate and NARA.10 The predicates in the antecedent clauses in (19a) and (19b) can, for example, be replaced respectively by tootyaku suru NO NARA “if it’s the case that he will arrive” and tootyaku sita NO NARA “if it’s the case that he arrived” with no change in meaning. The subordinating role played by NO here can be functionally correlated with the act of referring to the subordinated proposition, thereby shifting the level of the contingency relationship from phenomenal to meta-propositional.

92

WESLEY M. JACOBSEN

The role of complementizers in effecting such a shift in meaning can be observed in a variety of conditional forms besides NARA. The conditional forms NO DATTARA11 and TO SUREBA,12 for example, exhibit a consistently meta-propositional contingency relationship which can be attributed to the nominalizing function of the sentential complementizers NO, in the former, and TO, in the latter. As illustrated in the following example, it is again the time of the act or event of referring to the antecedent proposition, rather than the time of the event encoded in that proposition, that forms the basis for the contingency relationship in all these constructions:13 (20) 10zihatu no tokkyuu ni noru NARA/NO DATTARA/TO SUREBA koko o 8zihan ni denakerebanaranai. “If (it is the case that) we will catch the special express leaving at 10:00 (= if we are to catch the special express leaving at 10:00), we have to leave here at 8:30.)” Although not all conditional constructions expressing a meta-propositional contingency relationship involve complementizers, (recall the earlier example (11) of epistemic-type meaning even in the absence of such complementizers), the presence of such complementizers appears to be characteristic of meta-propositional meaning at least of the speech act variety. For conditional forms lacking such complementizers, such as TO and TARA, the normal interpretation is one involving a temporal precedence between the events expressed in the antecedent and consequent. Such an interpretation is also possible with BA, though less consistently so than with TO or TARA. (21) involves a clear temporal sequence between antecedent and consequent events, no matter which of the three conditional forms TO, TARA, or BA is used: (21) Kono kusuri o nomu TO/nonDARA/nomeBA genki ga deru. “If you take this medicine, you will feel better.” In (22) and (23), by contrast, the choice of the BA conditional appears to result in a relaxation of the strict precedence requirement that operates with TARA. Thus BA appears to allow an interpretation where Taroo goes before Hanako in (22) and to leave open the possibility in (23) that I will tell you the secret before you pay me, as long as I am assured that you will eventually pay me. TARA, by contrast, requires that the occurrence of the event in the antecedent clause temporally precede that of the consequent clause in both

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS

93

cases: (22) Hanako ga itTARA/ikeBA Taroo mo iku daroo “If Hanako goes, Taroo will go, too.” (23) Okane o dasiTARA/daseBA himitu wa osiete yaru. “If you pay me, I’ll tell you the secret.” The meaning of BA therefore overlaps with that of NARA in certain cases, allowing the occurrence of the consequent event to be dependent on the accessing of the information in the antecedent clause, rather than on the occurrence of the antecedent event itself. We will see in Section 5 that this meta-propositional tendency in BA can be correlated with other properties distinguishing BA from TARA and TO on the aspectual and discourse levels as well.

4.

Conditionals Expressing Actual Occurrence

As seen in Section 2, the constituent clauses of a conditional sentence are connected by a relationship of contingency whereby the state of affairs in the consequent clause is seen as being dependent on the state of affairs in the antecedent clause, either at the phenomenal or meta-propositional level. This relationship we take to be not just an incidental feature of the meaning of conditionals in natural language, but the central and defining feature of their meaning. As there is no necessary connection between such a relationship and the question of whether the antecedent state of affairs actually occurs or not, there is in principle no problem with viewing hypothetical meaning as an independent meaning parameter which often happens to, but does not necessarily, attach to conditional sentences in natural language. While certain languages may opt to encode the presence or lack of hypothetical meaning in their system of conditionals, as does English with if versus when, other languages may opt to ignore that distinction, choosing instead simply to encode the contingency relationship between clauses, as does Japanese. Stripped of their hypothetical meaning, conditional sentences expressing a contingency relationship of the phenomenal sort might appear to converge totally in their meaning with causal or temporal constructions. Causal meaning is nevertheless insufficient to explain such a contingency relationship even at the phenomenal level, as shown by examples such as the following:

94

WESLEY M. JACOBSEN (24) If/whenever I pick up the phone, I get a busy signal. (25) If/whenever the sunset is bright red, the weather is sunny the following day.

It does not follow our intuitive sense of what counts as causation to say that my picking up the phone causes a busy signal to occur, or that a red sunset causes the weather to be sunny the next day. Our ability to posit a causal relationship between two events involves a higher-order capacity for judging what counts as a necessary and sufficient condition for the caused event to occur. For conditional sentences, it is sufficient that there exist a pattern of cooccurrence such that an instance of the first event is consistently followed by an instance of the second. The connection made between two events in conditional constructions is thus fundamentally no more and no less than the speaker’s judgment from past experience that one event will be accompanied by occurrence of the other in a fixed order in time, and that possible future instances of the one event will be similarly followed by an instance of the other. Conditional meaning does not place any further constraints on the nature of the connection beyond that, leaving the question of how the events might be connected to empirical experience, with the endless possibilities that presents of varieties of co-occurrence relationships still to be encountered. The fundamentally temporal, as opposed to causal, nature of this connection is reflected in the fact that conditional constructions, once stripped of their hypothetical meaning, are typically better paraphrased by when constructions than by because constructions. In (26), for example, (b) is clearly closer to the non-hypothetical content of (a) than is (c): (26) a. If you push this button, the light will go on. b. When you push this button, the light will go on. c. ?Because you (will) push this button, the light will go on. The affinity between temporal and conditional meaning can also be seen in the close paraphrase relationship existing between if and whenever in English examples such as (24) and (25). This does not mean, though, that all when constructions are automatically conditional in character. The contingency relationship present in conditional constructions always involves an asymmetric relationship of antecedence and consequence, whereas when constructions merely require that two events be proximate in time. When constructions which express simultaneous, rather than sequential meaning, such as (27),

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS

95

lack the requisite sense of asymmetric ordering necessary for a contingent interpretation: (27) When the news hits the stands tomorrow, he will be gone on a business trip. Nor do when constructions always carry an interpretation that the two events are connected through a previously-established pattern of repeated occurrence. In the past tense, for example, when constructions typically represent the confluence of two events as a unique, first-time occurrence. Note how (28) and (29) lack the sense of a connection based on a pattern of previous occurrence which is present in examples such as (26b): (28) When the earthquake occurred, I was studying in my office. (29) When I entered college, the draft was abolished. Still, when the occurrence of the events is sequential, the possibility exists of a latent contingent relationship which has simply not been experienced previously, and is therefore being discovered for the first time. (30) When I pushed the button, a light went on. The existence of a relationship of contingency thus arises most easily in contexts which support the interpretation of a previous pattern of co-occurrence between two event types, either overtly, as in generic contexts such as (24) and (25), or covertly, as in future contexts such as (26a) and (26b), where a previous pattern of co-occurrence constitutes the grounds for making predictions about future developments, whether hypothetical or not. In Japanese, generic and one-time occurrence of a contingency relationship are, as a general rule, expressed in distinct conditional forms, TO for the former and TARA for the latter: (31) Kaidan o kakeagaru TO iki ga kireru. “If/whenever I run up the stairs, I run out of breath.” (32) Kaidan o kakeagatTARA iki ga kireru (yo). “If you run up the stairs, you will run out of breath.” Note that, even though no subjects are specified in (31) or (32), the generic nature of TO causes (31) to be most naturally interpreted as having a firstperson subject, reflecting a connection arising from the speaker’s own past

96

WESLEY M. JACOBSEN

experience of the events referred to, whereas (32), with TARA, is more naturally interpreted as a warning to another person who is about to engage in a particular instance of the act referred to in the antecedent. In either case, the ability to posit a conditional relationship between the two events arises from a connection forged through repeated experience of the co-occurrence of similar event types in the past. The use of TARA and TO to express actual past events, on the other hand, appears to run directly counter to this pattern. In such cases these forms express the first-time co-occurrence of two unique events without any sense of a previous pattern of repeated co-occurrence, parallel to the use of when in (30). (33) Botan o osiTARA/osu TO akari ga tuita. “When I pushed the button, a light went on.” The observation made in Kuno (1973) that the content of the consequent clause in such constructions is typically the source of surprise to the speaker underlines the unique nature of the co-occurrence relationship. As with the English when constructions in (28) - (30), it might seem that the lack of any pattern of co-occurrence between the two events would disqualify such constructions from being conditional in meaning as we have characterized such meaning so far. A peculiar constraint applies, however, to past-tense uses of TARA and TO which I have argued in previous papers points to the very experiential foundations of conditional meaning.14 In its essence, the constraint requires that the speaker have no control over the co-occurrence of the two events in question. Hasunuma (1993), for example, characterizes the relationship between the events as one where the consequent event expresses the content of some new awareness in the speaker following occurrence of the antecedent event. Accordingly, the consequent event cannot be one which the speaker planned or volitionally brought about. In the case of TARA, the first event must either be an action performed first-hand by the speaker or otherwise directly experienced by the speaker; in the case of TO it is sufficient that the speaker be in a position to objectively observe the sequence of both events from a consistent vantage point. The following examples from Hasunuma (1993) illustrate this difference: (34) Donguri wa korokoro korogaru TO/??korogatTARA ike ni otita. “When it had rolled a ways, the acorn fell into the lake.”

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS

97

(35) Ke no seetaa o sentakuki de aratTARA/??arau TO kirarenakunatta. “When I washed the wool sweater in the washing machine, I couldn’t fit into it any more.” In (34), the first clause involves a third-person subject, thus rendering the construction incompatible with TARA. In (35) the events are such as to make it difficult to interpret them as having been observed from a consistent vantage point (several other events are likely to have intervened between washing the sweater and putting it on, such as drying it, folding it, etc., creating a gap in time and location between the vantage points relevant to the first and second clauses), thus resulting in awkwardness with TO. In either case, the second clause represents a new or unexpected development for the speaker. The lack of control exercised by the speaker over the occurrence of the second event in past-tense TARA and TO constructions is both epistemic and volitional: epistemic in the sense that the speaker has no prior knowledge that the event is to occur, and volitional in the sense that the speaker cannot willfully bring about the event . This is in direct contrast to what we find in future conditional constructions, where the speaker always exercises one of these two forms of control over the occurrence of the second event. Epistemic control, illustrated in (36), is the experientially-rooted knowledge that the second event will occur given occurrence of the first event. In cases of volitional control, illustrated in (37), the occurrence of the second event is assured by virtue of the speaker’s own power to bring it about: (36) Kyuu ni tatiagaru TO memai ga suru. “If I stand up suddenly, I feel dizzy.” (37) Uti ni kaetTARA mata denwa suru. (=2) “When I get home I’ll call again.” While expressing a similar relationship of asymmetric co-occurrence between consequent and antecedent events, then, past and future conditional constructions exhibit precisely reverse patterns in terms of the control exercised by the speaker over the co-occurrence relationship. This might be taken to indicate a basic incompatibility in meaning between the two construction types, but in reality mirrors complementary aspects of the experiential prototype underlying our ability to express conditional meaning. Our ability to posit a relationship of contingency is, as we have argued, rooted in the experience of a pattern of repeated co-occurrence between two event types.

98

WESLEY M. JACOBSEN

Such a pattern cannot, however, extend infinitely into the past, but must originate at some point in a primal experience in which the co-occurrence is first observed, an experience in which the occurrence of the second event following the first will by definition be a new, unexpected development. This is precisely the scenario expressed in past-tense uses of TARA and TO conditionals in Japanese. The ability to express hypothetical meaning finds its origins in actual experience, and it is this fundamental connection between hypothetical and non-hypothetical meaning which is given unified expression in the morphology of Japanese conditionals.

5.

Parameters of Hypothetical Meaning in Japanese

Given that conditional meaning can be adequately characterized without reference to hypothetical meaning, and that many uses of conditional forms in Japanese in fact do not carry such meaning, the question now arises of what kinds of conditions give rise to hypothetical meaning where it does occur in Japanese conditionals. Contrary to the seemingly dichotomous opposition between hypothetical and actual meaning as lexicalized in English if and when, the picture which emerges of hypothetical meaning in Japanese is that of a scalar property which is present in greater or lesser degree according to how far a particular event is seen to be removed from the real world as it exists for the speaker. At the most “real” end of this scale are event types forming a prototype characterized by actual, phenomenal occurrence at a unique point in time. Departures from this prototype can be seen along various dimensions of meaning, the number and degree of which govern the degree to which hypothetical meaning attaches to a particular conditional clause. We will identify here four such dimensions of meaning, each independent from the other and from the element of contingency which defines conditional meaning, and yet interacting closely in defining how close or far the proposition in the conditional clause is to the set of propositions which are accepted as fact by the speaker. 5.1 Hypothetical Meaning and Tense Perhaps the most apparent dimension of meaning affecting the hypothetical character of a proposition — i.e., whether or not it will have a determinate

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS

99

truth value — is the tense of the proposition. There is a basic asymmetry arising from the unidirectional flow of time in the determinacy of the truth value of past- and present-tense propositions, on the one hand, and futuretense propositions on the other. The truth value of past/present-tense propositions can in principle be judged on the basis of their consistency with the record of what actually happened in the past or is happening at the present, limited only by the extent of the speaker’s knowledge of the actual past and present. Future-tense propositions have no such standard against which their truth value can be judged and are therefore inherently indeterminate in truth value, which is to say hypothetical. It is nevertheless possible to speak of varying degrees of hypothetical meaning in future contexts according to the level of confidence which the speaker attaches to the probability of occurrence of future events of various types. Such distinctions are encoded in English, once again, by when vs. if but are not formally marked in Japanese conditionals, as seen in the use of the same TARA form in each of the following examples: (38) Uti ni kaetTARA sugu denwa o suru. “When I get home, I’ll phone you right away.” (39) Taroo ni atTARA sono koto o hanasite oku. “If/when I see Taroo, I’ll talk to him about that.” (40) Ame ga hutTARA siai wa tyuusi ni naru. “If it rains, the match will be canceled.” (38) does not, in most normal contexts of utterance, entail any doubt as to whether the speaker will actually arrive at his/her home, although a hypothetical reading could be imposed by the addition of the adverbial mosi “if” (likely to be understood in a comical sense in this example). (39) and (40), by contrast, represent successively higher degrees of hypothetical meaning in most normal contexts. The degree of hypotheticality in examples (38) - (40) is inversely related to the strength of the expectation that the event in the antecedent clause represents a “normal” future outcome of the present state of affairs, assuming a standard unfolding of events (as determined from past experience) in which no unanticipated events such as accidents intervene. Typically, this expectation is stronger in contexts where the speaker is seen to exercise a high degree of control over the occurrence of the antecedent event. The existence of such an expectation is of course governed to a large extent

100

WESLEY M. JACOBSEN

by pragmatic factors such as the speaker’s beliefs and contextual elements which either support or fail to support the occurrence of the antecedent event as an accepted fact. The level of hypothetical meaning is not a purely pragmatic matter, though, as semantic factors such as the aspectual character of the predicate in the antecedent also play a role in determining the level of hypothetical meaning, as will be discussed in Section 5. 2. Such variations in the degree of hypothetical meaning are reflected to some extent in the choice of conditional form in Japanese. While TARA and TO are capable of being used either in future contexts or in actual, past-tense contexts such as those discussed in Section 4, BA correlates with a higher level of hypothetical meaning and is disallowed in past-tense contexts of actual occurrence in modern colloquial Japanese.15 All three forms TO, TARA, and BA are therefore acceptable in the future conditional in (41a) but only TO and TARA in its past-tense counterpart in (41b): (41) a. Kusuri o nomu TO/nonDARA/nomeBA zutuu ga osamaru. “If/when you take some medicine, your headache will subside.” b. Kusuri o nomu TO/nonDARA/*nomeBA zutuu ga osamatta. “When I took some medicine, my headache subsided.” In a given future context, BA is typically associated with a lower probability of occurrence of the event it marks — i.e., a higher degree of hypothetical meaning — than TARA, as reflected in the fact that a when paraphrase in English is only possible with TARA, not BA, in (42) and (43): (42) Tomodati ga dekireBA/dekiTARA gakkoo mo tanosiku naru desyoo. “If/when she makes some friends, school will probably become more fun.” (43) Boonasu ga dereBA/deTARA atarasii kuruma o kau. “If/when I get my bonus, I’ll buy a new car.” One might expect therefore that BA would be disallowed in the case of future events that are certain to occur, such as the arrival of a particular date or point in time. Interestingly, though, it is sufficient for the use of BA that the event in question be a future one, no matter how certain, as seen in its acceptability along with TO and TARA in the following examples:

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 101 (44) Haru ni naru TO/natTARA/nareBA kono yama no naka de mo atatakakunaru. “When it becomes spring (when spring comes), it will become warm even here in the mountains.” (45) Asita no gogo 3-zi ni naru TO/natTARA/nareBA kekka ga wakaru. “When it becomes 3:00 tomorrow afternoon, we will know the results (we will know the results at 3:00 tomorrow afternoon).” Although BA also differs from TO and TARA in (44) and (45) along meaning parameters other than that of futurity of tense (e.g., in the newness of the information presented, as discussed in Section 5. 4), these examples illustrate how future events invariably carry with them a minimal degree of hypothetical meaning setting them strictly apart from past, actual events. Even for an event as highly probable as the eventual arrival of a particular calendar date or time, there is no absolute certainty of the occurrence of the event until it becomes actualized as a present or past event. Behind this fundamental dichotomy between future and past occurrence is the assumption that there is one course of events (one “history”) which constitutes the past up through the present moment, whereas there is a multiple, indeed infinite, number of possible courses of events along which the future may develop, some of which may be more probable to occur than others, but none of which is categorically distinguished from the others in being certain to occur. Possible future courses of events can be seen as forming a branching network of histories which becomes progressively more complex as time advances into the future.16 Different scenarios of how the past might have occurred make it also possible to envision multiple histories branching from particular points in the past, but with the difference that only one of all such possible histories is unique in having actually occurred. Such multiple histories nevertheless come into play in situations where the speaker either does not know what actually took place in the past or makes the assumption that the past took a different course from what it actually did. Under either of these scenarios, the speaker is able to exploit his experience of the actual past to create conditional utterances expressing a contingent outcome of unknown or non-actual past situations, just as in the case of future situations. Hypothetical meaning in this way becomes attached not only to propositions whose truth value is indeterminate, but also to those whose truth value is inconsistent with (counterfactual to) what is known to be the case.

102

WESLEY M. JACOBSEN

The conditional form BA, whose hypothetical character is otherwise inconsistent with use in past contexts (see (41b)), may be used in the past just in those cases where the antecedent event is not one which is known to have actually occurred — i.e., in those cases where multiple past histories are under consideration. Where the antecedent event is one which forms part of the speaker’s first-person experience, the meaning is typically counterfactual, as pragmatic considerations make it unlikely that the speaker would not know what actually took place in the realm of his/her own experience. Such is the case in (46) and (47), where BA and TARA are both possible: (46) Motto hayaku okireBA/okiTARA ma ni atta no ni.17 “If we had gotten up earlier, we would have made it on time.” (47) Kuru to sitte-ireBA/sitte-iTARA tyanto sita syokuzi no yooi o sita no ni. “If I had known you were coming, I would have prepared a proper meal.” Where pragmatic considerations such as the above do not force a counterfactual interpretation, such as in third-person contexts, the possibility exists of a hypothetical interpretation arising from indeterminacy of the truth value of the antecedent proposition (one where the speaker does not know if the event in question actually occurred or not). Such an interpretation is possible in (48a), where BA is attached directly to the predicate hajimaru “start.” (48a) can be paraphrased using an overtly meta-propositional conditional form such as TO SUREBA in (48b). It is common, however, for BA in past contexts to be attached to a stative predicate form, such as that in (48c), where hazimaru “start” occurs with the resultative stative affix te-i(ru) . For reasons to be discussed in the following section, the interpretation given in such cases is unequivocally counterfactual: (48) a. Yotei doori ni kaigi ga hazimareBA hirugoro owatta daroo. “If the meeting started on time, it probably finished around noon. b. Yotei doori ni kaigi ga hazimatta TO SUREBA hirugoro owatta daroo. “If (we assume) the meaning started on time, it probably finished around noon.”

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 103 c. Yotei doori ni kaigi ga hazimatte-ireBA hirugoro owatte-ita daroo. “If the meeting had started on time, it would probably have finished around noon.” Whether the interpretation received is counterfactual in the narrow sense or merely due to the indeterminate nature of the truth value of the antecedent proposition, hypothetical meaning can in all tense contexts, past and future, be seen to arise where multiple histories are envisioned as possibly occurring or having occurred, in contradistinction to the unique historical character of the actual past and present. 5.2 Hypothetical Meaning and Verbal Aspect The temporal character of an event is relevant to the degree of hypothetical interpretation it receives not only in terms of its tense — i.e., its location in time relative to the time of the speech event — but also its aspect — i.e., its inherent temporal quality. Specifically, hypothetical meaning tends to be present in lower degrees for events which are seen to occur at a unique point or interval of time, and conversely higher for events whose occurrence is disassociated from unique points or intervals of time, as in the case of stative situations and iterated events. A parallel can be seen between this aspectual parameter of hypothetical meaning and the correlation discussed in the last section between hypothetical meaning and the existence of multiple histories that allow differing scenarios to be conceptualized at particular points of time in the past and the future. In either case, hypothetical meaning correlates positively with the presence of multiple states of affairs, in the one case characterizing single moments in time (a synchronic multiplicity), in the other different moments spread across time (a diachronic multiplicity). Events which culminate in a change of state of some kind, typified by verbs in Zeno Vendler’s achievement and accomplishment categories, tend to be anchored to unique points in time in a way that situations lacking any such change, such as states, are not. This difference is indicated, among other ways, by differing patterns of co-occurrence with temporal adverbs: (49) a. I arrived in town yesterday at noon/*for five hours. b. I was in town yesterday at noon/for five hours.

104

WESLEY M. JACOBSEN

A change-of-state achievement verb, exemplified by arrive in (49a), can be predicated of an instantaneous moment, but not of an interval of time. Stative predicates, such as was in town in (49b), can be predicated either of an interval of time or of an instantaneous moment, but the moment of which a state is predicated is non-unique in that the same state could also be predicated of a potentially infinite number of other moments surrounding that moment, whereas the change of state expressed by an achievement verb such as in (49a) can only be predicated of a single, unique moment in time. Predicates in conditional clauses which lack a change-of-state component in their aspectual structure tend to correlate with a higher degree of hypothetical meaning compared to those which contain such a component, as seen in the differing hypothetical interpretation give to the future-tense TARA conditionals in (50). (50a), which contains an achievement predicate, strongly favors a low hypothetical reading — i.e., one indicating a high probability of occurrence — in contrast to the stative predicate in (50b), which can only be interpreted in a purely hypothetical sense: (50) a. Mukoo ni tuiTARA tegami o kaku. “When I arrive there I’ll write a letter.” b. Hima ga atTARA tegami o kaku. “If I have time, I’ll write a letter.” This correlation is also reflected in the fact that only “change-events,” such as those expressed in the antecedent of (50a) and (51a), can appear in past contexts of actual occurrence, whereas stative situations, such as those in the antecedent of (50b) and (51b), are rejected in such contexts:18 (51) a. Syukusya ni tuiTARA/tuku TO tomodati kara denwa ga kakatte kita. “When I arrived at the dorm, I got a call from my friend.” b. *Hima ga atTARA/aru TO tomodati kara denwa ga kakatte kita. (intended reading) “When I had time, I got a call from my friend.” Further evidence for the correlation between hypothetical meaning and stativity is seen in the tendency, pointed out in Section 5. 1, for predicates in counterfactual constructions to be expressed in a stativized form through attachment of the resultative stative form te-iru . The boldface portions of the predicates uketeiTARA “if (he) had received” and tasukatteita “(he) would

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 105 have survived” in the counterfactual construction (52) are due to this form: (52) Motto hayaku teate o uketeiTARA tasukatteita ni tigai nai. “I’m sure he would have survived if he had gotten medical treatment more quickly.” It seems reasonable to treat states of affairs which are known not to have occurred as representing a further departure from reality than states of affairs which are simply not known to have occurred or whose future occurrence is not certain — i.e., to treat counterfactual propositions as representing a more extreme degree of hypothetical meaning than propositions whose truth value is merely indeterminate. The presence of stative forms in constructions such as (52) can therefore be seen to trigger hypothetical meaning in its most extreme form. Note that the hypothetical interpretations arising in the stative antecedents of (50b) and (52) cannot be altered by pragmatic elements such as context or speaker expectation, but are set by the semantic character of the predicate. As I have argued in an earlier paper,19 the fact that stative situations are not bound in their temporal meaning to any unique moment in time, as are change-events, makes them susceptible to being interpreted as removed from real time altogether and therefore as irreal in modality, giving rise to hypothetical meaning in conditional contexts . We noted in Section 5.1 the high degree of hypothetical meaning attached to the BA form as compared to the TARA and TO forms. The correlation between stative and hypothetical meaning can be observed in a tendency for BA to be preferred over TARA and TO with stative predicates such as adjectives, as in (53a), and in the acceptability of stative predicates in contexts where BA is rejected with non-stative (change) predicates, such as in request or command constructions, illustrated in (53b): (53) a. IsogasikereBA/oisogasikatTARA/*oisogasii TO hi o aratamete ukagattemo ii desu ga.20 “If you are busy, I can come some other day.” b. Taroo ni *aeBA/au kikai ga areBA yorosiku tutaete kudasai. “If you meet/have a chance to meet Taroo, please greet him for me.” There exists, however, one particularly revealing class of exceptions to the inability of BA conditionals to occur in contexts of actual occurrence, and that is where an event occurs iteratively, as in (54):

106

WESLEY M. JACOBSEN (54) a. Daigaku no koro mainiti zyugyoo ga owareBA kissaten de tomodati to matiawase o site maazyan o yatte ita mono da. “When I was in college every day when classes were over I would meet my friends at a local coffee shop and play mahjong.” b. Kore o tabereBA sore, sore o tabereBA are, to iu yoo ni tugi kara tugi e to kuti ni hoobatteitta. “When he ate this, he would eat that; when he ate that, then another thing, and so on, stuffing one thing into his mouth after another.”

The possibility of a hypothetical BA form occurring in contexts of actual occurrence just in case an event is iterated receives a natural explanation when considered in light of the general correlation we have noted between hypothetical meaning and non-unique occurrence. Neither stative situations nor iterated events are “bound” to a unique point in time as are events which culminate in a one-time change. In both cases, the disassociation of an event from any unique point in time is accompanied by a shift away from the domain of real time altogether in the direction of irrealis, hypothetical meaning. The correlation between iterative and hypothetical meaning is in fact observable in domains of Japanese grammar beyond the four conditional forms we have considered so far. Consider, for example, the use of the concessive conditional form TEMO in contexts of future occurrence such as (55): (55) (Tatoe) ame ga hutTEMO siai wa yoteidoori ni okonawareru “Even if it rains, the game will be played as scheduled.” (55) can be understood to refer to the possible (hypothetical) occurrence of a given event at a particular point in time in the future. TEMO can also be used in past contexts where the associated event is understood to have actually occurred, but in such cases it is normally associated with the occurrence of a series of events which exhibit some quality on an increasing scale of intensity. This is most clearly seen in the uses of TEMO with quantifiers such as donna ni “to (what) extent” and ikura “(how) much”, as in the following:

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 107 (56) Ikura/ donna ni takai mono o katte ageTEMO yorokonde kurenakatta. “No matter how much/how expensive a thing I bought her, she wasn’t happy.” TEMO is not prohibited outright from contexts of unique (one-time) occurrence in the past, but in such cases the event in question is seen to occupy a position on the extreme end of a scale of events exhibiting some relevant property. In (57), for example, the event of buying something for the intended recipient is seen as occurring at the high end of a scale of events which exhibit to an increasing degree the (intended) property of being able to bring happiness to someone: (57) Sekkaku katte ageTEMO yorokonde kurenakatta.21 “Even though I went to the trouble of buying it for her, she (still) wasn’t happy.” A similar affinity between hypothetical meaning and non-unique or iterated occurrence can be seen to operate in the case of the form TEWA. This form functions in its basic use as a conditional-like connector attached to clauses expressing a hypothetical situation upon which some future negative or undesirable state of affairs is contingent. (58) Ima no koto o sensei ni hanasareTEWA komaru/ikenai/taihen na koto ni naru. “If you tell the teacher what just happened, I’ll be upset/it won’t do/it will be terrible.” Where the connective TEWA is used in actually occurring contexts, we find, parallel to what we observed for BA and TEMO, that an iterative meaning is again present: the event expressed in the first clause is seen to follow that in the second clause in a repeated manner: (59) Paatii ni kanarazu kiTEWA taberareru dake tabete kaette itta. “He would always come to parties, eat as much as he could and return home.” A wide-ranging correlation can thus be seen in Japanese morphological conditional forms between hypothetical meaning and the element of “nonunique” occurrence. The tendency for events or situations which are predi-

108

WESLEY M. JACOBSEN

cated of multiple points in time to be viewed as less real in conditional contexts than events whose occurrence is anchored at a single point or interval of time can be seen as a manifestation on the aspectual level of the similar tendency for hypothetical meaning to arise in certain tense contexts where branching histories are conceptualized, giving rise to multiple states of affairs which may alternatively characterize a given moment in time. That this correlation is more than a local phenomenon limited to Japanese can be seen in the use of English would as both a marker of hypothetical and iterative meaning, seen in (60a) and (60b) respectively: (60) a. John would leave this company without hesitation if he were given an offer elsewhere b. John would come over every evening after work and shoot the breeze with me until late into the night. It is probably safe to conjecture that this correlation between multiple occurrence and hypothetical meaning points to some cognitively fundamental link in the way time and reality are perceived across languages. 5.3 Hypothetical Meaning and Phenomenal vs. Meta-propositional Contingency A third dimension of meaning which correlates with the presence of hypothetical meaning in Japanese conditionals is the distinction between phenomenal and meta-propositional contingency described in Section 2. To the degree that it is some meta-propositional act or event in the antecedent clause, rather than the phenomenal content of the antecedent, that enters into a contingency relationship with the consequent clause, the conditional construction is assigned a hypothetical interpretation — one where the truth value of the antecedent proposition is presented as being indeterminate. Conversely, conditionals which are assigned a low degree of hypothetical meaning correlate consistently with a phenomenal interpretation, one where the antecedent and consequent events are seen as entering into a temporal order of literal precedence and antecedence. The clearest evidence of this can be seen in past contexts of actual (nonhypothetical) occurrence such as (41b), repeated here as (61), where conditional forms which are either meta-propositional in meaning or tending in that direction, namely NARA, NO DATTARA, TO SUREBA, and BA, are re-

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 109 jected. Only the conditional forms TARA and TO, which are open to a phenomenal interpretation, are allowed here: (61) Kusuri o *nomu NARA/*nomu NO DATTARA/*nomu TO SUREBA/ *nomeBA/nomu TO/nonDARA zutuu ga osamatta. (=41b) “When I took some medicine, my headache subsided.” In future contexts, likewise, a higher degree of hypothetical meaning attaches to BA conditionals, which tend toward a meta-propositional interpretation, than TARA conditionals, which are characteristically used to express phenomenal contingency: (62) Boonasu ga dereBA/deTARA atarasii kuruma o kau. (=43) “If/when I get my bonus, I’ll buy a new car.” As compared to NARA, however, the meaning of BA occupies a more intermediate position on the scale of phenomenal to meta-propositional, exhibiting a tendency in certain cases toward behavior more characteristic of phenomenal-type conditionals. Recall that in examples such as (23), repeated here as (63), BA conditionals are open to two interpretations, one where the order of the events in question is iconic to the order of the grammatical clauses (like the phenomenal interpretation given to TARA), and another where they may be understood to occur in the opposite order (an interpretation associated with meta-propositional conditionals): (63) Okane o daseBA himitu wa osiete yaru. (=23) “If you pay me, I’ll tell you the secret.” To the degree that phenomenal-like behavior is present in BA, though, it is restricted to contexts such as future or past iterative,22 where hypothetical meaning is already present for independent reasons having to do with tense and aspect. The meaning of BA does not range sufficiently toward the phenomenal end of the scale to allow its appearance in contexts of past actual occurrence such as (61). The hypothetical character of meta-propositional type conditionals is clearly related to the semantically subordinate character of the proposition appearing in the antecedent clause, as illustrated by the paraphrase relationship existing between (64a) and (64b) for a typical meta-propositional conditional of the NARA type:

110

WESLEY M. JACOBSEN (64) a. Kimi ga iku NARA, boku mo iku. “If you go, I’ll go too.” b. Kimi ga iku to iu NARA, boku mo iku koto ni suru. “If you say you will go, I decide (hereby) that I will go too.”

What is at issue in such conditionals is not the direct expression of the content of the antecedent proposition, but rather the process of presenting or accessing that proposition as information. The effect is to create some distance between the speaker and the proposition in question, with the further consequence that the content of that proposition becomes less “real” for the speaker. Prototypically real events, conversely, are those which are apprehended directly by the speaker in the phenomenal world, and expressed outrightly in the antecedent clause, unmediated by any semantically higher predicate, as in (65): (65) Soto ni deTARA ame ga hutteita. “When I went outside, it was raining.” The distancing function inherent in the meta-propositonal interpretation of conditionals makes it possible to encode information which may count as “real” by more objective standards in forms normally reserved for hypothetical or iterative meaning. The events represented in the antecedent clauses of the following constructions, for example, are ones which are actual and occur at a unique point in time, thus tending toward a “real” interpretation in terms of the tense and aspect dimensions discussed in Sections 5. 1 and 5. 2: (66) a. Soo ieBA watasi mo sorerasii hito ga tosyokan de urouro siteiru no o saikin mikaketa. “Now that (“if”) you say that, I too have seen someone like that hanging around the library recently.” b. Koo natTEWA/nareBA kyuuen butai o dooin saseru yori hoka wa nai. “Now that (“if”) things have come to this, there is no alternative to mobilizing a relief force.” c. Kimi ga soo itTEMO boku wa zettai ni iku. “Even though you say that, I’m definitely going.” The actual nature of the events in the antecedent clauses here is strengthened by the presence of deictics such as soo “that way” and koo “this way.” The fact that these events are nevertheless “cast” in a hypothetical form using BA, TEWA and TEMO illustrates a tendency to assign a non-definite truth status

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 111 to states of affairs which have not yet been incorporated as true into the speaker’s domain of knowledge. This use of hypothetical-like forms is common, for example, when referring to information which is attributed to the hearer, as in (66a) and (66c), or in expressing a state of affairs which has just entered the domain of awareness of the speaker and which the speaker has not yet fully processed, as in (66b). The hypothetical expression given to the information here thus illustrates differences in the way information is treated according to whose “territory” the information is seen to belong to, along the lines discussed in Kamio (1979, 1990), or according to whether the information represents shared or unshared beliefs between the speaker and hearer, along the lines discussed in Akatsuka (1985, 1986). To the groundwork laid by Akatsuka and Kamio, we add the observation that the distancing of information from the speaker’s territory or domain of belief in conditional sentences is a consequence of the semantic distancing inherent in and unique to the meta-propositional interpretation of conditional sentences. 5.4 Hypothetical Meaning and Old/New Information The interaction observed in the last section between hypothetical meaning and the informational value of the antecedent proposition to the discourse participants is manifested in yet another dimension of discourse-related meaning, that referred to traditionally as the continuum of old and new information. Recall that conditional sentences where the level of hypothetical meaning is lowest — those expressing actual past occurrence of two sequential events — are subject to a constraint that the consequent clause cannot represent an event or situation which is subject to the control of the speaker. This constraint is met automatically when the subject of the consequent clause is a third person, as in (67a), although predicates with a first person subject may appear in such clauses just in case they express an action or event not controlled by the speaker, as in (67b): (67) a. Rooka de suretigatta hito ga, yoku miru TO/miTARA, Kenzikun datta. “When I looked closely at the person I passed in the hall, it was Kenji.” b. Daremo inai uti ni kaeru TO/kaetTARA nantonaku sabisiku natta. “When I returned to the empty house, I somehow became lonely.”

112

WESLEY M. JACOBSEN

Viewed from a functional perspective, these types of conditionals can be seen to present an answer to the question: given the occurrence of an event A, what event B follows? Relative to each other, at least, the consequent clause represents focal, new information and the antecedent clause old information. Such an informational flow is not limited to past contexts, but is possible in future contexts as well, as illustrated in the reading given to the following examples where the predicate receives neutral, unstressed intonation: (68) a. Kono sigoto ga owatTARA oyogi ni iku. “When this work is over I’m going swimming.” b. Kodomo ga ookiku natTARA/naru TO izen no yoo ni mata soto de hatarakeru yoo ni naru. “When the children grow up, I’ll be able to work outside the home again as before.” Again, the examples in (68) may, under this reading, be seen to answer the question: given the occurrence of event A, what event B follows? Not all conditionals, however, fit the above pattern of informational flow. Some instead provide an answer to the question: given a certain event B under discussion, what prior event A is sufficient to bring about B? This type of functional pattern can be seen in TARA and TO conditionals, but is most typically associated with the BA form, as illustrated in (69). As noted by Koide, Komatsu, and Saita (1981), the BA form often carries with it the implication of a desired or positive development. This is a direct consequence of its function, illustrated here, of providing information as to the means by which an event referred to or evoked by the previous context can be brought about. Such information is typically sought just in those cases where the event in question is viewed as desirable in some relevant sense: (69) a. Koko kara mireBA huzisan ga yoku mieru. “If you look from here, you can see Mt. Fuji clearly.” b. Issyookenmei rensyuu sureBA kitto zyoozu ni nareru yo. “If you practice hard, you will certainly become good (at it).” The informational flow in (69a) and (69b) is clearly the opposite of that seen in the conditionals in the previous paragraph. Here it is the antecedent which bears the new information, with the consequent expressing what is relatively old information. This is confirmed by the possibility of questioned information appearing in the antecedent of conditionals of this type. The sentences in

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 113 (69) can, for example, be seen as providing answers to the questioned antecedent clauses in (70): (70) a. Doko kara mireBA huzisan ga yoku mieru? “From where (lit., if you look from where) can Mt. Fuji be seen clearly?” b. Doo sureBA zyoozu ni nareru?. “How (lit, if I do what) can I become good (at it)?.” The possibility of questioned information in the antecedent of a conditional sentence provides convincing evidence that conditional clauses should not be categorically identified with topics in the manner proposed in Haiman (1978, 1986).23 It is no accident that the conditional form BA, which we saw earlier to fall consistently at the high end of the scale of hypothetical meaning, should also be the form which correlates with a high degree of “newness” of information in the clause it marks. There is a natural correlation between the non-actual, and therefore as yet unknown, occurrence of an event, and the relative “newness” of the information it represents. By the same token, an event whose occurrence is already in the realm of fact lends itself easily to being treated as representing information which is presupposed — i.e., old. These are, of course, general tendencies which may be obscured by other discourse factors: it is conceivable, for example, for an event which forms part of the objective realm of fact not to be known as a fact to one of the discourse participants, or for an event which has not yet become actual to be treated as presupposed information for the purposes of a particular discourse setting. Yet the tendency is clearly present and observable in examples (67) and (68), where the absence of or low degree of hypothetical meaning in the antecedent clause correlates with an old-new informational flow, and in (69) and (70), where an antecedent marked with a highly hypothetical form correlates with a new-old informational flow. Note also how focal stress, normally understood as an indicator of new informational status, correlates with a higher degree of hypothetical interpretation and lack of such stress with a lower hypothetical reading in future TARA conditionals such as (43), repeated here: (71) Boonasu ga deTARA atarasii kuruma o kau. (=43) “If I get my bonus, I’ll buy a new car.” (possible interpretation with focal stress on deTARA)

114

WESLEY M. JACOBSEN “When I get my bonus, I’ll buy a new car.” (normal intonation)

Such interplay between hypothetical meaning and informational value is manifested in a particularly intriguing fashion in certain instances of pasttense TARA conditionals which appear to violate the constraint prohibiting same-subject control in contexts of actual occurrence.24 Recall that while neither TO nor TARA in such contexts allow consequent events under the intentional control of the speaker, TARA is more restrictive in requiring that the antecedent clause have a first-person subject, or otherwise involve the speaker in some direct fashion. Since the consequent event cannot be one controlled by the speaker, the result is that TARA in past-tense contexts exhibits the behavior of a switch-reference marker requiring different subjects in the antecedent and consequent clauses — first person in the former and non-first person in the latter.25 This accounts for the difference in acceptability in the judgment of most speakers between TARA and TO in the following example: (72) a. Tegami o yomu TO/?yonDARA uresisoo ni tobiagatta. “When she read the letter, she jumped up for joy.” b. Otoko wa heya ni hairu TO/?haitTARA uwagi o nuida.26 “When the man entered the room, he removed his jacket.” Interestingly, however, even speakers who find the TARA versions of (72a) and (72b) unacceptable in isolation sense a great improvement when a context providing a contrast to the antecedent clause is set up: (73) a. Sore made wa saenai kao o siteita noni, tegami o yon DARA uresisoo ni tobiagatta. “Although she had looked sullen up to then, when she read the letter, she jumped for joy.” b. Kankanderi no soto de wa nuganakatta noni, heya ni haitTARA uwagi o nuida. “Although he didn’t remove it when he was outside where the sun was hot, when he entered the room, he removed his jacket.” Note that the NONI clauses which preface the conditional constructions in (73a) and (73b) either mention or evoke information of the sort which occurs in the consequent clause of the following conditional construction — i.e., the mention of the facial expression of the individual in (73a) and the mention of

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 115 uwagi o nugu “remove his jacket” in (73b). The effect is to cast the consequent clause as old information, against which the antecedent clause provides some new information which contrasts with the element marked by the particle WA in the NONI clause. This means, essentially, that the flow of information is reversed in these cases from that observed in past-tense TARA conditionals, where it is the antecedent clause that presents old information. Significantly, the conditional predicate forms yonDARA and haitTARA in (73a) and (73b) tend to receive focal stress, an attribute which we saw correlates with the possibility of a hypothetical interpretation in examples such as (71). What we find, then, is that when the antecedent clause is cast as new information, constructions in TARA which are otherwise non-hypothetical in character take on a property which is associated with future TARA conditionals expressing non-actual, hypothetical occurrence — namely, that it is possible for the consequent clause to take the same subject as, and be controlled by the same individual as, the subject of the antecedent clause, as in (74): (74) Siken ni gookaku siTARA, kookyuu resutoran de minna ogotte yaru. “If I pass the exam, I’ll treat everyone at a first-class restaurant.” This provides further evidence of a link between hypothetical meaning and “new” informational status, a link which is in this case mediated by contrastive meaning.

6.

Conclusion

I have proposed in this paper that the connection intuitively felt to exist between antecedent and consequent propositions in conditional constructions is at its root no more and no less than a relationship of co-occurrence in time wherein one event is seen to consistently follow, and therefore depend on, the occurrence of another. While this relationship may take on the character of cause and effect, it encompasses a greater range of possible co-occurrence patterns than those of cause and effect, one which cannot be defined a priori but is open to the entire range of co-occurrence patterns which may be encountered in human experience. The experience of a pattern of repeated cooccurrence between two event types in the past forms the basis for predicting that an occurrence of one event will be followed by the other even in future,

116

WESLEY M. JACOBSEN

yet-to-be-experienced situations. This pattern of repeated co-occurrence necessarily originates in an initial experience in which the co-occurrence relationship is first discovered, pointing to a cognitive link between hypothetical meaning and actual experience which is directly reflected in the past-tense use of conditional forms in Japanese. At its basic level, therefore, conditional meaning is rooted in a temporal relationship between events as phenomena, a relationship we have called one of “contingency.” This fundamentally temporal relationship extends beyond the phenomenal level, however, to include superordinate acts or events which can be represented as higher predicates to which the content of the constituent propositions of a conditional are subordinated. Such “meta-propositional” contingency relationships include not only conditionals which have traditionally been referred to as epistemic or speech-act types, but also cases where the consequent clause represents the content of a decision or a deduction drawn from a premise. The distinction between phenomenal and meta-propositional meaning represents one dimension along which formal differences can be observed among Japanese conditionals. Given that the knowledge of a contingency relationship between two propositions is most typically employed in predicting the outcome of events in the future, and that future events are by definition “hypothetical,” there is a close connection of a pragmatic nature between hypothetical and contingent meaning. Yet the connection is not a necessary one, and the view of conditionals I have proposed here allows hypothetical meaning to be viewed as an element of meaning fundamentally independent from, if frequently converging with, conditional meaning. As I have attempted to show, hypothetical meaning as it manifests itself in Japanese conditionals is not a concept which can be exclusively accounted for either by contextual factors such as whether or not the speaker presupposes the truth of the content of the conditional clause, on the one hand, or by any “objective” dichotomy of actual vs. nonactual occurrence in the real world, on the other. It is instead a category of meaning which interacts with subjective and temporal domains of human experience in numerous and still poorly understood ways which exceed the capacity of any reductionist theories to explain, whether truth-functional or discourse-based. We have looked specifically at four such domains of meaning in this paper: the tense character of a clause as it enables or precludes the conceptualization of branching histories; the aspectual character of the predicate as seen on a continuum of unique versus multiple occurrence; the

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 117 continuum between phenomenal and meta-propositional contingency; and the relative character of the antecedent and consequent clauses on a continuum of new to old information. We cannot pretend to have unearthed all the domains of meaning relevant to the notion of hypotheticality, much less to have accounted for their interaction. This is a task which must be left to further research, but the groundwork laid here will hopefully contribute to further work on that question.27

Notes 1.

Conditional inflections are rendered in capital letters in the Japanese examples in this paper simply for the purpose of ease of identification, and not to indicate intonational features such as stress.

2.

Areas of overlap between conditional and temporal meaning are discussed in Reilly (1986) for English and Harris (1986) for the Romance languages.

3.

Some studies of Japanese conditionals in fact reject examples such as (2) and (3) outright from membership in that class. This is the position taken, for instance, in Akatsuka (1985, 1986).

4.

Since a logical equivalence exists between (NOT(P→Q)) and (P AND NOT Q), the truthfunctional approach would incorrectly predict that a denial of the conditional in (6) is equivalent to affirming that an earthquake of magnitude 6 will occur and the building will not collapse.

5.

See Dowty (1979) for a linguistic perspective on some of the difficulties with formulating a workable definition of causation.

6.

These conditional types participate, for example, in Sweetser’s (1990) classification of conditionals into three domains of meaning — the “content domain,” the “epistemic domain,” and the “speech-act domain.”

7.

Though various idiosyncrasies of each of these conditional types will be touched on in the course of our discussion, a comprehensive treatment of them is beyond the scope of this paper. For a review (in Japanese) of the considerable literature devoted to Japanese conditionals, see Arita (1993b).

8.

In a spirit similar to that of our proposed distinction between phenomenal and metapropositional contingency, Masuoka (1993) distinguishes the meaning of TARA, BA, and NARA as that of a conditional relationship occurring respectively at the level of phenomena, propositions, and judgments (handan). Masuoka cites the tripartite classification of conditionals made in Sweetser (1990), referred to in Footnote 6, as a motivation for his own analysis.

9.

TARA and BA are inflected forms which carry no tense at all. TO requires that its preceding predicate occur in the non-past tense form, thus in effect neutralizing any tense opposition.

118 10.

WESLEY M. JACOBSEN This is related to another fact about the morphology of Japanese conditional forms: as shown in (i) - (iii), the form NARA in fact functions as a suppletive form of the copula DA occurring within the inflectional paradigm of the BA conditional (note that there are three distinct predicate categories in Japanese — verb, adjective, and noun + copula). (i) iku “go” → ikeBA “if one goes” (ii) yasui “is cheap” → yasukereBA “if it’s cheap” (iii) hon da “is a book” → hon NARA “if it’s a book” The copular flavor of the NARA form carries over into uses of NARA to mark verbal and adjectival clauses as well, so that the preceding clause is felt to be an (underlyingly) nominalized form marked by the complementizer NO, as in onaka ga suiteiru (NO) NARA “if it’s the case that you’re hungry.” The NO NARA form is in turn clearly related to the sentence-final complementizer + copula form NO DA commonly used in providing explanations and clarifications (e.g., Onaka ga suiteiru NO DA “It’s that I’m hungry.”). For a historical perspective on the semantics of NARA, see Quinn (forthcoming).

11.

This form consists of the sentential complementizer NO followed by the TARA conditional form of the copula DA. As with NARA, the copular nature of the DATTARA form requires that the preceding constituent be nominal in category, either a garden variety noun (e.g., Boku DATTARA “if it were me”) or a clause nominalized by NO (e.g., Onaka ga suiteiru NO DATTARA “if it’s the case that you’re hungry”).

12.

This form consists of the sentential complementizer TO (normally used to mark the complements of verbs of saying or thinking, not to be confused with the conditional form TO), followed by the BA conditional form of the verb SURU, meaning “do” or “make.” The combination TO SURU takes on the meaning “assume” and is used to present a hypothesis or proposition which is assumed for the sake of argument (e.g., Onaka ga suiteiru TO SURU “Say (for the sake of argument) that you were hungry”). In its conditional use, any of the three forms TO SUREBA, TO SITARA, and TO SURU TO are possible. Although some minor differences can be observed in the usage of these three forms which can be attributed to independent constraints on the usage of BA, TARA, and TO conditionals, they are identical in their meta-propositional function of referring to a proposition through the act of assuming it, as opposed to presenting the content of the proposition as a phenomenon.

13.

Hasunuma (1985) characterizes NARA and TO SUREBA as having the function of presenting a state of affairs as a condition on a speech act, as opposed to expressing a relationship of sequentiality or causality between states of affairs, as do TO, TARA, and BA. Her characterization, which applies equally to NO DATTARA, can be seen as a special case of the meta-propositional reading which these conditional forms impose on their constituent propositions.

14.

See Jacobsen (1990b, 1992).

15.

This constraint was not operative in earlier stages of the language, as seen in the following example from Matsushita (1930), quoted in Arita (1993), where BA is used in a context of actual past occurrence. (I)

Ame da to iu kara kasa o motte dekakereBA nihonbare da. “Since they said it was going to rain, I went out with my umbrella. When I did so, (I found that) the sky was clear.”

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 119 Such uses of BA are heard occasionally in poetic contexts such as song lyrics, but have a distinctly classical flavor. BA traces its ancestry farther back in time than the other three modern conditional forms and appears to have originally covered the semantic territory now divided among the four forms. 16.

Such an analysis of branching histories is adopted, for example, in Dowty (1979) in his discussion of progressive meaning.

17.

The sentence-final form NO NI combines counterfactual meaning with a sense of disappointment that the actual outcome of events was less felicitous than the positive state of affairs expressed in the consequent clause of the conditional.

18.

As I argue in Jacobsen (1990), stativity is a scalar phenomenon which can be seen to be present in greater or lesser degrees to the extent a situation is seen as inherently changeable or not. Agentive situations tend to be seen as inherently changeable, since they are subject to the arbitrary will of a human agent who may decide at any time to put an end to the situation. The presence or lack of agentive meaning accounts for the differing aspectual behavior of the existential verbs iru and aru, the former animate and agentive and the latter inanimate and non-agentive. Iru is acceptable in past nonhypothetical conditional constructions such as (i), even though aru, as seen in (51b), is not. (i)

Soko in iTARA, sensei ni koe o kakerareta. “As I was (standing) there, I was accosted by my teacher.”

19.

See Jacobsen (1990a).

20.

TARA is not unacceptable in (53a). The claim here is simply that there is a statistically higher tendency for BA to occur in these contexts than any other conditional form, including TARA.

21.

I note in passing that some speakers exhibit resistance to the use of TEMO in contexts of unique occurrence such as (57), preferring instead the use of the NONI, which lacks the scalar implications of TEMO . (i)

Sekkaku katte ageta NONI yorokonde kurenakatta. (Gloss same as for (57))

22.

For examples of BA constructions receiving a phenomenal-type interpretation in past iterative contexts, see the earlier (54a) and (54b).

23.

See Jacobsen (1992) for a detailed discussion of the difficulties posed to Haiman’s analysis by Japanese conditionals.

24.

The first to point out this phenomenon, to my knowledge, was Koide, Komatsu, and Saita (1981).

25.

The function of TARA to mark switch reference is pointed out and discussed in Iwasaki (1993).

26.

Examples (72b) and (73b) are due to Koide et al. (1981).

27.

I wish to express my thanks to Yuko Iguchi, Yuki Johnson, and Yoshio Satoh for the assistance they provided at various stages in the preparation of this paper in gathering data and judging the appropriateness of the various conditional forms appearing in the examples in this paper.

120

WESLEY M. JACOBSEN

References Akatsuka, Noriko 1985 “Conditionals and epistemic scale”. Language 61: 625-39. Akatsuka, Noriko 1986 “Conditionals are discourse-bound”. In E. Traugott, A. Ter Meulen, J. Reilly, and C. Ferguson (eds), 333-52. Alfonso, Anthony 1974 Japanese language patterns. Tokyo: Sophia University L. L. Center of Applied Linguistics. Arita, Setsuko 1993a “Nihongo no zyookenbun to tisiki (Japanese conditionals and knowledge)”. In T. Masuoka (ed), 41-71. Arita, Setsuko 1993b “Nihongo zyookenbun kenkyuu no hensen (Trends in the research of Japanese conditional sentences)”. In T. Masuoka (ed), 225-278. Bedell, George, Eichi Kobayashi, and Masatake Muraki. (ed) 1979 Explorations in linguistics: Papers in honor of Kazuko Inoue. Tokyo: Kenkyuusha. Brentari, Diane, Gary Larson, and Lynn MacLeod (ed) 1992 The joy of grammar: A festschrift in honor of James D. McCawley. Amsterdam: John Benjamins. Cole, Peter and Jerry L. Morgan (eds) 1975 Syntax and semantics 3: Speech acts. New York: Academic Press. Comrie, Bernard 1986 “Conditionals: A typology”. In E. Traugott, A. Ter Meulen, J. Reilly, and C. Ferguson (eds), 77-102. Dowty, David 1979 Word meaning and Montague grammar. Dordrecht, The Netherlands: D. Reidel. Grice, H. Paul 1975 “Logic and conversation”. In P. Cole and J. L. Morgan (eds), 45-58. Haiman, John 1978 “Conditionals are topics”. Language 54: 564-89. Haiman, John 1986 “Constraints on the form and meaning of the protasis”. In E. Traugott, A. Ter Meulen, J. Reilly, and C. Ferguson (eds), 215-28. Harris, Martin B. 1986 “The historical development of SI-clauses in romance”. In E. Traugott, A. Ter Meulen, J. Reilly, and C. Ferguson (eds), 265-284. Hasunuma, Akiko 1993 “‘Tara’ to ‘to’ no zizituteki yoohoo o megutte (Concerning the factual use of TARA and TO)”. In T. Masuoka (ed), 73-97. Hasunuma, Akiko 1985 “‘Nara’ to ‘to sureba’ (NARA and TO SUREBA)”. Nihongo Kyooiku 56.

ASPECTS OF HYPOTHETICAL MEANING IN JAPANESE CONDITIONALS 121 Iwasaki, Shoichi 1993 Subjectivity in grammar and discourse — theoretical considerations and a case study of Japanese spoken discourse. Amsterdam: John Benjamins. Jacobsen, Wesley M. 1990a “The multifaceted character of stativity”. In O. Kamada and W. Jacobsen (eds), 76-99. Jacobsen, Wesley M. 1990b “Zyookenbun ni okeru kanrensei ni tuite (On the notion of connection in conditional sentences)”. Nihongogaku 9: 93-108. Jacobsen, Wesley M. 1992 “Are conditionals topics? The Japanese case”. In D. Brentari, G. Larson, and L. MacLeod (eds), 131-160. Kamada, Osamu and Wesley M. Jacobsen (eds) 1990 On Japanese and how to teach it: In honor of Seiichi Makino. Tokyo: Japan Times. Kamio, Akio 1979 “On the notion of speaker’s territory of information: A functional analysis of certain sentence-final forms in Japanese.” In G. Bedell, E. Kobayashi, and M. Muraki (eds), 213-31. Kamio, Akio 1990 Zyoohoo no nawabari riron: Gengo no kinooteki bunseki (The theory of territory of information: A functional analysis of language). Tokyo: Taishukan. Koide, Keiichi, Noriko Komatsu, and Izumi Saita 1981 “To, ba, tara: Danwa ni okeru sentaku yooin o motomete (To, ba, tara: Factors affecting their choice in discourse)”. Stanford Inter-university Center for Japanese Language Studies (Tokyo, Japan) Kiyoo. 4: 30-66. Kuno, Susumu 1973 The Structure of the Japanese language. Cambridge, MA: MIT Press. Masuoka, Takashi 1993 “Zyooken hyoogen to bun no gainen reberu (Conditional expressions and the conceptual level of sentences)”. In T. Masuoka (ed), 23-39. Masuoka, Takashi (ed) 1993 Nihongo no zyooken hyoogen (Conditional expressions in Japanese). Tokyo: Kurosio Publishers. Matsushita, Daisaburoo 1930 Hyoozyun nihon koogohoo (Standard colloquial Japanese grammar). Tokyo: Chuubunkan Shoten/Benseisha. McCawley, James D. 1981 Everything that linguists have always wanted to know about logic* (*but were ashamed to ask). Chicago: The University of Chicago Press. Nazikian, Fumiko Asari 1994 A functional analysis of Japanese conditionals TARA, BA, NARA, and TO. Dissertation, University of Sydney.

122

WESLEY M. JACOBSEN

Quinn, Charles Forthcoming A reader’s grammar of classical Japanese. Ann Arbor, Michigan: University of Michigan Center for Japanese Studies. Reilly, Judy S. 1986 “The acquisition of temporals and conditionals”. In E. Traugott, A. Ter Meulen, J. Reilly, and C. Ferguson (eds), 309-332. Sweetser, Eve 1990 From etymology to pragmatics. Cambridge: Cambridge University Press. Takubo, Yukinori 1993 “Danwa kanri riron kara mita nihongo no hanzizituteki zyookenbun (Japanese counterfactual conditionals viewed from the perspective of discourse management theory)”. In T. Masuoka (ed), 169-183. Traugott, Elizabeth, Alice Ter Meulen, Judy Reilly, and Charles Ferguson (eds) 1986 On conditionals. Cambridge: Cambridge University Press. Vendler, Zeno 1957 “Verbs and times”. Reprinted in Z. Vendler (ed), 97-121. Vendler, Zeno (ed) 1967 Linguistics in philosophy. Ithaca: Cornell University Press.

ENGLISH PASSIVES WITH INDEFINITE SUBJECTS

123

A Context-Based Account of English Passives with Indefinite Subjects Aiko Utsugi Keio University

1.

Introduction

The aim of this paper is to explore the contexts in which passives with indefinite subjects are used. The following examples from newspaper articles have subjects that are indefinite in form: (1) (2) (3)

A comment should be made by the president. A funeral will be arranged by his son. Another example was provided by Prof. Lee.

Kuno and Kaburaki (1977) discuss the likelihood of definite NPs occurring in the subject position from the standpoint of old/new information. They note that one of the functions of passivization is to reorder the underlying subject and the underlying object so that the constituents that represent old information appear clause-initially and those that represent new information clausefinally. As for the passive, they note: (4)

...this explains why many passive sentences have definite derived subjects and indefinite by-passive agentives, as: (a) Mary was attacked by a mugger in Central Park. (b) ?? A girl was attacked by the mugger. (Kuno and Kaburaki, 1977:648)

* I am grateful to Dr. Moore of Cambridge University for many helpful comments on an earlier version of this paper. I am also grateful to Wesley M. Jacobsen for many detailed stylistic comments on the pre-final version.

124

AIKO UTSUGI

Out of context, therefore, it might appear that active sentences such as (5)-(7) in which definite NPs are placed in the subject position should be preferred to the corresponding passives in (1)-(3): (5) (6) (7)

The president should make a comment. His son will arrange a funeral. Prof. Lee provided another example.

A context-based consideration, however, calls this view into question. We will come back to examples (1)-(3) later.

2.

Associative Chain Formation with the Context that Precedes

Let us first look at some further examples from actual data that have indefinite subjects and definite agents: (8)

One very simple example is provided by Sir Frederick Bartlett in his experiments on thinking. The subject is required to complete an arrangement of words, of which a fragment is given, by selection from a random group. (Formal Transcendental & Dialectical Thinking, E.E. Harris, 1987) (9) Similar views have been developed by P.F. Strawson and Paul Ziff. Can any such view be correct? (Reference and Contrast, C. Chastain, 1975) When we look at these examples, we soon realise that the subjects are implicitly related to what has been mentioned previously:1 (8)’ One very simple example One very simple example of what has been discussed in the preceding context (9)’ Similar views Views similar to the view already introduced in the preceding context In this sense, the subjects in these examples are not entirely new in spite of the fact that they are indefinite in form, and therefore these passives are consis-

ENGLISH PASSIVES WITH INDEFINITE SUBJECTS

125

tent with and result from the general old/new information structure which is proposed, for instance, by Kuno and Kaburaki (1977). Prince observed that an overwhelming majority of the subjects used in a fragment of a naturallyoccurring oral narrative she examined were what she called Evoked entities; that is, those subjects were, in her terms, already ‘on the counter’. On the other hand, few of the non-subjects were already Evoked entities. None of the subjects were Brand-new, while nearly a third of the nonsubjects were (1979:273). We may say that examples (8) and (9) are consistent with the general information structure observed by Prince because the subject is an indefinite NP that represents an entity connected to a preceding portion of the discourse that is, in Prince’s terms, already ‘on the counter’. In order to explain more clearly the occurrence of the type of passive under discussion, I would like to introduce the notion ‘associative chain’ hinted at in Hawkins’ (1978) use of the term ‘associative anaphora,’ which we will discuss in the next sub-section. I suggest that discourse is made up of associative chains; that is, discourse is organized in such a way that newly introduced elements are linked to already-mentioned elements in such a way that the two overlap to some extent either overtly or implicitly. (Later we will discuss ‘associative chains’ with the context that follows as well.) The following remark in Prince (1979) is insightful: (10) It seems that hearers follow a Conservation Principle, according to which they try to use the entities they already have rather than not. If they cannot, they prefer to make new entities out of old ones, and, if that is impossible, then they try to link the new entity to an old one. (Prince, 1979:275) Prince further claims that the function of a number of syntactic constructions is to implement what she calls the Conservation Principle: (11) ...speakers seem to have command of a variety of syntactic constructions that appear to be used at times simply to remove relatively unfamiliar entities from being represented by subject NPs: existential there, it-clefts, left dislocation out of subject position, certain relative clause constructions, and a variety of so-called ‘run-on’ sentences. (Prince, 1979:276) What I am going to propose here is that the passive construction should be

126

AIKO UTSUGI

included in Prince’s list of syntactic constructions, because it also performs the same task; that is, it serves to remove what she calls ‘unfamiliar’ entities from subject position. In the passive construction ‘unfamiliar’ entities can either occur in agentive by phrases or be eliminated, and entities that are ‘familiar’ because, for instance, they are related to previous mentions, have a tendency to be placed in subject position. We may thus assume that one of the functions of the English passive is to form an associative chain with the preceding context and thereby improve the flow of discourse by placing entities that are, in Prince’s terms, ‘familiar’ in subject position. This suggests that if the passive is used in a context where the subject participates successfully in an associative chain with the preceding context, that passive will sound natural even when the subject is indefinite in form. Going back to our earlier examples, repeated below, the same account can be given. The subjects are indefinite in form, but are all semantically context-bound. In other words, in order to construe this type of passive properly consideration of the preceding context is essential: (1) (2) (3)

A comment should be made by the president. A funeral will be arranged by his son. Another example was provided by Prof. Lee.

In the newspaper articles in which these passives are used, A comment is construed as ‘A comment on what has been already discussed’, A funeral is construed as ‘A funeral of the person who has already been mentioned’, and Another example normally presupposes at least one already mentioned example. These subjects thus participate successfully in an adequate associative chain with the preceding context. 2.1 Associative Chains: Views of Other Linguists NPs that are related to previous mentions have been referred to in different ways by other linguists. Hawkins (1978), for instance, uses the term ‘associative anaphora’. By this he means that the NPs are associated with entities already mentioned but are not strictly anaphoric with them because they do not directly refer back to the entities themselves. We must note that the term ‘anaphora’ is not used here in the sense of generative grammar where it generally refers to reflexives and reciprocals. One example of Hawkins’ ‘associative anaphora’ is as follows:

ENGLISH PASSIVES WITH INDEFINITE SUBJECTS

127

(12) The man drove past our house in a car. The exhaust fumes were terrible. (Hawkins, 1978:123) Hawkins’ ‘associative anaphora’ thus involves general knowledge of the world — in the case of (12), knowledge that a car gives off exhaust fumes. In this sense his ‘associative anaphora’ is similar to Prince’s (1981) concept of ‘inferrable’. Prince defines ‘inferrable’ as follows: (13) A discourse entity is Inferrable if the speaker assumes the hearer can infer it, via logical — or, more commonly, plausible — reasoning, from discourse entities already Evoked or from other Inferrables. (Prince, 1981:236) Her examples of ‘inferrable’ entities are as follows: (14) I got on a bus yesterday and the driver was drunk. (15) I went to the post office and the stupid clerk couldn’t find a stamp. (Prince, 1981:233, 237) According to Prince’s account, the driver is inferrable from a bus, and the stupid clerk is inferrable from the post office. Hawkins’ associative anaphoras and Prince’s inferrable entities thus involve general knowledge of the world, and therefore they are different from the passive subjects in our examples. As Prince notes clearly, her ‘inferrable’ entities are mostly culture-based (1981:242). What is common between these and the passive subjects in our examples is that they are all context-bound; that is, they require alreadymentioned entities to be construed properly.

3.

Associative Chain Formation Involving Indefinite Passive Subjects

Regarding the examples given by Prince (1981) and Hawkins (1978), I draw attention to the fact that they are all definite in form. Their discussions are restricted to properties of referents as discourse entities and they do not refer to specific linguistic forms, except in Hawkins’ following statement: (16) ...this (= associative anaphora) is undoubtedly the most frequent use that is made of the. (Hawkins, 1978:123)

128

AIKO UTSUGI

One of the characteristics of the passives we have been discussing is that they have indefinite subjects. Brown and Yule (1983:185) simply note that inferrable entities are regularly introduced by definite expressions. A distinction should be made, however, between definite NPs and indefinite NPs that are inferred from or related to previous mentions. Indefinite NPs do not carry a linguistic anaphoric marker such as the definite article the, so I posit that indefinite NPs have a stronger tendency to be placed in the subject position in order to make the link with previous mentions obvious in the absence of any overt anaphoric markers. Consider the following examples: (17) a. There was a bomb threat at our college. A student was questioned by the police. b. There was a bomb threat at our college. The police questioned a student. A student is a newly introduced entity which is indefinite in form. Nevertheless the passive sentence shown in (17a) is more adequately linked to the preceding sentence than the active sentence in (17b) where the police is placed in the subject position. When a student is put in the subject position, the student is more readily understood to be a suspect in the bomb threat mentioned in the preceding sentence. Because a student is inferrable from our college, a more satisfactory link is created between the two sentences, while the police is not inferrable in this way from any particular entity in the preceding sentence. A similar example is proposed by Corblin (1990). He remarks that indefinites can also occur in what he calls ‘associative referential chains,’ as in the following example: (18) The house is beautiful, but a window is broken. (Corblin 1990:87) He notes that there is no clear answer to the question of whether such a chain should be called anaphoric or referential. It is clear, however, that in this example a window is, under the most natural interpretation, understood to be ‘a window on the house that has been just mentioned’, and a window thus clearly forms an associative chain with the house. Corblin does not, however, discuss such chain formations in relation to the passive structure. Let us see if corresponding active sentences have the same effect:

ENGLISH PASSIVES WITH INDEFINITE SUBJECTS

129

(19) a. We moved into a beautiful house yesterday. A window had been broken by a neighbor. b. We moved into a beautiful house yesterday. A neighbor had broken a window. It may be possible to interpret (19b), for instance, as ‘A neighbor had broken a window on their own house’. This interpretation is less likely in (19a) because the associative chain between the house and a window is more direct in (19a) than in (19b). If the active is used in the sense shown in (19a), (20b) below is probably preferable to (20a), because the windows overtly indicates the link to the previously mentioned entity a beautiful house by means of the definite article. (20) a. We moved into a beautiful house. A neighbor had broken a window. b. We moved into a beautiful house. A neighbor had broken one of the windows (on our beautiful house). Examples (17)-(19) show that an entity is more readily linked to a previous mention when it occurs in the subject position than in the object position. This observation is consistent with and results from the general information structure mentioned earlier. When an entity that is inferrable from the preceding context occurs in the object position, therefore, it needs some kind of marker such as the definite article in the above example to indicate the link with a previous mention. I would like to add that highly relevant to our discussion of passives with indefinite subjects here is Ward and Prince’s (1991) analysis of the topicalization of indefinite NPs. They maintain that definiteness and topicalization are independent of each other. They give several examples in which indefinite NPs are topicalized, such as the following: (21) There are a couple of nice points in here. One point I can say something about. The other I’m not sure. (Ward and Prince, 1991:171) Although ‘topicalization’ and passivization are two distinct operations, the following argument, in which Ward and Prince attempt to account for topicalization in relational terms, seems to apply to our discussion of the passive subject:

130

AIKO UTSUGI (22) ...[we] argue that what is relevant is neither the morphological/ lexical features of the NP nor the specificity of the entity represented by that NP, but rather the particular relationship which must obtain between the entity in question and other entities in the discourse model. (Ward and Prince, 1991:168)

In the passive operation, too, even when the subject is indefinite, the sentence may sound natural if what is referred to by the subject has a particular relationship with a previous mention. Though indefinite in form, such subjects are pragmatically related to previous mentions.

4.

Associative Chain Formation with the Context that Follows

We have seen some examples in which NPs that are related to previous mentions are placed in subject position. We have also seen that this relationship is clear even when the subjects are indefinite. In this section we will look at some further examples of passives with indefinite subjects. We will take into consideration here a broader range of contexts, including those that follow, as well as precede, the subject in question. First consider the following examples from actual data: (paragraph initial) (23) The electoral calendar was blank between the final Presidential primaries in June and last week’s Democratic convention, but history was being made by Ross Perot and his volunteers. Mr. Perot’s candidacy and... (The Boston Globe, 8 July 1992) (paragraph initial) (24) A philosophical justification of the new morality has been offered by John Moore (1987). He spoke of... (Notes and Records of the Royal Society of London, 1992) (paragraph initial) (25) Another adverse consequence of increased paternal involvement was described by Russell (1982, 1983) and Kelly (1981) in studies of Australian families. In both studies, highly participant fathers had more conflicts with their children. (Reassessing fatherhood, C. Lewis, 1987)

ENGLISH PASSIVES WITH INDEFINITE SUBJECTS

131

(paragraph initial) (26) Hopes that the economy may be poised for recovery were hit yesterday by an unexpectedly sharp rise in unemployment last month... (The Independent, 18 Dec. 1992) Unlike our earlier examples, the subjects of the passive sentences in (24), (25) and (26) are overtly related to previous mentions, that is, the new morality, increased paternal involvement, and the economy respectively. However, as no issues of significance are raised by referring to these as ‘indefinite subjects’ rather than as ‘subjects whose head NPs are indefinite’, I would like to treat these examples under the category of ‘passives with indefinite subjects’. There is a pattern common to these passive sentences, namely that the agent phrase, not the passive subject, provides the discourse topic. Here I use the term ‘discourse topic’ as suggested by Dik: (27) A discourse, taken in the wide sense of any kind of coherent text (a story, a monologue, a dialogue, a lecture, etc.), is “about” certain entities. ...For those entities about which a certain discourse imparts information we may use the term Discourse Topic. (Dik, 1989:266) In our discussion the term ‘discourse topic’ is used in the sense of ‘the topic of the paragraph’ or ‘what the paragraph is about’. In (23)-(26), I did not present the entire paragraphs in which these examples occur nor their preceding paragraphs. I would like, therefore, to provide an explanation of the contexts in which these sentences occur. In (23), Ross Perot is first introduced in the agent phrase of a passive sentence that occurs in paragraph initial position, followed by a sentence about his candidacy. The whole paragraph is about Ross Perot. Likewise, in (24), the agent John Moore occurring in the first sentence of the paragraph serves as the subject of several following sentences. The whole paragraph is about John Moore. In (25), Russell and Kelly occur in agent phrases of the paragraph initial sentence, and their studies is the subject of the following sentence, thereafter becoming the topic of the whole paragraph. Likewise, in (26), an unexpectedly sharp rise in unemployment occurring in the agent phrase becomes the topic of the whole paragraph. To summarize, these examples share the following three properties:

132

AIKO UTSUGI

ii(i) the passive occurs in the paragraph initial sentence, i(ii) the subject (or the head NP of the subject) of the passive is indefinite in form but is overtly or implicitly related to a previous mention, (iii) the agent phrase introduces a new entity, which becomes the topic of the paragraph in which that passive occurs. The pattern seen in certain English passives is therefore one where the passive agent phrase introduces a discourse topic. This pattern is not restricted to passives with indefinite subjects, but when such passive occurs in paragraph initial position, it has a tendency to exhibit this pattern. Highly relevant to our discussion of the discourse-topic introductory role is the term New Topic proposed by Dik (1989). (D-Topic stands for a discourse topic): (28) If a discourse is to be about a certain D-Topic, that D-Topic will at some point have to be introduced for the first time. Such a first presentation of a D-Topic will be called a New Topic (NewTop)... (Dik, 1989:267) Dik also notes that a New Topic typically takes a non-initial position, and among the examples of various strategies he cites that conspire to bring the New Topic to such a position is the following passive construction: (29) ...The building is now occupied by the Bodleian Library. (Dik, 1989:269) Thus Dik explicitly shows that elements that occur in passive agent phrases can be New Topics. At the sentence level, The building is likely to be the topic of the sentence. However, when we take into consideration the whole discourse in which this particular sentence is used, it is possible for Bodlleian Library appearing in the agent phrase to represent the discourse topic. In order to avoid conflating the notion of sentence topic and that of discourse topic, I would like to refer to an argument made by Siewierska (1991). She criticizes Dik’s treatment of New Topics as follows (the numbering is my own): (30) ...A controversial aspect of Dik’s treatment of (30)’ is that the New Tops do not meet the ‘aboutness’ criterion for topicality at clause level. For example, (a) is not about Mary and her sisters. It is only in the subsequent clause, as in (b), that this expression functions as a clausal topic.

ENGLISH PASSIVES WITH INDEFINITE SUBJECTS

133

(30)’ a. In the garden sat Mary and her sisters. b. They were having an animated conversation. ...The New Top, in contrast to the other three topic functions is not a clausal topic. (Siewierska, 1991:161-163) In her criticism of Dik’s account, Siewierska overlooks the important point that Dik’s New Topic is a discourse notion. The difference in viewpoint between Dik and Siewierska stems from the level at which the topic role is determined. When Dik says Mary and her sisters is a new topic in the sentence In the garden sat Mary and her sisters, he takes into consideration the entire discourse, not only the particular sentence in which Mary and her sisters occurs for the first time. On the other hand, according to Siewierska, Mary and her sisters is not the topic of the sentence In the garden sat Mary and her sisters, but is the topic only of the following sentence in which Mary and her sisters is repeated (in the form of a pronoun). Dik’s category of New Topic encompasses the possibility of an element functioning as the topic of an entire passage. Thus I take the same position as Dik and consider the possibility that the element in a passive agent phrase may become the topic of the discourse in which the passive is used. Let us take a closer look at example (23), in which the subject history is not overtly linked to a previous mention. In order to understand how this subject is construed here, I must repeat an observation made in the preceding sections. Although indefinite in form and not overtly linked to what has already been talked about, the subject here is understood to be the history of the presidential election already introduced and not, for example, world history in general: (23) The electoral calendar was blank between the final Presidential primaries in June and last week’s Democratic convention, but history was being made by Ross Perot and his volunteers. Mr. Perot’s candidacy and... (The Boston Globe, 8 July 1992) The active version would be as follows: (23)’ The electoral calendar was blank between the final Presidential primaries in June and last week’s Democratic convention, but Ross Perot and his volunteers were making history. Mr. Perot’s candidacy and...

134

AIKO UTSUGI

What the author of this article has in mind as a reference for history would be less clear in (23)’ than in (23). When history occurs as an object without any marker indicating a link to a previous mention, it is less readily understood to be ‘the history of the presidential election’. Again, this results from the general information structure discussed earlier. This phenomenon can also be accounted for in terms of an associative chain. History in (23), is more readily interpreted as ‘electoral history’ than history in (23)’. Although history is indefinite in form, the sentence enters more readily into an associative chain when history is made the subject of a passive sentence than when the freshlyintroduced entity Ross Perot and his volunteers occurs as the subject of a corresponding active sentence. Furthermore, Ross Perot occurring in the passive agent phrase is the subject of the immediately following sentence and becomes the topic of the whole paragraph in which the passive appears as the first sentence. Example (23), together with (24)-(26) on page 10, exhibit a certain discourse pattern of ‘associative chain formation’ that is illustrated in the following diagram of (23). Similar diagrams could be drawn for examples (24), (25) and (26). This diagram shows that this particular type of passive serves as a means of linkage between two paragraphs. What is particularly important here is that if an active is used as shown in (b) in Figure 1, the natural flow of discourse is broken. It is also important to note that since the agent phrase in these passives functions to provide a discourse topic, unlike many other types of English passive, agentless passives are not possible in these cases. Presidential election (topic of the newspaper article)

> t2 Mr. Perot

t1 electoral calendar (linked to t1)

> tn

(introduce t2)

(a) history was being made by Ross Perot and his volunteers. (b) Ross Perot and his volunteers were making history. (‘t’ stands for paragraph topic) Figure 1.

ENGLISH PASSIVES WITH INDEFINITE SUBJECTS 5.

135

Conclusion

This paper has focused on one type of English passive with unique discourse characteristics: although the subject of such a passive is indefinite in form, it is semantically related to previous mentions. We analyzed this type of passive from the standpoint of both preceding and following contexts. As a concluding remark, I would like to add that the purpose of this paper was not to propose a rule or a condition on rules regarding the use of passives with indefinite subjects. Rather it is intended to account for the unique linguistic behavior of this particular type of passive by demonstrating the importance of context-dependent considerations. Keeping this purpose in mind, I have restricted myself to pointing out ‘tendencies’ rather than proposing ‘rules’ or ‘conditions’.2 Reference was made to Prince’s (1981) inferrable entities and Hawkins’ (1978) concept of associative anaphora. What is common to Prince’s inferrable entities, Hawkins’ associative anaphora, and passive subjects of the type discussed in this paper is that they are all context-bound and cannot therefore be construed properly without reference to something already mentioned in the preceding context. It has been made clear from actual data that one tendency observed in the use of the English passive is for elements that are related to what has already been mentioned to occur in subject position. As repeatedly mentioned, this is consistent with and results from general information structures widely agreed upon, and this can explain why passives with indefinite subjects can sound natural in certain contexts. We have also discussed the discourse-topic introductory role of passive agent phrases, for which consideration of the following context is essential. All the tendencies discussed in this paper can be explained on the basis of a requirement to link linguistic entities properly to both preceding and following contexts, a requirement which is met by means of associative chain formation. Our observation of passives with indefinite subjects thus demonstrates that one of the functions of the English passive is to ensure a natural flow of discourse by adequately linking linguistic entities to both preceding and following contexts.

136

AIKO UTSUGI

Notes 1.

Dusková (1971:136) considers such expressions as ‘similar’ and ‘such’ to be anaphoric. She does not explain why she does so. I assume that she calls these expressions anaphoric because they require an already-mentioned entity to be construed properly.

2.

Haas (1973:147) notes that there are two kinds of regularity or norm in language: (i) constraints or rules, which characterise the form of acceptable utterance, and (ii)tendencies, or habits which characterise their meanings. For a more detailed account of the differences between the notions of ‘rule’ and ‘tendency’, see Utsugi (1994). The suggestion is made there that the application of rules is not affected by contextual factors but that in the discussion of tendencies, context-depedent considerations are essential.

References Brown, Gillian and Yule, George 1983 Discourse analysis. Cambridge: Cambridge University Press. Cole, Peter (ed) 1981 Radical pragmatics. New York: Academic Press. Corblin, Francis 1990 “Anaphoric and referential chains in discourse”. Rivista di Linguistica 2: 67-89. Dik, Simon C. 1989 The theory of functional grammar Part I: The structure of the clause. Dordrecht: Foris. Dusková, Libuse 1971 “On some functional and stylistic aspects of the passive in present-day English”. Philologica Pragensia 14: 117-143. Hawkins John A. 1978 Definiteness and indefiniteness: A study in reference and grammatical predication. London: Croom Holm. Kuno, Susumu 1972 “Functional sentence perspective”. Linguistic Inquiry 3: 269-320. Kuno, Susumu and Etsuko Kaburaki 1977 “Empathy and syntax”. Linguistic Inquiry 8: 627-672. Prince, Ellen F. 1979 “On the given/new distinction”. CLS 15: 267-278. Prince, Ellen F. 1981 “Toward a taxonomy of given/new information”. in P. Cole (ed), 223-255. Utsugi, Aiko 1994 The structure and use of the English passive with special reference to passives in be with a by-agent. Dissertation, Cambridge University. Ward, Gregory L. and Ellen F. Prince 1991 “On the topicalization of indefinite NPs”. Journal of Pragmatics 16: 167177.

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

137

Observations on Three Anaphoric Expressions in Japanese* Etsuko Kaburaki Tomoda Shukutoku University

1.

Introduction

Observe the following sentences: (1)

(zero-pronoun) Midorii wa Øi (ga)1 ima kariteiru ie ni moo 20-nen mo sundeiru. Ø now is renting house 20-years no less than has lived ‘Midori has lived in the house which she is renting now for no less than 20 years.’

(2)

(zibun) Midorii wa zibuni ga ima kariteiru ie ni moo 20-nen mo sundeiru. self ‘ Midori has lived in the house which she is renting now for no less than 20 years.’

(3)

(honnin) Midorii wa honnini ga ima kariteiru ie ni moo 20-nen mo sundeiru. self ‘ Midori has lived in the house which she is renting now for no less than 20 years.’

* This paper is a revised and greatly extended version of Tomoda (1989, 1990, 1992). I am greatly indebted to Akio Kamio, Tomoko Yasutake, Harumi Sawada, Masa-hito Kubota, Kenichi Takami, and William J. Herlofsky for their valuable comments on the earlier versions of the paper.

138

ETSUKO KABURAKI TOMODA

All these sentences are grammatical in the interpretation in which the zeropronoun, zibun and honnin are coreferential with Midori. It seems, however, that (1), in which the zero-pronoun is used, is the most natural sentence of the three. The use of zibun or honnin causes some semantic difference. In some contexts, the use of zibun or honnin is more appropriate. In this paper, I will discuss the distributional properties of these three anaphoric expressions in Japanese.2

2.

A Comparison of Three Anaphoric Expressions

2.1 Subjectivity Kaburaki(1973a) first noted that in some contexts zibun behaves as a subjective expression. The notion of subjective expression was discussed later in Kaburaki (1973b), Kuno and Kaburaki (1975), Kuno (1978), and Kuno (1987). It might be defined as follows: Subjective Expression: Subjective expression is a description from the side of the participant in a sentence with whom the speaker is empathizing.3 Though it is impossible to enumerate all the members of the class of subjective expressions here, it is possible to give some examples; emotive adjectives like ‘itosii (beloved)’, ‘natukasii (dear old)’, and ‘netamasii (enviable)’ and nouns indicating human relationships like ‘sensei (teacher)’ and ‘otoosan (father)’ seem to be included in this class.4 The following examples, which are based on Kuno (1972)’s Awareness Constraint, illustrate the distributional parallelism between zibun and sensei: (4)

a. Tarooi wa, minna ga zibuni no waruguti o itteiru toki, Taro people self speak-ill-of when tonari no heya de nusumigiki site ita. next room in eavesdropping was ‘Taro was eavesdropping in the next room when people were speaking ill of him.’ b. *Tarooi wa, minna ga zibuni no waruguti o itteiru toki, Taro people self speak-ill-of when

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

139

tonari no heya de (tomodati to) zatudan site ita. next room in (with his friends) chatting was ‘Taro was chatting (with his friends) in the next room when people were speaking ill of him.’ (5)

a. Taroo wa, minna ga sensei no waruguti o itteiru toki, Taro people teacher speak-ill-of when tonari no heya de nusumigiki site ita. next room in eavesdropping was ‘Taro was eavesdropping in the next room when people were speaking ill of his teacher.’ b. Taroo wa, minna ga sensei no waruguti o itteiru toki, Taro people teacher speak-ill of when tonari no heya de (tomodati to) zatudan site ita. next room in (with his friends) chatting was ‘Taro was chatting (with his friends) in the next room when people were speaking ill of his teacher.’

In the interpretation in which Taroo and zibun are coreferential, (4a) is grammatical but (4b) is not. This is because the speaker is empathizing with Taro only in (4a). Examine (5a) and (5b). In the interpretation in which sensei means Taro’s teacher, (5a) is grammatical but (5b) is not. This leads us to claim that in this context sensei refers to Taro’s teacher when the speaker is empathizing with Taro. Zibun is also parallel to sensei in that in simple sentences zibun uniquely refers to the subject and sensei means the subject’s teacher (Subject Antecedent Condition or SAC). Consider the following: (6)

a. Tarooi wa Ziroo ni zibuni no hanasi o sita. Taro Ziro self’s story did ‘Taro talked to Ziro about himself.’ b. Zirooi wa Taroo ni zibuni no hanasi o sareta. Ziro Taro self’s story was-done ‘Ziro was talked to by Taro about himself.’

(7)

a. Taroo wa Ziroo ni sensei no hanasi o sita. Taro Ziro teacher’s story did ‘Taro talked to Ziro about his teacher.’

140

ETSUKO KABURAKI TOMODA b.

Ziroo wa Taroo ni sensei no hanasi o sareta. Ziro Taro teacher’s story was-done ‘Ziro was talked to by Taro about his teacher.’

Zibun refers to Taroo in (6a), while it refers to Ziroo in (6b). Similarly, sensei in (7a) is Taro’s teacher and sensei in (7b) is Ziro’s teacher. Zibun and other subjective expressions behave similarly even in N.A. McCawley’s (1972) well-known counterexamples to SAC. (8a), (9a), and (10a) given below are three of N.A. McCawley’s examples. Note that otoosan (father) in the b versions of (8), (9), and (10) is a kind of subjective expression and behaves like zibun. (8) to (10) resist SAC because the speaker is empathizing with Hiroshi or the Premier Sato. In other words, the empathy constraint is stronger here than SAC. (8)

a. Zibuni ga gan kamosirenai koto ga Hirosii o nayamaseta. self cancer might-have Hiroshi worried ‘That he might have cancer worried Hiroshi.’ b. Otoosan ga gan kamosirenai koto ga Hirosi o nayamaseta. father cancer might-have Hiroshi worried ‘That his father might have cancer worried Hiroshi.’

(9)

a. Koibito ga zibuni o uragitta koto ga Hirosii o hungaisaseta. girlfriend self betrayed Hiroshi angered ‘That his girlfriend betrayed him angered Hiroshi.’ b. Koibito ga otoosan o uragitta koto ga Hirosi o hungaisaseta. girlfriend father betrayed Hiroshi angered ‘That his girlfriend betrayed his father angered Hiroshi.’

(10) a. Zibuni no ninki ga oti-dasita to iu uwasa ga Satoo-Syusyooi self’s popularity started to drop rumor Premier Sato o iraira sasete iru. irritates ‘The rumor that his popularity has started to drop irritates Premier Sato.’ b. Otoosan no ninki ga oti-dasita to iu uwasa ga Satoo-Syusyoo father’s popularity started to drop rumor Premier Sato o iraira sasete iru. irritates ‘The rumor that his father’s popularity has started to drop irritates Premier Sato.’

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

141

Otoosan in the b versions of (8) and (9) is Hiroshi’s father and otoosan in (10b) is Premier Sato’s father respectively. Thus far, we have given examples which indicate the distributional parallelism between zibun and other subjective expressions. Curiously enough, honnin shows the same behavior as illustrated in the following examples: (11) a. Tarooi wa, minna ga honnini no waruguti o itteiru toki, Taro people self speak-ill-of when tonari no heya de nusumigiki site ita. next room in eavesdropping was ‘Taro was eavesdropping in the next room when people were speaking ill of him.’ b. *Tarooi wa, minna ga honnini no waruguti o itteiru toki, Taro people self speak-ill-of when tonari no heya de (tomodati to ) zatudan site ita. next room in (with his friends) chatting was ‘Taro was chatting (with his friends) in the next room when people were speaking ill of him.’ (12) a. Tarooi wa Ziroo ni honnini no hanasi o sita.5 Taro Ziro self’s story did ‘Taro talked to Ziro about himself.’ b. Zirooi wa Taroo ni honnini no hanasi o sareta. Ziro Taro self’s story was-done ‘Ziro was talked to by Taro about himself.’ (13) a. Honnini ga gan kamosirenai koto ga Hirosii o nayamaseta. self cancer might-have Hiroshi worried ‘That he might have cancer worried Hiroshi.’ b. Koibito ga honnini o uragitta koto ga Hirosii o hungaisaseta. girlfriend self betrayed Hiroshi angered ‘That his girlfriend betrayed him angered Hiroshi.’ c. Honnini no ninki ga oti-dasita to iu uwasa ga self’s popularity started to drop rumor Satoo-Syusyooi o iraira sasete iru. Premier Sato irritates ‘The rumor that his popularity has started to drop irritates Premier Sato.’

142

ETSUKO KABURAKI TOMODA

(11a) and (11b) show that the occurrence of honnin depends on the presence or absence of empathy for Taro. (12a) and (12b) show that honnin also obeys SAC. Even in examples like (13a), (13b) and (13c), which violate SAC, honnin is parallel to zibun and other subjective expressions. What about the zero-pronoun? Note that there is no grammatical difference between (14a) and (14b): (14) a. Tarooi wa, minna ga Øi (no)waruguti o itteiru toki, Taro people speak-ill-of when tonari no heya de nusumigiki site ita. next room in eavesdropping was ‘Taro was eavesdropping in the next room when people were speaking ill of him.’ b. Tarooi wa, minna ga Øi (no) waruguti o itteiru toki, Taro people speak-ill-of when tonari no heya de (tomodati to ) zatudan site ita. next room in (with his friends) chatting was ‘Taro was chatting (with his friends) in the next room when people were speaking ill of him.’ These examples show that the zero-pronoun is not an empathy pronoun. Finally, the contrast between (15a) and (15b) and that between (16a) and (16b)6 also lend support to the claim that zibun and honnin indicate speaker’s empathy. (15a) and (15b) are taken from Kuno (1987). (15) a. Kawada-kyoozyui wa, si-go suguni, zibuni no ie kara Prof. Kawada death-after self’s house from soogi-zyoo ni utusareta. funeral home transferred ‘Prof. Kawada was transferred from his house to a funeral home immediately after his death.’ b. *Kawada-kyoozyui wa, si-go suguni, zibuni no soogi-zyoo Prof. Kawada death-after self’s funeral home ni utusareta. transferred ‘Prof. Kawada was transferred to his funeral home immediately after his death.’

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

143

(16) a. Kawada-kyoozyui wa, si-go suguni, honnini no ie kara Prof. Kawada death-after self’s house from soogi-zyoo ni utusareta. funeral home transferred. ‘Prof. Kawada was transferred from his house to a funeral home immediately after his death.’ b. *Kawada-kyoozyui wa, si-go suguni, honnini no Prof. Kawada death-after self’s soogi-zyoo ni utusareta. funeral home transferred ‘Prof. Kawada was transferred to his funeral home immediately after his death.’ Here again, the use of the zero-pronoun cancels the difference in grammaticality between (15a) and (15b) and that between (16a) and (16b). Note that (17a) and (17b) given below are grammatical: (17) a. Kawada-kyoozyui wa, si-go suguni, Øi (no) ie7 kara Prof. Kawada death-after house from soogi-zyoo ni utusareta. funeral home transferred ‘Prof. Kawada was transferred from his house to a funeral home immediately after his death.’ b. Kawada-kyoozyoi wa, si-go suguni, Øi (no) soogi-zyoo ni Prof. Kawada death-after funeral home utusareta. transferred ‘Prof. Kawada was transferred to his funeral home immediately after his death.’ 2.2 Contrastive and Emphatic Use As is well known, there are several uses of zibun which are not controlled by empathy. One of such uses is the contrastive and emphatic use. Observe the following:

144

ETSUKO KABURAKI TOMODA (18) *Yasuei wa, zibuni no soosiki no toki, dare mo hana o okuru Yasue self’s funeral when nobody flowers hito ga nakatta. send ‘Nobody sent Yasue flowers on her funeral.’

The ungrammaticality of (18) is due to the fact that zibun cannot refer to the dead person in this context. The empathy constraint is working here. Compare this with the following example in which zibun refers to Yasue: (19) Yasuei wa, hito no soosiki no toki ni wa yoku hana o okutta ga, Yasue other people’s funerals when often flowers sent but zibuni no soosiki no toki wa, dare mo hana o okuru hito ga nakatta. self’s funeral when nobody flowers sent ‘Yasue often sent flowers on other people’s funerals, but nobody sent her flowers on her funeral.’ (19) is grammatical because zibun here is used in its contrastive and emphatic use. Interestingly enough, honnin shows the same behavior as zibun, which is clear from the following contrast: (20) *Yasuei wa, honnini no soosiki no toki, dare mo hana o okuru Yasue self’s funeral when nobody flowers hito ga nakatta. send ‘Nobody sent Yasue flowers on her funeral.’ (21) Yasuei wa, hito no soosiki no toki ni wa yoku hana o okutta ga, Yasue other people’s funerals when often flowers sent but honnini no soosiki no toki wa, dare mo hana o okuru self’s funeral when nobody flowers hito ga nakatta. send ‘Yasue often sent flowers on other people’s funerals, but nobody sent her flowers on her funeral.’ What about the zero-pronoun? If we change zibun and honnin in (18) and (20) for the zero-pronoun, the following grammatical sentence results:

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

145

(22) Yasuei wa, Øi (no) soosiki no toki, dare mo hana o okuru hito Yasue funeral when nobody flowers ga nakatta. send ‘Nobody sent Yasue flowers on her funeral.’ Compared to (18) to (21), (22) is an objective statement of the fact. The following sentence, which corresponds to (19) and (21), is ungrammatical because the zero-pronoun cannot have the contrastive and emphatic sense:8 (23) *Yasuei wa, hito no soosiki no toki ni wa yoku hana o okutta ga, Yasue other people’s funerals when often flowers sent but Øi (no)soosiki no toki wa dare mo hana o okuru hito ga nakatta. funeral when nobody flowers send ‘Yasue often sent flowers on other people’s funerals, but nobody sent her flowers on her funeral.’ 2.3 Humanness Although zibun can refer to higher animals, honnin uniquely refers to humans. This may be because of the expression ‘hon-nin’, where ‘-nin’ means ‘human’. (24), in which ‘sasori’ is considered a higher animal, is ungrammatical for this reason: (24) *Sasorii wa, minnna no tame ni aayatte, sora no mezirusi scorpion others in that way mark in the sky ni natte, honnini no karada o moyasi-tuzukete irunda ne. become self’s body continue to burn ‘The scorpion seems to continue to burn his body in that way as a mark in the sky to save other animals.’ In contrast, the following examples involving zibun and the zero-pronoun are grammatical: (25) Sasorii wa, minna no tame ni aayatte, sora no mezirusi ni natte, scorpion others in that way mark in the sky become zibuni no karada o moyasi-tuzukete irunda ne. self’s body continue to burn ‘The scorpion seems to continue to burn his body in that way as a mark in the sky to save other animals.’

146

ETSUKO KABURAKI TOMODA (26) Sasorii wa, minna no tame ni aayatte, sora no mezirusi ni natte, scorpion others in that way mark in the sky become Øi (no)karada o moyasi-tuzukete irunda ne. body continue to burn ‘The scorpion seems to continue to burn his body in that way as a mark in the sky to save other animals.’

2.4 Person Zibun can refer to the first person, second person and third person, whereas honnin can refer only to the third person. (27a) and (27b) are ungrammatical for this reason:9 (27) a. *Watasii wa, honnini no kurasu no tannin no sensei to, I self’s class teacher with doomo umaku yatte-ikenai. get along cannot ‘I cannot get along with my class teacher.’ b. *Honnin no heya gurai honnin de soozi sinasai. self’s room self clean ‘Clean your room for yourself.’ Both of these sentences become well-formed if we replace honnin with zibun. The following sentence, with the zero-pronoun, is also grammatical: (28) Watasii wa, Øi (no) kurasu no tannin no sensei to, I class teacher with doomo umaku yatte-ikenai. get along cannot ‘I cannot get along with my class teacher.’ (27b) remains ungrammatical even if honnin is replaced with the zeropronoun. This is because the contrastive and emphatic sense is cancelled here.10 In the following new example, the zero-pronoun clearly refers to the hearer: (29) Anatai wa, Øi (no) heya o omoti desu ka. you room have ‘Do you have your own room?’

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

147

There are examples in which honnin appears to be coreferential with the first person. The following example is one of them: (30) Watasii to iu honnini ga ima kono basyo ni tatte iru no desu. I self now here standing is ‘I myself is standing here now.’ (30) seems to be an apparent counterexample to the above hypothesis.The grammaticality of (30) may be due to the fact that the speaker considers himself as a third person in this sentence. In other words, the speaker seems to be looking at himself objectively here. 2.5 Semantic Difference Honnin’s contrastive and emphatic meaning seems stronger than that of zibun. Moreover, honnin has an inherent meaning of ‘this person and not others’. The zero-pronoun, on the other hand, has no contrastive and emphatic meaning. Compare the following triplet: (31) a. Kodomo ga tiisai no de, Naokoi wa, Øi (no) ie de dekiru sigoto children young because Naoko house work o sagasite-iru. looking for-is ‘Naoko is looking for some piecework to do at home because she has young children.’ b. Kodomo ga tiisai no de, Naokoi wa zibuni no ie de children young because Naoko self’s house dekiru sigoto o sagasite-iru. work looking for-is ‘Naoko is looking for some piecework to do at home because she has young children.’ c. Kodomo ga tiisai node, Naokoi wa honnini no ie de children young because Naoko self’s house dekiru sigoto o sagasite-iru. work looking for-is ‘Naoko is looking for some piecework to do at home because she has young children.’

148

ETSUKO KABURAKI TOMODA

Honnin seems to have the strongest meaning of contrast and emphasis of the three examples given above. 2.6 Backward Anaphora There are cases in which honnin resists backward anaphora in spite of the fact that the use of honnin does not cause semantic anomaly. Consider (32) (from Ken-ichi Takami, personal communication): (32) ??Honnini ga osieta gakusei dake o, Yamada-senseii wa kaisya self taught students only Prof. Yamada company ni suisen-sita. recommended ‘Prof. Yamada recommended to the company only the students whom he had taught.’ (32) is very awkward. In contrast, the following examples with zibun and the zero-pronoun are well-formed: (33) a. Zibuni ga osieta gakusei dake o, Yamada-senseii wa self taught students only Prof. Yamada kaisya ni suisen-sita. company recommended ‘Prof. Yamada recommended to the company only the students whom he had taught.’ b. Øi (ga) osieta gakusei dake o, Yamada-senseii wa taught students only Prof. Yamada kaisya ni suisen-sita. company recommended. ‘Prof. Yamada recommended to the company only the students whom he had taught.’ If the word order of (32) is changed, the following grammatical sentence results: (34) Yamada-senseii wa, honnini ga osieta gakusei dake o kaisya Prof. Yamada self taught students only company ni suisen-sita. recommended

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

149

‘Prof. Yamada recommended to the company only the students whom he had taught.’ What is important here is that (34) does not involve backward anaphora. Many examples of this sort can be found. For instance, the a versions of (35) to (37) are either ungrammatical or very awkward, whereas their b versions are all grammatical: (35) a. *Honnini ga syoo o torenakatta no de, Hanakoi wa taisoo self prizecould not win because Hanako much zannen-gatta. disappointed ‘Hanako was much disappointed that she could not win the prize.’ b. Hanakoi wa, honnini ga syoo o torenakatta no de, taisoo Hanako self prize could not win because much zannen-gatta. disappointed ‘Hanako was much disappointed that she could not win the prize.’ (36) a. ??Honnini ga kariteiru heya de, Tarooi wa paatii o hiraita. self was renting room in Taro party gave ‘Taro gave a party in the room he was renting.’ b. Tarooi wa, honnini ga kariteiru heya de, paatii o hiraita. Taro self was renting room in party gave ‘Taro gave a party in the room he was renting.’ (37) a. ??Honnini ga muri o ittei-iru koto o self unreasonable thing is saying mitometa ue de, Yookoi wa kangae o aratamenakatta. admitting Yoko opinion did not change ‘Yoko didn’t change her mind, although she admitted that it was unreasonable.’ b. Yookoi wa, honnini ga muri o itte-iru koto Yoko self unreasonable thing is saying o mitometa ue de, kangae o aratamenakatta. admitting opinion did not change

150

ETSUKO KABURAKI TOMODA ‘Yoko didn’t change her mind, although she admitted that it was unreasonable.’

The following examples involving relative clauses are also ungrammatical. These sentences have no forward-anaphora counterparts: (38) *Hanako ni honnini no e o okutta otokoi ga kono daigaku Hanako self’s picture sent man this university ni kinmu site iru. in works ‘The man who sent Hanako his picture works in this university.’ (39) *Honnini no syukudai o oeta gakuseii ga konsaato ni dekaketa. self’s homework finished student concert went ‘The student who had finished his homework went to the concert.’ (38) and (39) become grammatical with zibun and the zero-pronoun. Observe the following: (40) a. Hanako ni zibuni no e o okutta otokoi ga kono daigaku Hanako self’s picture sent man this university ni kinmu site iru. in works ‘The man who sent Hanako his picture works in this university.’ b. Hanako ni Øi (no) e o okutta otokoi ga kono daigaku Hanako picture sent man this university ni kinmu site iru. in works ‘The man who sent Hanako his picture works in this university.’ (41) a. Zibuni no syukudai o oeta gakuseii ga konsaato ni dekaketa. self’s homework finished student concert went ‘The student who had finished his homework went to the concert.’ b. Øi (no) syukudai o oeta gakuseii ga konsaato ni dekaketa. homework finished student concert went ‘The student who had finished his homework went to the concert.’

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

151

It is important to notice, however, that not all the uses of honnin resist backward anaphora. For instance, (13a), (13b) and (13c) are not ungrammatical, although some native speakers might find them a bit awkward. The following sentence is also grammatical in spite of the fact that its structure is similar to that of (35a): (42) Honnini ga kaigi ni derarenakatta no de, Tarooi wa Ziroo o self meeting could not attend Taro Ziro kawari ni syusseiki-saseta. attend-made ‘Taro made Ziro attend the meeting, because he could not attend it.’ We have so far noted that honnin often resists backward anaphora. Why? I conjecture that this is because honnin has a clear inherent meaning ‘ the person concerned and whoever fits the requirement’ (as was pointed out to me by Ken-ichi Takami, personal communication) and that this meaning often prevents it from becoming a pure anaphor.11 Thus, although honnin happens to be coreferential with Taroo in (42), honnin here is a noun having inherent meaning rather than a pure anaphor. Honnin here could be replaced with something like ‘Taroo honnin (Taro himself)’.

3.

The Order of Priority

Observe the following sentences: (43) a. Darekai ga Hanakoj ni zibuni no tomodati no hanasi o sita. someone Hanako self’s friend talked ‘Someone talked to Hanako about his friend.’ b. Darekai ga Hanakoj ni honninj no tomodati no hanasi o sita. someone Hanako self’s friend talked ‘Someone talked to Hanako about her friend.’ Note that zibun in (43a) refers to dareka, whereas honnin in (43b) refers to Hanako. Where does this contrast come from? Our hypothesis is that zibun and honnin function as subjective expressions in certain contexts. Zibun in simple sentences must refer to the subject in accordance with SAC, which is consistent with the Surface Structure Empathy Hierarchy.12 As is predicted,

152

ETSUKO KABURAKI TOMODA

zibun in (43a) is coreferential with dareka. I assume that (43b) has a ‘marked’ interpretation for the following reason: Honnin in (43b) violates SAC and refers to the object because of its strong emphatic meaning. In other words, the use of honnin in contexts where zibun is more naturally used reinforces the special emphatic meaning of the word, resulting in the change of the anaphoric relationship. What is making the problem more complex is that it is difficult for the speaker to empathize with the referent of dareka. In the following example, where Yooko is the subject, honnin is uniquely interpreted as coreferential with Yooko: (44) Yookoi wa Hanako ni honnini no tomodati no hanasi o sita. Yoko Hanako self’s friend talked ‘Yoko talked to Hanako about her friend.’ I assume that although honnin is originally subject-centered, it can refer to the object only when it is difficult for the speaker to empathize with the referent of the subject. What matters here is the order of priority. 1. SAC 2. Who is the subject? (Can the speaker easily empathize with the referent of the subject? ) 3. Inherent meanings of the anaphoric expression (such as contrast and emphasis) The three factors above seem to interact with each other in various ways in determining the antecedent. But what about subjective expressions other than zibun and honnin? Consider the subjective expression natukasii used in the following example: (45) Dareka ga Hanako ni natukasii tomodati no hanasi o sita. someone Hanako dear old friend talked ‘Someone talked to Hanako about her dear old friend.’ In (45) natukasii does not express dareka’s feeling but Hanako’s. If dareka is replaced with Yooko, natukasii becomes an expression from Yoko’s point of view. (46) Yooko wa Hanako ni natukasii tomodati no hanasi o sita. Yoko Hanako dear old friend talked ‘Yoko talked to Hanako about her dear old friend.’

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

153

Finally, let us consider the case where the zero-pronoun is used instead of zibun and honnin: (47) a. Yookoi wa Hanakoj ni Øi (no) tomodati no hanasi o sita. Yoko Hanako friend talked ‘Yoko talked to Hanako about a friend.’ b. Darekai ga Hanakoj ni Øi/j (no) tomodati no hanasi o sita. someone Hanako friend talked ‘Someone talked to Hanako about their/a friend.’ The predominant interpretation of (47a) is the one with the zero-pronoun referring to the subject, Yooko. In (47b), on the other hand, the antecedent of the zero-pronoun is both dareka and Hanako. That is to say, tomodati here is a common friend of theirs. One more interpretation is possible, in which dareka and Hanako are talking about the topic of friendship.

4.

Conclusion and Some Residual Problems

In this paper, I have discussed the distributional properties of three anaphoric expressions in Japanese; zibun, honnin, and the zero-pronoun. Further investigations of these three expressions will surely reveal some essential properties of the Japanese anaphoric system. The long-range purpose of this research is the clarification of the following: 1. Is there any reason why there are three such anaphoric expressions in Japanese? 2. How are these expressions acquired by the native speaker of Japanese? 3. What position does the present study occupy within the universal grammar framework?13 Let us assume that there are degrees of anaphoricity. Honnin may occupy the lowest position of the three. Next is zibun. The zero-pronoun may occupy the highest degree position.14 This is illustrated by the following diagram: anaphoric non-anaphoric expression expression zero-pronoun -------------zibun-------------honnin------------ noun

154

ETSUKO KABURAKI TOMODA

In the above diagram, nouns are supposed to be non-anaphoric expressions, while the zero-pronoun is regarded as a pure anaphoric expression. One of the striking differences between pronouns and nouns may be in their semantic transparency. Pronouns are semantically transparent, whereas nouns clearly have their inherent meanings. The following three facts seem to support this view: 1. Inherent Meaning As was noted in section 2, honnin not only indicates speaker’s empathy like zibun, but also carries the contrastive and emphatic meaning which is stronger than that of zibun. Furthermore, it has the meaning ‘the person concerned and whoever fits the requirement’. Thus, honnin may have the richest meaning of the three anaphoric expressions. Needless to say, the zero-pronoun does not have inherent meaning. 2. Backward Anaphora In 2.6., it was argued that honnin tends to resist backward anaphora. Although there remain many problems about this phenomenon such as those discussed in Tomoda (1993), this point also seems to suggest the low anaphoricity of honnin. 3. Anaphoric Expressions Without Antecedents Having no antecedents implies that the expression is non-anaphoric. While the zero-pronoun must have some antecedent and is always used anaphorically, zibun and honnin are sometimes used without their antecedents. Note that zibun in (48) and (49) has no antecedent: (48) kagami no naka no zibun mirror in self ‘oneself in the mirror’ (49) zibun-tyuusin no kangae self-centered opinion ‘selfish opinion’ Honnin without antecedents are used more frequently. I believe that this is the typical use of honnin. Observe the following:

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

155

(50) Sore wa honnin sidai desu ne. it self depends on ‘It depends on him.’ (51) Byooki-ryooyoo-tyuu no honnin ni kawatte, sick in bed self in place of goaisatu moosiagemasu. greet ‘Let me say something in my sick friend’s place.’ (52) Gookaku-tuutisyo wa, honnin ga torini kite kudasai.15 certificate self get ‘Successful candidates should come to the office in person to get their certificates.’ Moreover, there are cases in which ‘honnin’ is written on a package for the sender’s name when someone sends something to himself. This use of honnin can also be regarded as a non-anaphoric one, although it refers to the writer. Finally, I will present two residual problems before closing the paper. The first problem is about the two inconsistencies found in the present paper. 1. The two uses of zibun and honnin (empathy use and contrastive and emphatic use) seem to contradict each other. What does this mean? 2. Honnin uniquely refers to the third person. This use of honnin appears to conflict with its empathy use. What does this mean? If the answers to the above questions can be found, they would help to clarify the essential properties of the Japanese anaphoric system. The second problem is the use of go-honnin as in the following:16 (53) Tarooi wa, minna ga go-honnini no waruguti o itteiru toki, Taro people self speak-ill-of when tonari no heya de (tomodati to) zatudan site ita. next room in (with his friends ) chatting was ‘Taro was chatting (with his friends) in the next room when people were speaking ill of him.’ (53) is grammatical in contrast with (11)b. Although I have no explanation for this phenomenon at the moment, I conjecture that the addition of go-, one of the honorific prefixes, to honnin cancells the empathy and makes the whole sentence objective. I will leave this problem for future research.

156

ETSUKO KABURAKI TOMODA

Notes 1.

‘Ø’ indicates the zero-pronoun. The particle in the parenthesis never appears when the zero-pronoun is used.

2.

A great deal of papers have been written concerning the distributional properties of zibun. These papers include Kuroda (1965), Kuno (1972), McCawley (1972), Oyakawa (1973, 1974), Inoue (1976), Iida (1992) and Sakakibara (1995). The zero-pronoun has been discussed in Kuroda (1965), Ohso (1976), Kameyama (1988), Iida (1992) and others. However, as far as I know, this paper is the first attempt to discuss the behavior of honnin. Although several other anaphoric expressions like toonin, kare, kanozyo, zisin and otagai are used frequently in Japanese, I will not deal with these expressions in the present paper.

3.

The concept of empathy was first described in Kuno and Kaburaki (1975).

4.

For example, ‘daisukina’ is a subjective expression while ‘sukina’ is not. These two words differ only in the degree of emotion.

5.

‘Honnin’ sometimes resists SAC in these contexts. We will take up this problem in section 3.

6.

Kuno (1987) proposed the principle ‘Aliveness Requirement of Empathy’ to account for this contrast.

7.

Incidentally, the expression ‘zitaku (self’s house)’ may be more appropriate here.

8.

This fact was also pointed out by Ohso (1976).

9.

I am indebted to Ken-ichi Takami (personal communication) for this observation and example (27b).

10.

In the following example, ‘heya’ is contrasted with other places: (i) Ø (no) heya gurai Ø (de) soozi sinasai. room should clean ‘You should clean the room.’ This sentence can be used to mean the following: (ii) Toire wa iyadesyoo ga, heya gurai soozi sinasai. toilet don’t like room should clean ‘I understand you don’t like to clean the toilet, but you should clean the room.’

11.

Zibun also has its inherent meaning, but it can become anaphoric more easily than honnin. In other words, zibun seems to have more transparent meaning than honnin. The meaning of the zero-pronoun is the most transparent of the three in that it has no inherent meaning. I will take up this problem in section 4.

12.

See Kuno and Kaburaki (1975).

13.

For example, since the notion of empathy belongs to the universal grammar, the clarification of the Japanese anaphoric system controlled strongly by empathy will surely contribute to the study of the universal syntax.

OBSERVATIONS ON THREE ANAPHORIC EXPRESSIONS IN JAPANESE

157

14.

There may arise much discussion as to what is a zero-pronoun and what is not. See Kuno (1995), for example.

15.

I am indebted to Ken-ichi Takami (personal communication) for this example.

16.

This fact was pointed out to me by Tomoko Yasutake (personal communication).

References Gabriele, Leslie, Debra Hardison, and Robert Westmoreland (eds) 1995 Syntax 2 and semantics/pragmatics: Proceedings of the sixth annual meeting of the Formal Linguistics Society of Mid-America Vol. 2. Bloomington, IN: Indiana University Linguistics Club. Iida, Masayo 1992 Context and binding in Japanese. Dissertation, Stanford University. Also published from CSLI Publications, 1996. Inoue, Kazuko 1976 “Reflexivization: An interpretive approach”. In M. Shibatani (ed), 117200. Kaburaki (Tomoda), Etsuko 1973a Some remarks on the distribution of the Japanese reflexive zibun. Master’s thesis, Meiji Gakuin University. Kaburaki (Tomoda), Etsuko 1973b “Nihongo saiki-daimeisi zibun ni tuite no teigen”. Nebulous (MeijiGakuin Daigaku-in Eibungaku-Senkooka, Tokyo ) 2: 19-52. Kameyama, Megumi 1988 “Japanese zero pronominal binding: Where syntax and discourse meet”. Papers from the second international workshop on Japanese syntax: 4774. Kuno, Susumu 1972 “Pronominalization, reflexivization, and direct discourse”. Linguistic Inquiry 3: 161-195. Kuno, Susumu 1978 Danwa no bunpoo. Tokyo: Taishukan. Kuno, Susumu 1987 Functional syntax: Anaphora, discourse and empathy. Chicago: The University of Chicago Press. Kuno, Susumu 1995 “Null elements in parallel structures in Japanese”. In R. Mazuka and N. Nagai (eds), 209-233. Kuno, Susumu (ed) 1975 Harvard studies in syntax and semantics 1. Cambridge, MA: Department of Linguistics, Harvard University. Kuno, Susumu and Etsuko Kaburaki (Tomoda) 1975/1977 “Empathy and syntax.” in S. Kuno (ed), 1-73. Also in Linguistic Inquiry 8: 627-672.

158

ETSUKO KABURAKI TOMODA

Kuroda, Shige-Yuki 1965 Generative grammatical studies in the Japanese language. Dissertation, MIT. Mazuka, Reiko and Noriko Nagai (eds) 1995 Japanese sentence processing. Hillsdale, NJ: Lawrence Erlbaum. McCawley, Noriko A. 1972 A study of Japanese reflexivization. Dissertation, University of Illinois. Ohso, Mieko 1976 A study of zero pronominalization in Japanese. Dissertation, The Ohio State University. Oyakawa, Takatsugu 1973 “Japanese Refexivization, 1”. Papers in Japanese Linguistics 2: 49-135. Oyakawa, Takatsugu 1974 “Japanese Reflexivization, 2”. Papers in Japanese Linguistics 3: 129201. Sakakibara, Sonoko 1995 “Pragmatic-only account of the Japanese reflexive pronoun zibun”. In L. Gabriele, D. Hardison, and R. Westmoreland (eds), 150-161. Shibatani, Masayoshi (ed) 1976 Syntax and semantics 5: Japanese generative grammar. New York: Academic Press. Tomoda, Etsuko K. 1989 “Honnin”. Syukutoku Tanki-Daigaku Kenkyuu Kiyoo 28: 65-71. Tomoda, Etsuko K. 1990 “Sarani honnin ni tuite”. Syukutoku Tanki-Daigaku Kenkyuu Kiyoo 29: 25-29. Tomoda, Etsuko K. 1992 “Nihongo no syooou-hyoogen no yoohou ni kansite”. Syukutoku TankiDaigaku Kenkyuu Kiyoo 31: 31-37. Tomoda, Etsuko K. 1993 “Nihongo ni okeru gyakkou-shooou ni kansuru oboe-gaki”. Syukutoku Tanki-Daigaku Kenkyuu Kiyoo 32: 61-69.

QUALIFICATION AND POINT OF VIEW

159

Qualification and Point of View* Jacqueline Guillemin-Flescher University of Paris VII

The Theoretical Framework The theoretical framework of this paper is the ‘théorie des opérations énonciatives’ initiated by Emile Benveniste but essentially developed by Antoine Culioli. For a detailed account of the theory in English, one should refer to Antoine Culioli (1995). Within the scope of this paper, we can only give an outline of the parameters on which the theory is based, with special emphasis on the points that are relevant to the discussion that follows. The theory is embedded in a conception of language that rejects the dichotomy between ‘langue’: the abstract system of a given language and ‘langage’: linguistic activity based on usage. It is grounded in a relationship between the speaker and his discourse, but the speaker is included in the theory in so far as there are formal traces of his presence. Viewed in this perspective, he is both a reference point in the calculation of time, space and personal pronouns and a cognitive origin in the expression of modality. For instance, the present tense: John is playing the piano coincides with the time of utterance, the preterite implies, either that the process is located at a time previous to the time of utterance as in: John played the piano or, in the case of a specific event: John broke the vase / John left the room that the time of the event is independent of the time of utterance. This approach entails a distinction between a well formed sentence such as: Two men are in the garden (syntactically correct) and a well formed * I am indebted to Claude Charreyre, Claude Rivière, Janine Bouscaren and more particularly to Stéphane Robert of the University of Paris VII for their valuable comments on this paper, and to John Holstead and Paul Volsik of the University of Paris VII, and Nigel Turner of the University of Paris XII, for testing examples.

160

JACQUELINE GUILLEMIN-FLESCHER

utterance: There are two men in the garden. A well formed utterance takes into account location of the predicative relation with respect to the situation of enunciation. This situation includes the three parameters mentioned above: the speaker, the time of utterance and the place of utterance. The meaning of a word in discourse is not given but constructed by means of various specifications. The initial construct is a notion, a mental representation to which one associates physico cultural properties of a cognitive nature; for instance the notion chocolate to which properties such as eatable / sweet / brown might be associated. In some cases the properties will vary from one culture to another. For instance the notion policeman will tend to suggest neutral or negative properties in a French context, positive ones in a British context. Determination i.e. specification, is not simply a question of syntactic markers such as determiners or quantifiers. It is based on two kinds of operation: quantification and qualification, whereby the indefinite character of the notion is reduced. For instance, the existential sentence: There is a bar of chocolate on the table will specify the notion chocolate by giving it existential status and by relating it to a format on the one hand and spatial location on the other (quantification). It therefore acquires greater determination. A notion can be fragmented into a class of units or occurrences as in: chocolates should not be consumed in large quantities. Chocolates are here abstract units derived from the fragmentation of the notion chocolate. The units expressed by the plural are viewed with respect to the properties common to a class of elements rather than with respect to individual entities (i.e. they are viewed as a kind or type; in other words qualitatively rather than quantitatively). In the following example however: The chocolates I gave you have different fillings, chocolates are viewed as a set of entities and no longer as a kind or type (quantification). This distinction can apply to occurrences represented by nouns or verbs. For instance, if we say: He is very gifted: he dances and sings and plays the piano, we are not defining events that are time bound and therefore quantified, but activities that are typical of the subject he, and are therefore essentially qualitative. In He sang Ave Maria at the concert last night, the process He sang is located with respect to time and place. This implies a specific occurrence that has acquired existential status (quantification). The notion is organized with respect to a propotype which corresponds to the minimal agreement that can be reached amongst speakers as to the

QUALIFICATION AND POINT OF VIEW

161

properties of a particular entity i.e. if you talk about a brown sweater it implies a minimum agreement amongst speakers as to the notion sweater and as to the notion brown. The propotype can function in two different ways: it can either be a reference point which enables speakers to identify an occurrence with the prototype of the notion as in David is a Frenchman (organizing center) or as the extreme degree of the property, the property par excellence: for instance David is so French!. In the second case it will be defined as an attractor (attracting center). If one compares David is a Frenchman with David is so French!, it will be obvious that the identification of David with an occurrence of the type French corresponds to an objective statement, whereas David is so French! implies a subjective evaluation of David’s degree of correspondence to the type (i.e. high degree). Likewise, if you say: He sings whenever he is in the bathroom, you are stating an objective fact, whereas if you say How he sings! you are referring to an intensive quality of the process. The process is not viewed objectively but with respect to the assesment of a specific speaker. In this instance the predicative relation involves the point of view of the speaker. Point of view is directly related to modality, since modality within this theoretical framework concerns the speaker’s attitude to his utterance. In the case of an objective statement as in the example above David is a Frenchman, the speaker guarantees the validity of the predicative relation but there is no specific point of view involved since it can be validated by any speaker by virtue of the fact that it can be verified by external criteria. In the case of David is so French! the criteria are purely subjective. They involve the speaker’s point of view only. Modality interacts with other grammatical categories such as aspect and determination. When a process is viewed as a whole and specified with respect to time and space as: John bought a new car on Monday, there is no specific point of view involved. However, when the process is viewed with respect to a specific phase, as in John is mowing the lawn / has mown the lawn, the process is viewed from the point of view of the speaker. One can only determine whether an end point has been reached or not with respect to a specific point of view. Simple verb forms such as the simple present or past are compatible with point of view only when they are viewed qualitatively as in How he sings! / How he sang! . If they refer to a completed action as in He broke the vase, no specific point of view is involved : *How he broke the vase!.

162

JACQUELINE GUILLEMIN-FLESCHER

Point of view is an essential factor in the paper that follows. Our objective is to show that point of view is linked to a specific speaker and therefore to a subjective assessment of the utterance. The term or process that is evaluated must moreover be compatible with a qualitative value. ***** 1.

Introduction

The predication of a property can be expressed in a number of ways: (1)

Carlotta Adams was an American girl... (A. Christie, Lord Edgware Dies, p.7)

(2)

He’s so good-looking... (Ibid., p.17)

(3)

How beautiful the sound is ! (V. Woolf, A Haunted House, p.24)

(4)

There was a bitterness in his last words... (A. Christie, Lord Edgware Dies, p. 31)

There are obviously significant differences between these examples. The question is whether they can be explained purely on the grounds of syntactic constraints. Our assumption is that they cannot.1 Basic distinctions have to be established between objective statement, subjective assertion and exclamation. These imply a complex network of relations that involve determination, modality, and to a certain extent, aspect. Semantic properties, syntactic restrictions and prosody are conditioned by these relations, in which point of view plays an essential role. A contrastive approach will be used wherever possible, to bring out specific theoretic points that are relevant to general linguistics and to show both the extent and the limitations of generalisation.

2.

Objective Statement

Adjectives will necessarily be a significant syntactic category in the type of predication we are here concerned with. Their semantic properties and the syntactic restrictions that they entail cannot, however, be pre-established. They depend on the nature of the utterance: objective statement or subjective

QUALIFICATION AND POINT OF VIEW

163

assertion. In an objective statement such as: (5)

Lord Edgware was a tall man of about fifty. (A. Christie, Lord Edgware Dies, p. 33)

the semantic property associated with the adjective tall is viewed as a characteristic that can be established by external criteria. In: (1)

Carlotta Adams was an American girl.

American is viewed as an established fact. No specific point of view is here involved. The relation between the subject and the predicate is presented as one that holds without reference to a specific point of view — in other words, that can be verified on objective grounds. Certain adjectives are obviously more compatible than others with a reference system that does not imply subjective evaluation. However, even adjectives referring to height and nationality can be used with modal implications.2 The modal distinction between the two uses is directly linked to constraints that concern determination. As the following examples show, a difference often appears in this respect between English and French: (6)

She was a tall girl, ungainly, with a long waist and long legs. Elle était grande et un peu gauche, avec un buste allongé et de belles jambes... (J. Le Carré, The Spy who came in from the Cold, p. 26 / 41)

(7)

Myself, I’m a Brit first, a Pole second and everything else comes afterwards. Moi je suis anglais d’abord, polonais ensuite, et tout le reste vient après. (J. Le Carré, The Russia House, p. 55 / 60)

It must be noted that a literal translation would generally be considered ungrammatical for example (6) and (7): * Elle était une grande fille...3 * Je suis un anglais d’abord, un polonais ensuite... example (6) would, however, be acceptable with a deictic pronoun: C’était une grande fille... (This point will be explained on page 163.)

164

JACQUELINE GUILLEMIN-FLESCHER

In English the notions tall, Brit, Pole are fragmented into classes of entities, one of which is singled out to exemplify the property that is predicated. The indefinite article + the noun indicate that one element has been extracted from the class. In French, on the other hand, the adjective simply indicates a quality of the subject without any specification of quantity.4 The same difference applies when the quality is expressed by a noun in both languages: (8)

Harry’s a lawyer Harry est homme de loi (J. Le Carré, The Russia House, p.84 / 94)

(9)

And you are not and never have been a member of a Communist party...? Et vous n’êtes pas, ni n’avez jamais été, membre du parti communiste ? (Ibid., p. 62 / 69)

(10) I am not a nihilist. I am a humanist. Je ne suis pas nihiliste, mais humaniste (Ibid., p. 200 / 232) The zero + noun construction in French cannot be used freely. It is subject to two restrictions which are generally defined as follows: a) b)

the qualifying noun must be related to a subject referring to a Human Animate the semantic properties of the qualifying noun must refer to what is usually called “a social function”.

The traditional formulation needs qualification. It cannot be considered as a linguistic criterion. Example (10) moreover, shows that this restriction does not necessarily apply. It would seem more adequate to refer to properties that are - either characteristic of a human type: Bernard Kouchner est médecin - or that correspond to a recognized status: Marie-José Perec est médaille d’or This would explain why the following examples are ungrammatical:

QUALIFICATION AND POINT OF VIEW

165

* Il est héros * Il est voisin de palier The noun héros cannot have a classifying function since it implies subjective evaluation. The term voisin is purely relational: in other words, definable only in relation to a specific Human Subject. The fundamental criterion for the use of a zero + noun construction is a property or set of properties that can be defined objectively. The difference that appears between English and French seems at first sight to indicate that constraints governing predication of a property5 vary from one language to another. This would not, however, be stating the problem adequately. In French one simply qualifies the subject. The zero marker before the noun expressing the property implies that the property is viewed qualitatively.6 In English, the second term is quantified. A second criterion must, however, be taken into account. Even when the construction: C’était une grande fille is used in French (cf. p. 161), there is a notable difference between the two languages. The deictic pronoun C’ foregrounds a term defined in a previous context and not the existence of the defining property. One must note, moreover, that this construction is only possible when referring to a third person, as above, or to a non human element, as in C’était un beau jardin. One point remains to be argued. How can one account for the fact that a purely objective statement about the existence of a defining property cannot be expressed by an existential sentence? If we compare: (11) She was a quiet girl with a pleasant low voice. (A. Christie, Lord Edgware Dies, p. 20) with the existential sentence: (12) There was a kind of nervous tension about him... (Ibid., p. 26) we note that in both cases the property (quietness / nervous tension) is related to a Human Subject7 (She / him). This is necessarily the case since you cannot refer to the property of a Human Subject without establishing an explicit link between the two terms. But whereas in (11) the term she is in initial position in (12) him is in final position in a prepositional phrase. The essential difference lies in the status of the predicate. All existential sentences relating to a property predicated of a Human Animate have a

166

JACQUELINE GUILLEMIN-FLESCHER

common characteristic. Let us quote a few examples that make this clear: (13) There was something faintly disagreeable in her manner. (A. Christie, Lord Edgware Dies, p. 23) (14) There was something strangely likeable about the young man. (Ibid., p. 22) (15) ...there was something vaguely effeminate that I disliked about the softness of his voice. (Ibid., p. 33) (16) There was a curious scrutinizing quality in the girl’s gaze... (Ibid., p. 21) (17) ...there was a very definite hostility in those pale blue eyes. (Ibid., p. 21) In the above examples as in (4) and (12) a) the semantic properties of the predicate and b) the adjective or adverb that qualifies it, when such is the case, express subjective appreciation of a property. The judgment is moreover formulated with reference to a given context. Our hypothesis is that when one states the existence of a defining property, without subjective appreciation or reference to a given context, the subject is in initial position and the predicate follows the copula be. Existential sentences on the other hand, as our examples clearly show, express a personal judgment concerning a Human Animate as in (12) and (14) or concerning the property of a Human Animate ((13) (15) (16) (17)) and at the same time state the existence of the property. In examples such as (6) (7) and (11) the subject, though more determined than the predicate, since it is already identified, can appear in its usual position because it is a necessary term in the relation. The degree of symmetry in a clause containing the copula be will depend on the degree of definiteness of the terms related. When the predicate is expressed by an adjective or a noun without an article, there can be no symmetry because it has a purely qualitative value. When the adjective is preceded by an indefinite article and followed by a noun, as in a quiet girl, the post-verbal element has the necessary degree of definiteness (i.e. a representative element in a class) to constitute a distinct term. The reversal of the order observed in existential

QUALIFICATION AND POINT OF VIEW

167

sentences will be justified only to express subjective appreciation of the predicate. One notes in this respect that, in a narrative context, the modal properties of adjectives used in the subject verb object construction, for instance quiet in (11) will tend to be neutralized: a quiet girl will be interpreted as an entity in a sub-class of girls. Property predication does not necessarily imply the use of an adjective or a noun however. In the following examples it is expressed by the semantic properties of the verb only. (18) ...he drank like a fish: this was confirmed by the bartender. (J. Le Carré, The Spy who came in from the Cold, p. 24) (19) Though my mother still knits white socks for me ... and I am a child, I love and I hate. (V. Woolf, The Waves, p. 12) There is an obvious difference between the two examples since (18) has a counterpart with a be construction that has similar meaning: he was a drunkard and (19) does not. Our concern here will be with the class of verbs that has a counterpart. This class is not homogeneous. When verbs that can be used transitively refer to an activity that involves an external manifestation in such a way that they characterize a class of human animates, the counterpart will be nominal: He teaches → he is a teacher He smokes → he is a smoker He writes → he is a writer When they can be used transitively but refer to an inner disposition rather than an external manifestation, the counterpart will be adjectival and generic meaning will have to be indicated by a direct object in the verbal construction, and by an indirect object in the adjectival structure: He criticizes everyone / everything → he is critical of everyone/ everything He resents everyone → he is resentful of everyone With verbs that can only be used intransitively, generic meaning will have to be expressed in the verbal construction by an adverbial modifier:

168

JACQUELINE GUILLEMIN-FLESCHER He adapts easily → he is adaptable He communicates easily → he is communicative He talks a lot → he is talkative

The constraints are somewhat more complex than this; for instance the counterpart can entail a change of meaning: cf. he cooks / he is a cook; some verbs are compatible with both adjectival and nominal counterparts:8 he works hard → he is hard-working / he is a hard-worker. We shall not examine the constraints in detail, however, since our objective is to compare the verbal predication with the adjectival / nominal counterpart when both are available. The two constructions are often presented as equivalent. But the fact that only verbs with classifying properties have a nominal counterpart tends to indicate that there is a difference. A verbal predication is not stabilized into a defining property.9 It is fundamentally compatible - with qualification and quantification: He smokes a pipe He smoked three cigarettes - with present or past tense: He smokes He smoked The nominal construction on the other hand definies one entity in a class by a defining property. The difference between the verbal and the adjectival construction is less obvious; one may indeed question the legitimacy of a different analysis for the two structures. Translations however reveal that, when both can be used, a marked preference is shown in English for the nominal and adjectival structures, in French for the verbal structure. This again pleads in favour of establishing a distinction. As already mentionned verb + Ø expresses the notion of predication; be + adjective however implies that the property is stabilized. This modifies the status of the subject in relation to the predicate. In the adjectival construction, the active role of the subject is completely neutralized. In the verbal construction, an ambiguity remains. When no object is specified, the qualitative value of the verb is necessarily underlined, but the active role of the subject is not completely cancelled out.

QUALIFICATION AND POINT OF VIEW 3.

169

Subjective Assertion

As we have seen, stating the existence of a quality does not involve modal evaluation. No specific point of view is singled out. The relation between subject and predicate is regarded as objectively valid. This implies that it can be verified from any point of view and that it is consequently unquestionable. Subjective assertion, on the other hand, implies that the predication is presented from the point of view of the speaker only. This type of predication will have specific constraints that are particularly complex. The first fact that has to be noted is that it is not restricted to modal predicates, that is, to predicates whose semantic properties are associated with subjective evaluation. A distinction has to be made between three types of subjective predication: a) Qualification of a subject He is rich b) Qualification that involves intensity of degree He is so rich c) Predication restricted to evaluating degree How rich he is ! 3.1 Qualification of a Subject Our previous examples exemplified a tendency in English to state the existence of a property by identifying the subject to an occurrence of a class, when in French one simply qualified the subject. This distinction does not, however, hold systematically. The adjectival construction can also be used in English but under specific conditions, whereas in French these constraints do not apply. Let us compare the nominal structure in examples (5), (10) and (1) with adjectival “equivalents”. (5)

a. Lord Edgware ... was a tall man of about fifty. Just for a moment Liz thought it was Ashe. He was tall and fair and wore one of those raincoats with leather buttons. (J. Le Carré, The Spy who came in from the Cold, p. 160)

170

JACQUELINE GUILLEMIN-FLESCHER (10) a. I am not a nihilist. I am a humanist. How long’s he been talking like this ? Always ? Or is it only recent? He has always been idealistic. That is his nature. (J. Le Carré, The Russia House, p. 200) (1)

a. Carlotta Adams was an American girl... Leamas guessed he was Russian, but he wasn’t sure. (J. Le Carré, The Spy who came in from the Cold, p. 67)

D. Bolinger mentions the fact that with Human Subjects one will more readily use the structure: “He is an American” than “He’s American” except when the adjective has the semantic feature of “typical of such people”. He gives the following example: Why does he behave that way? — Because he’s (so) Irish.10 We shall see later, that this semantic feature does indeed, in some cases, account for the restriction. As our examples show, however, it is not a necessary condition for predicative use of the adjective, whether it concerns nationality or other properties. The difference between the two patterns (indefinite article + noun / zero marker + adjective) seems to be based on a broader distinction. When the semantic features of an adjective are more readily associated with objective criteria, they can nevertheless be used subjectively to qualify a Human Subject and more particularly under the following conditions: a) in a complement clause introduced by a verb of cognition, perception, supposition, assertion: think, suppose, remember, believe, say etc. as in (1a) b) in a clause that refers back to a previous statement as in (10a) c) in a dialogue situation, when an inter-subjective modality is involved as in: He is American, you know Example (5a) is a little more complex. The clause Liz thought it was Ashe, in the previous sentence, implies that the predication He was tall and fair ... is related to a specific point of view. Private verbs and verbs of assertion indicate that the predication is neither independent nor objective. An objective predication will be expressed in a main clause, that is independent both of previous context, and of a specified source of utterance. In a complement clause such as Leamas guessed he was Russian, “he was Russian” (zero marker + adjective) is no longer presented as an

QUALIFICATION AND POINT OF VIEW

171

unquestionable fact. Even if the main verb were said, the predication would still be related to the source of assertion Leamas and would therefore be presented as a relative assertion.11 A predication that refers back to a previous statement also has modal implications. It will necessarily qualify the first predication or an element in the first predication and can therefore no longer be a pure statement of fact.12 In our example He is American, you know, the inter subjective relation marked by you know, precludes objective statement. If He + be American is brought to the knowledge of the hearer, it implies that it is not viewed as an established fact. It is simply asserted by the speaker. 3.2 Assertion with Intensive Evaluation 3.2.1 “Modal” Adjectives13 We noted earlier that adjectives whose semantic features imply subjective evaluation can nevertheless be used in an objective statement, providing they are included in a nominal structure such as: he is a stupid man. They will naturally be particularly compatible with subjective assertion, however, and therefore with an adjectival structure: He’s stupid / he’s crazy When used with this type of predication they will be subject to specific constraints. 1) The predicate will necessarily be marked by a rising intonation on the first syllable implying subjective appreciation with intensity of degree. He’s stupid ! He’s crazy ! This is the only binding constraint but an extensive corpus reveals that 2) in up to 80% of the examples, the predicate will also be preceded by a modal adverb marked by an intensive stress. The adverb can express either a) indefinite degree: such as so (20) ‘He’s so proud of that’, the Ladybird said, smiling at James. (R. Dahl, James and the Giant Peach, p. 33)

172

JACQUELINE GUILLEMIN-FLESCHER (21) ‘It ‘s so funny’ she gasped ... (A. Christie, Lord Edgware dies, p. 95)

or

b) conformity with the notion of the predicate. The nature of the conformity will, however, be a vital criterion. It can either (i) relate the specific occurrence of the property to the central, fundamental value of the notion, for instance: (22) She’s really beautiful or

(ii) relate it to the notion in a way that does not single out a central value as in:

(23) She’s just spiteful ... trying to take it out of us both. (A. Christie, Death on the Nile, p. 54) (24) ‘What is his attitude in this ?’ ‘He’s furious - simply furious.’ (Ibid., p. 46) and (25) They must be absolutely mad ! “There’s nothing to be afraid of here !” (Ibid., p. 74) (26) ‘But, my dear, it’s perfectly charming!’ (V. Woolf, A Haunted House, p. 54) In the case of just or simply, the relation is one of strict conformity between the contextual occurrence and the prototypical value. The degree of identification is: “neither more nor less”. In the case of absolutely, perfectly, on the other hand, the relation is one of complete conformity: the specific occurrence of the property is identical in every respect with the concept. In neither case is the property related to the core of the notion, that is, to its central exemplary value, as in she’s really beautiful.14 The unlimited degree of the quality expressed by so in so proud (20) andso funny (21) also relates the occurrence of the property to the central value of the notion. This value can only be constructed, however, in the absence of a clause establishing a limit to the scale of possible degrees: compare

QUALIFICATION AND POINT OF VIEW

173

(27) ‘It was so beautiful last Friday’, she said, ‘that I determined I must go there’. (V. Woolf, A Haunted House, p. 115) and (27') It was so beautiful ! In (27) the clause: that I determined I must go there imposes a limit to the scale of possible degrees: i.e. it was beautiful to the extent that... When no such limit is imposed, the non-finite character of the scale will entail intensive evaluation and lead back to the exemplary value. In other words, when personal assessment is implied by the semantic features of the predicate, the degree of subjectivity in the assertion will have to be marked by the intonation15 and will frequently be further marked by a modal adverb. 3.2.2 “Defining” Adjectives Just as modal adjectives can, under the condition defined, be compatible with a predicative statement, defining adjectives can be used in a subjective predication. But whereas with modal adjectives constraint 1 (rising intonation on the first syllable) is binding and constraint 2 (presence of a modal adverb marked by an intensive stress) is optional, with defining adjectives, constraint 2 is binding and constraint 1 no longer applies. Subjective intonation expressing intensive degree is incompatible with this category of predicates unless it is preceded by a modal adverb.16 If this condition is not observed, an accent marking intensive degree will lead to ungrammaticality. * He’s American * He’s tall * The sky’s blue A further restriction applies, however, concerning the nature of the adverb. Only adverbs expressing indefinite degree like so and adverbs expressing the central value of the notion like really can be used with defining adjectives: He’s so American He’s really American One notes that quite is incompatible with defining adjectives * He’s quite American17 * The sky’s quite blue

174

JACQUELINE GUILLEMIN-FLESCHER

While it can be used with adjectives that are associated with both modal and defining properties, such as tall, handsome, the value constructed will be one of restricted degree and not intensive degree as in He’s quite tall He’s quite handsome Adverbs that do not single out the central value, as in (23), (24), (25), and (26), are also incompatible with defining adjectives * He is simply American * He’s just American * He’s absolutely / positively American18 It follows from these constraints and restrictions that subjective assertion with intensive degree can only appear with defining adjectives when there is an adverbial modifier. Moreover the adverb must express either indefinite degree or the core of the notion. Strict coincidence with the qualitative value of the predicate such as implied by just / simply; or entirety, completion, absence of restriction such as expressed by absolutely / perfectly, do not bring out the qualitative value of the predicate. Indefinite degree on the other hand, implies that no limit can be established in evaluating the degree of the property. Hence the intensive quality. Colour adjectives, like nationality adjectives, will need an adverbial modifier to be compatible with subjective assertion of degree. * The sky is blue The sky is so blue The restrictions concerning the nature of the adverb also apply: the modifier will have to express intensive degree; adverbs expressing strict conformity or complete coincidence with a qualitative value are unacceptable with colour adjectives: * The sky is just / simply blue * The sky is absolutely blue When the colour is inherent to a concept, it will be compatible with gradation, providing the qualitative evaluation is related to a specific situation. This explains why

QUALIFICATION AND POINT OF VIEW

175

The sky is so blue ! The grass is so green ! are acceptable while * Her dress is so blue ! is not. In the last example blue simply characterizes a specific occurrence of the concept: dress. It is not inherent to the concept itself and is therefore incompatible with degree. Two significant differences appear between French and English where subjective assertion of a quality is concerned. Whereas modal adjectives implying extreme degree of a property have to be marked by a stress or a stress + modifier in English, this is not the case in French. Il est fou / il est bête can be marked simply by a rising tone without intensification. The other difference concerns the way in which intensive degree is expressed. With colour adjectives for instance, it will often be expressed by repetition. Sa robe était rouge, mais rouge ! The colour red is in the first instance predicated of the term her dress, and is subsequently identified with the exemplary value: that is, to the colour red par excellence. The second term will necessarily receive emphatic stress. 3.2.3 Adjectives Implying Change of State Predicates expressing a property resulting from change of state are interesting in relation to subjective predication. Two types of predicates correspond to this category: - the first involves two states that are mutually exclusive: alive / dead single / married open / closed The change from one to the other coincides with a lexical change; alive --> dead, and results in complete stabilization. It therefore precludes progression beyond the point that differentiates the two states.

176

JACQUELINE GUILLEMIN-FLESCHER This type of predicate will generally be compatible a) with adverbs that express quantification of the process: The door is almost closed

It is the degree of completion of the process that is then evaluated. One notes that quantitative evaluation is possible only if it indicates that the point separating the two states has not yet been reached: He is practically dead / married He is practically married implies he is still single; the change from one state to another has not yet taken place. Once the limit between the two has been reached the resulting state cannot be quantified. Hence the impossibility of he is half-dead / he is half-married, except with a metaphorical value: i.e. he is completely exhausted / he has all but taken the official step. Only the first of the two examples, however, corresponds to a common idiom. b) with adverbs that appear to evaluate degree of completion such as She is dead, very dead Elle est morte, bien morte He is well and truly married Il est bel et bien marié But the problem is in fact a little more complex. The first of the English examples is obviously very marked stylistically and therefore marginal. It is nevertheless an authentic example.19 It is possible in English because very can function modally whereas the French translation * Elle est morte, très morte would be unacceptable. It would not be considered marginal but ungrammatical. Très unlike bien is in fact incompatible with subjective appreciation. Both very and bien are possible only because the property morte / dead has already been predicated of the subject. They function as a modal reassertion of the property. The second category of predicates also expresses resultant state, but without lexical change.20 The accomplishment of the process that leads to a resultant state:

QUALIFICATION AND POINT OF VIEW

177

a) does not coincide with a specific point in time b) does not imply stabilization of the process and hence c) does not involve a lexical change ; for instance with predicates such as - tanned - exhausted - embittered the resultant state of the process (be tanned / exhausted / embittered) is open to progression He is more and more tanned / exhausted / embittered Contrary to the first category of predicates that is compatible with qualification and hence with a subjective point of view, only when reassertion of the property is expressed, the second category is compatible with qualitative grading without restriction and hence with subjective predication: He’s so tanned / He’s so embittered / I’m so exhausted This modal difference is directly linked to the aspectual characteristics that distinguish the two categories. A resultant state will be compatible with subjective appreciation only if the process can be viewed as a continuum without a definite point that separates previous state from resultant state and without an end point. When the resultant state of a process: e.g. die → be dead, implies a break with a previous state: i.e. alive, then both qualification and subjective predication are possible only when they refer back to a previous property predication. When there is no definite dividing line between two states, then subjective qualification is possible without restriction. He’s so tanned He’s completely exhausted / He’s absolutely exhausted Once more indefiniteness, qualification and subjectiveness are closely linked. Subjective evaluation of a property that results from a process implies the possibility of assigning a degree. Degree cannot however be assigned to a stable property since it is not subject to variation. Moreover intensive qualification requires that the degree should be indefinite i.e. that there should be no reference to an end point that establishes a limit to the scale.

178 4.

JACQUELINE GUILLEMIN-FLESCHER Exclamation

4.1 Subjective Assertion and Exclamation Whereas subjective assertion can express a quality with or whitout intensive degree, exclamatory predications referring to qualitative degree are centred on modal evaluation of the predicate. The frontal position of - the subject in the first case He is young and - the indefinite adverb + predicate in the second How young he is ! indicate a thematic difference. Since one can only evaluate a quality providing its existence is established, exclamations necessarily relate to a previous predication. This predication will either be explicit in previous discourse or inferred from the situational context. Even when it is not explicit, it will necessarily have a theoretic status that must be taken into account. The main differences between the two types of predication are therefore that a)

in subjective predication both the attribution of the property and its modal assessment can be expressed in the same predication. In exclamations the existence of the property is necessarily pre-established and the exclamatory predication is restricted to quality evaluation.

b)

In subjective predication there can either be - qualification of the subject He is young - or qualification of the subject with intensive degree He is so young Exclamations are compatible with intensive degree only How young he is !

QUALIFICATION AND POINT OF VIEW

179

4.2 Constraints on Exclamations It has been pointed out by R. Baker (1970), J. Bresnan (1970) and N. Chomsky (1973) that there is a morphological identity between interrogative pronouns and complementizers. J.C. Milner modifies the model in order to integrate pronouns21 that are typically used in exclamations. The characteristic that explains this common use can easily be explained. In each case, the pronoun covers an open class of entities or degrees without a specific unity or point of reference in a scale being singled out. (28) What reason should she have for lying about it ? Pour quel motif nous aurait-elle menti ? (A. Christie, Lord Edgware Dies, p.107/125) (29) It was a younger Maurice ... , wearing the ... smile of a man ... who hasn’t had time to decide what expression to assume. C’était un Maurice plus jeune, ... qui arborait le sourire ... d’un homme ... qui n’a pas eu le temps de décider quelle expression adopter. (P.D. James, Innocent Blood, p. 31/ 49) (30) What nonsense! Quelle sottise! (R. Dahl, Matilda, p. 146/146) The same analysis can be applied to adverbs such as how as the following examples show: (31) How was she dressed ? Comment était-elle habillée ? (A. Christie, Lord Edgware Dies, p.73/83) (32) I don’t know how my mother did it Je ne sais pas comment ma mère arrivait à se débrouiller (R. Dahl, Boy, p. 54/65) (33) How delightful it sounds... ... avec quelle joie je vous écoute! (A. Christie, Destination Unknown, p. 125/149) In the exclamation (ex. 33) intensive degree is marked by the adverb How. It

180

JACQUELINE GUILLEMIN-FLESCHER

represents a scale of possible degrees of the property delightful. Whereas in a question, a specific degree is expected to be assigned by the addressee, in an exclamation no such specification is either expected or possible. It is precisely the absence of a specific reference point that implies an unlimited range of conceivable degrees; hence the intensive evaluation and the subjective viewpoint.22 This explains why a question concerning a quality must refer to a classifying property and not to a modal one. As has already been shown, the distinction between the two is not clear cut, but adjectives whose semantic properties imply marked subjective evaluation are generally incompatible with questions: * Is he infuriating ? except with a negative marker and a suspended intonation Isn’t he infuriating ? It will then be equivalent to an intensive assertion and the speaker simply submits the predication to the addressee for his assentment. The constraint governing adjectives also holds with verbs. But whereas adjectives will be incompatible with an exclamation if they refer to a classifying property, verbs will be excluded - if they are used transitively. - if they express a process that is viewed aspectually as accomplished. The relation between classifying properties and transitivity is not immediately obvious. If transitivity is considered as a process engendered by a subject or source and producing an effect on the object rather than as a syntactic category of verbs,23 the analogy becomes clear. If the object is modified by the process, it implies - that the process is viewed quantitatively, - that the result will be definable by external criteria. In other words, the process is viewed as reaching an end point and its properties are transferred to the object: He opened the door → the door is open Since the qualitative value compatible with exclamations requires that no

QUALIFICATION AND POINT OF VIEW

181

specific reference point be established, it is incompatible with a transitive relation. A classifying property can also be defined objectively. Neither, therefore, complies with the qualitative constraint. If we compare the following examples: (34) How the children stare ! (V. Woolf, A Haunted House, p. 20) (35) How she laughed ! (V. Woolf, Mrs Dalloway, p. 37) (36) How you hate her ! (V. Woolf, A Haunted House, p. 20) (37) How I love my Rob ! (V. Woolf, Mrs Dalloway, p. 67) with (38) *How she cut her finger ! (39) *How the dragonfly fell into the pond ! it becomes apparent that the presence of an object does not necessarily entail quantification of the process.24 In (34) and (35), the process is necessarily viewed qualitatively since there is no object. In (36) and (37) the process does not imply modification of the object. It is not therefore viewed as having an end point. In (38) and (39), however, the process entails a stable result. Since it does not comply with the qualitative constraint, it is compatible neither with a subjective viewpoint nor with an exclamation. 4.3 Degree Evaluation and Exclamation The pronouns we have just examined and the evaluation of intensive degree which they represent are usually considered to be characteristic of exclamations. Two questions remain to be answered, however, -

are they a necessary condition of exclamatory predication ? are predications that foreground evaluation of degree necessarily exclamatory ?

To the first question we answer no. Exclamations can express qualitative evaluation without these pronouns, for instance:

182

JACQUELINE GUILLEMIN-FLESCHER (40) ‘Nonsense !’ said Alice... (L. Carroll, Alice in Wonderland, p. 202) (41) ‘Impossible !’ said Mrs Bucket. (R. Dahl, Charlie and the Great Glass Elevator, p. 100) (42) ‘Idiot !’ said the Queen... (L. Carroll, Alice in Wonderland, p. 200)

The qualitative evaluation is expressed by a modal noun in (40) and (42) and by a modal adjective in (41). The difference between these exclamations and those that include a pro-form such as what / how is that subjective assessment of the previous predication is here expressed simply by a noun or adjective denoting a quality assigned to that predication or to the subject of the predication without scaling possible degrees.25 In other words a quality is assigned, but without evaluation of degree. Our last example (43) ‘Are you mad !’ said Mr Wonka. (R. Dahl, Charlie and the Great Glass Elevator, p. 128) is characterised by syntactic inversion. Except for the exclamation mark, it is syntactically identical with a “yes/no” question. Since a straightforward yes/ no question can only imply a value judgment with adjectives that have a double status (i.e. both modal and classifying) such as intelligent, interesting, beautiful, when the adjective does not fall into this category, the question will be interpreted as an exclamation. Our second objection is that exclamations are not necessarily concerned with qualification. The following examples will generally be recognized as exclamatory, yet they neither assign a quality nor evaluate the degree of a quality: (44) ‘There she is !’ said Mr Wonka... (Ibid., p. 114) (45) Off with her head ! (L. Carroll, Alice in Wonderland, p. 200) (46) ... who cares ! (R. Dahl, James and the Giant Peach, p. 40) -

Example (44) is a locative predication with thematic fronting of the adverb

QUALIFICATION AND POINT OF VIEW -

183

(45) is an exclamatory injunction, with thematic fronting of the adverbial particle and Ø verb.

With the exception of the exclamation mark, (46) has the characteristics of a rhetorical question (interrogative pronoun + modal verb). Since a question concerning a modal property cannot be answered, no specific value can be assigned to the wh-question and it is therefore interpreted as a negative assertion.26 We come now to the second point. In example (33) we noted that intensive evaluation and high degree were directly linked. Moreover, the exclamatory nature of this example implies that it is highly subjective. One question remains to be answered however: “Are predications that assign high degree to a property necessarily exclamatory ?” A specific point that is relevant to this problem has already been the object of theoretical debates, namely the status of “indirect exclamations”. J. C. Milner (1978) considers them as exclamations in their own right. Our own feeling is that exclamations in the strictest sense cannot be indirect. In the following example: (47) I know how well she played. (V. Woolf, A Haunted House, p. 31) the predication I know X implies that a limit is imposed on the degree of the quality. The assessment is therefore quantitative rather than intensive. As we mentioned before, intensive degree requires the scale of possible degrees to be infinite. It must be an open class with no end point. By imposing an end point to the scale, it seems to us that the predication I know X precludes the complement clause from exclamatory status. Let us now quote two other examples that do not qualify for exclamatory status, at least as we would define it: namely, the predication must be related to a specific situation, which can either be the immediate situational context or a previous explicit assertion. (48) How lovely goodness is in those who, stepping lightly, go smiling through the world ! (V. Woolf, A Haunted House, p. 30) (49) How shocking, and yet how wonderful it was to discover that these real things, Sunday luncheons, Sunday walks, country houses, and tablecloths were not entirely real... (Ibid., p. 47)

184

JACQUELINE GUILLEMIN-FLESCHER

Evaluation of degree is expressed in both examples How lovely / how shocking. Neither of these predications, however, refers to a previous explicit assertion or to a situational context. The intensive quality is related in (48) to a sub-class of human subjects: those who, stepping lightly, go smiling through the world, in (49) to a predication that is not strictly related to a specific situation. This brings us to a fundamental issue: the reason for determining how degree predications and exclamations are related. We have shown to what extent they can be subject to different constraints. They have, however, one common feature that is essential to subjective predication. They are directly related either to one element of a situation or to a whole situation in a previous context. The context can be determined explicitly or by a situational context. Even when not expressed linguistically, it has a theoretic status, and is an essential condition of these two types of predication. 4.4 Ø Verb Exclamations When an adjectival exclamation relates to one element in a context, this element will necessarily be explicit and entail the use of a verb. (50) How beautiful the sound is ! (V. Woolf, A Haunted House, p. 24) (51) ‘How unscrupulous you are, Mother !’ (A. Christie, Death on the Nile, p. 71) With adjectival exclamations relating to a whole situation, this will no longer be the case, since one does not need to identify the element that is qualified. (52) ‘Miss Adams died in her sleep last night....’ ‘How awful ø!’ (A. Christie, Lord Edgware Dies, p. 81) (53) ‘I, too, am going to make it my business to find out.’ ‘Are you ? How funny ø !’ (Ibid., p. 91) With nominal exclamations, providing the noun does not refer to modal properties, the element or situation is identified by the noun, but there is often a choice between a be predication and a zero predication:

QUALIFICATION AND POINT OF VIEW

185

(54) ‘What a splendid sight ø !’ said Mr Wonka. (R. Dahl, Charlie and the Great Glass Elevator, p. 76) (55) ‘What a great sight it is !’ (Ibid., p. 76) When a choice exists, the degree of subjectivity will be more marked with Ø verb, because the quality will be directly related to the previous context, and not indirectly via the anaphoric pronoun. This raises the question of the theoretical status of nominal phrases. Why should a non-verbal phrase be more subjective than the “equivalent” construction with an explicit verb? The fact that it is linked to direct speech and to previous or situational context, implies that the relation established between two terms is determined hic and nunc and is therefore (a) motivated by a specific situational context (b) established by the speaker in relation to that context Consequently it cannot be considered as a factual statement based on external criteria that have general validity. It is simply the speaker’s point of view as it appears to him at the moment of speaking. The paradox lies in the fact that the relation to the speaker is unmarked. The relation established with the immediate context, however, precludes an objective point of view.

5.

Conclusion

As we suggested in our introduction, each type of predication that has here been examined involves an interplay of three grammatical categories; modality, determination and aspect. Objective statement entails quantification, adjectives referring to classifying properties, aspect involving an end point. Subjective assertion and exclamation are linked to qualification, to adjectives expressing modal properties, and to process that does not involve a permanent change of state. Transitivity is characterictic of objective predication, intransitiveness of subjective predication. As we have seen, the modal or defining properties associated with nouns, adjectives, and verbs are not necessarily inherent semantic properties. They can be determined contextually and are in many cases compatible with both types of predication. Exclamations and, more generally, intensive degree are compatible with quantification only in so

186

JACQUELINE GUILLEMIN-FLESCHER

far as it is indefinite. Scaling degrees in a non-finite class implies subjective assessment and therefore leads back from quantity to quality. Both prosody and thematic emphasis will vary according to the point of view. The differences that appear between French and English do not invalidate these constraints. They are linked to other factors. Our hypothesis is that in English, relations tend to be determined in view of the re-construction of referential values by the hearer or reader, in French the point of reference is essentially the speaker and consequently the construction of referential values. Hence the need in English to state the existence of a quality whereas in French the existence of a quality can be pre-constructed.

Notes 1.

This is valid also for other linguistic phenomena. S. Kuno (1976) makes an explicit statement to this effect. See also S. Kuno (1987) and A. Kamio (1994).

2.

By modal implications we here mean involving a subjective point of view: for instance, the example given in our outline of the theoretical framework at the beginning of this paper: David is so French! implies that the speaker is not viewing the properties expressed by the adjective French as an objective property but as a set of characteristics that he evaluates as being typical of the French nation.

3.

Except with the meaning: She was a big girl now.

4.

This distinction can also be observed when the subject does not refer to a Human Animate: It’s almost a disease / C’est presque pathologique (Film on television) C’est délicat... / It’s a tricky business... (Le Monde April 4th 1992 / Guardian Weekly April 11th 1992)

5.

henceforward: property predication.

6.

A singular noun without an article is generally considered to have adjectival status: Cf. R.L. Wagner and J. Pinchon (1962) and B. Pottier (1962). As L. Kupferman (1979) points out, however, a predicative adjective can be preceded by a modifier such as très. It can only modify a noun when it does not have literal meaning, for example: Elle est très comédienne / Il est très artiste. The decisive factor, therefore, seems to be the distinction between a noun referring to a defining property (indefinite article + noun) and a noun referring to a qualifying property (zero + noun) and not between nominal or adjectival status.

7.

By Human Subject we mean a term referring to a Human Subject. The formulation is simplified for stylistic reasons.

8.

Although the adjectival counterpart necessarily entails the use of the copula be, the terms adjectival counterpart / adjectival structure, are used to differrentiate be + adjective (e.g.

QUALIFICATION AND POINT OF VIEW

187

be talkative) from a purely verbal construction (e.g. he drinks). 9.

By stabilized we mean: that is not subject to alteration, i.e. that is valid without restriction. For instance He gambles, expresses an activity that is typical of the subject he but cannot be considered as a stable defining property in the same sense as he is a gambler. Whereas he gambles can be restricted by specific conditions: He gambles whenever the fancy takes him, He is a gambler is not subject to change: *He is a gambler whenever the fancy takes him (cf. J. Guillemin-Flescher, 1997).

10.

D. Bolinger (1967: 33).

11.

Cf. on the subject S. Kuno (1987) and A. Kamio (1994). Both authors discuss the problem of perceptual / cognitive strategies and the relevance of this criterion to subordination / indirect discourse. See also, J. Guillemin-Flescher (1984a) and (1984b). It should be noted that the distinction between factive and non-factive verbs (cf. J. B. Hooper and S. A. Thompson 1973) does not apply to this particular constraint: *Leamas said/knew he was a Russian.

12.

The fact that the subject is a pronoun does not prevent the predication from being independent, providing it is not a deictic pronoun (cf. p.163).

13.

The inverted commas indicate that the modal properties are not necessarily pre-determined.

14.

The central exemplary value of the notion is not necessarily definable by an objective property or set of properties. It is a purely conceptual value that is viewed as an archetype. While the concept is shared by all, the nature of the properties associated with it is subject to variation. For a detailed discussion, cf. A. Culioli (1990), especially pp. 69-70. See also R. Needham (1972).

15.

Although our examples are mostly taken from written texts, prosody is such a fundamental criterion in subjective predication that it seemed legitimate to include this criterion throughout this discussion.

16.

Except when rising intonation expresses a quality that does not conform with expectations, for instance: he’s late / he’s deaf / he’s American. The quality can then either refer to an accidental property (late) or to a permanent characteristic (American), but in both cases, the prosody will simply reflect the discrepancy between the subject’s expectations and the quality revealed in a particular situation.

17.

Native speakers asked whether they accepted ‘he’s quite American’ systematically felt the need to add something: He’s quite American really He’s quite American in his tastes The restrictive value is then constructed, as for adjectives associated with both modal and defining properties (tall / handsome) see above.

18.

He’s just / simply American could possibly be acceptable with an adequate context, providing there is no stress on just or simply. It would, however, be difficult to imagine a context that would be compatible with He’s absolutely / positively American.

19.

I owe this example to C. Charreyre, University of Paris VII

188

JACQUELINE GUILLEMIN-FLESCHER

20.

M. Riegel (1985) makes a similar distinction, though he formulates it in different terms.

21.

The author, however, refers to pronouns that are common to interrogative clauses, relative clauses and exclamations. If the model is to be extended to exclamations, the analogy with relative clauses raises a problem: what and how are compatible with questions, exclamations and complement clauses, but not with relative clauses except with “free relatives” such as ‘what I have told you is strictly confidential’ (R. Dahl, Matilda, p. 32); the analogy with relative clauses applies to questions. Which, on the other hand, is compatible with questions, relative clauses and complement clauses but not with exclamations. Who and qui are compatible with rhetorical questions but not with exclamations assigning intensive degree.

22.

For a more detailed study of exclamations see A. Culioli (1974), (1990), and in a different theoretical perspective J.C. Milner (1978); see also M.P. Chatras (1984) for a contrastive approach.

23.

For a similar view cf. P. Hopper and S. A. Thompson (1980). Our view is, however, more radical; for instance the authors compare the two examples Jerry likes beer and Jerry knocked Sam down (Hopper and Thompson 1980: 253) and note that the second example is higher in Transitivity than the first. Since the process like does not establish a transfer from one participant to another, we do not consider a) as simply lower in Transitivity than b) but as intransitive. This and the fact that beer is less defined semantically than Jerry would explain why b) is compatible with a passive construction and a) is not: Sam was knocked down by Jerry / *Beer was liked by Jerry.

24.

By quantification of the process, we mean a process that is viewed as a specific occurrence: He drank a glass of whisky as opposed to the qualitative value: He drank (like a fish).

25.

It may be noted that when the quality is assigned to an addresse the qualitative assessment will be an insult rather than an exclamation.

26.

Cf. J. Guillemin-Flescher (1995): “Questions rhétoriques et évaluation modale”. Other types of exclamation can be quoted; for instance the following patterns: Ce courage! / Cette chance qu’elle a!

References Anderson, Stephen R. and Paul Kiparsky (eds) 1973 A festschrift for Morris Halle. Cambridge, MA: MIT Press. Baker, Carl R. 1970 “Notes on the description of English questions”. Foundations of Language 6: 197-219. Benveniste, Emile 1966 Problèmes de linguistique générale, Vol.1, Paris: Gallimard. Bolinger, Dwight 1967 “Adjectives in English : Attribution and predication”. Lingua 18: 1-34.

QUALIFICATION AND POINT OF VIEW

189

Bouscaren, Janine, Jean-Jacques Franckel, and Stephane Robert (eds) 1995 Langues et langages: problèmes et raisonnements en linguistique, Mélanges offerts à Antoine Culioli, Paris: Presses Universitaires de France. Bresnan, Joan 1970 “On complementizers”. Foundations of Language 6: 297-321. Chatras, Marie Pascale 1984 “L’exclamation et la prédication de propriété”. Cahiers Charles V N° 6, Université Paris VII: 83-109. Chomsky, Noam 1973 “Conditions on transformations”. In S. R. Anderson and P. Kiparsky (eds), 232-286. Culioli, Antoine 1974 “A propos des énoncés exclamatifs”. Langue Française N° 22, Paris, Larousse: 6-15. Culioli, Antoine 1990 Pour une linguistique de l’énonciation. Paris : Ophrys. Culioli, Antoine 1995 Cognition and representation in linguistic theory. Amsterdam and Philadelphia: John Benjamins, translated by Michel Hiddle with the assistance of John T. Stonham. Groussier, Marie-Line and Claude Rivière (eds) 1996 La notion: Actes du colloque tenu les 2 et 3 février 96, Université Paris VII,. Guillemin-Flescher, Jacqueline 1981 Syntaxe comparée du français et de l’anglais : problèmes de traduction, Paris: Ophrys. Guillemin-Flescher, Jacqueline 1984a “Enonciation, perception et traduction”. Langages N° 73, Paris, Larousse: 74-97. Guillemin-Flescher, Jacqueline 1984b “Traduire l’inattestable”. Cahiers Charles V N° 6, Université Paris VII: 127-146. Guillemin-Flescher, Jacqueline 1995 “Questions rhétoriques et évaluation modale”. In J. Bouscaren, J-J. Franckel, and S. Robert (eds), 435-457. . Guillemin-Flescher, Jacqueline 1997 “De la qualité à la qualité”. In M-L. Groussier and C. Rivière (eds), 146-155. Hjelmslev, Louis 1948 Essais linguistiques. Paris : Editions de minuit (1971, 1988). Hooper, Joan and Sandra A. Thompson 1973 “On the applicability of root transformations”. Linguistic Inquiry 4: 465497. Hopper, Paul and Sandra A. Thompson 1980 “Transitivity in grammar and discourse”. Language 56: 251-299.

190 Kamio, Akio 1994

JACQUELINE GUILLEMIN-FLESCHER

“The theory of territory of information : The case of Japanese”. Journal of Pragmatics 21: 67-100.

Kuno, Susumu 1973 The structure of the Japanese language. Cambridge, MA: M.I.T. Press. Kuno, Susumu 1976 “Subject, theme and the speaker’s empathy: A re-examination of relativization phenomena”. In C. N. Li (ed), 417-444. Kuno, Susumu 1980 “Discourse deletion”. In S. Kuno (ed), 1-144. Kuno, Susumu 1987 Functional syntax: Anaphora, discourse and empathy. Chicago: The University of Chicago Press. Kuno, Susumu (ed) 1980 Harvard Studies in Syntax and Semantics, vol.III. Cambridge, MA: Department of Linguistics, Harvard University. Kupferman, Lucien 1979 “Les constructions il est médecin / c’est un médecin : Essai de solution”. Cahier de linguistique, Montreal, Presses de l’Université de Québec N°9: 131-164. Kuroda, Shigeyuki 1968 “English relativization and certain related problems”. Language 44: 244266. Li, Charles N. 1976 Subject and topic. New York: Academic Press. Milner, Jean C. 1978 De la syntaxe à l’interprétation. Paris : Editions du Seuil. Needham, Rodney 1972 Belief, language and experience. Oxford: Blackwell. Pottier, Bernard 1962 “L’absence d’article en français et sa motivation”. Revue de langues romanes XXVI: 158-162. Riegel, Martin 1985 L’adjectif attribut. Paris: Presses Universitaires de France. Wagner, Rene-Louis and Pinchon Jacqueline 1962 Grammaire du français classique et moderne. Paris: Hachette.

Corpus: (When two dates are given, the first refers to the date of publication, the second to the edition used) Carroll, Lewis Alice’s adventures in wonderland. Bilingual edition, Paris: Aubier Flammarion, 1970.

QUALIFICATION AND POINT OF VIEW

191

Christie, Agatha Death on the Nile, London: Collins (Fontana), 1937, 1985. Lord Edgware Dies, London: Collins (Fontana), 1933, 1990. French translation by L. Postif, Paris: Librairie des Champs-Elysées, 1933, 1991. Destination Unknown, London: Collins (Fontana), 1954, 1958. French translation by M. Le Houbie, Paris: Librairie des Champs-Elysées, 1968, 1991. Dahl, Roald. Boy, Harmondsworth: Penguin Books, (Puffin Books), 1984, 1986. French translation by J. Hérisson, Paris: Gallimard (folio junior), 1985. Charlie and the Great Glass Elevator, Harmondsworth: Penguin Books, (Puffin Books), 1973, 1978. James and the Giant Peach, Harmondsworth: Penguin Books, 1973, 1974. Matilda, Harmondsworth: Penguin Books, (Puffin Books), 1988, 1989. French translation by H. Robillot, Paris: Gallimard (folio junior), 1988. Fowles, John The Ebony Tower, London: Pan Books, 1974, 1986. French translation by A. Saumont, Paris: Albin Michel, 1978. James, P.D. Innocent Blood, Harmondsworth: Penguin Books, 1980, 1989. French translation by L. Rosenbaum, Paris: Fayard Mazarine, 1984. Le Carre, John The Russia House, London: Coronet Books, 1989, 1991. French translation by M. Perrin et I. Perrin, Paris: Gallimard, 1989. The Spy Who Came In from the Cold, New-York: Bantam Book, 1964, 1990. French translation by M. Duhamel and H. Robillot, Paris: Gallimard, 1963, 1964. Levi-Strauss Claude Les structures élémentaires de la parenté , Paris: Mouton, 1967, 1981. English translation by J.Harle Bell, J.R. von Sturmer, and R. Needham, Boston: Beacon Press, 1969. Woolf, Virginia A Haunted House, Harmondsworth: Penguin Books, 1944, 1975. Mrs Dalloway, Harmondsworth: Penguin Books, 1964, 1976. The Waves, Harmondsworth: Penguin Books, 1931, 1973.

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 193

II. OTHER FUNCTIONAL ASPECTS OF LANGUAGE

194

NORIKO AKATSUKA

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 195

Towards a Theory of Desirability in Conditional Reasoning Noriko Akatsuka University of California, Los Angeles

1.

Introduction

I have been working on conditionals for over fifteen years.1 At the abstract level, I could say that I chose conditionals because I was interested in the following questions: 1. What is linguistic pragmatics? 2. Is semantics separable from pragmatics? 3. Why and how is language a reflection of the human mind? Though conditionals themselves represent an extremely fascinating research topic, I chose to study them not for their own sake, but instead as a key to finding answers to these fundamental issues in general linguistics. Everyone will agree that the notion of truth has played a central role in the long tradition of the study of semantics, in particular, in the inquiry of conditional reasoning. First and foremost, the “if-then” construction has been regarded as a linguistic device for valid reasoning and logical argumentation. Consequently, the natural language conditional has been typically compared to the mathematical conditional, “ p ⊃ q”, where the relevant notions are truth-values, i.e., True and False. Until quite recently, much research on conditionals in linguistics as well as in such related fields as psychology and language acquisition has been conducted under the strong influence of that rationalistic tradition (cf. Traugott, ter Meulen, Reilly and Ferguson 1986, Haiman 1978). As for counterfactual conditionals, their study has been virtually the monopoly of logicians and philosophers (cf. Jackson 1991), and the

196

NORIKO AKATSUKA

major concern of these disciplines has been the logical problem of truth conditions, i.e., to specify under what conditions a particular instance of counterfactual reasoning is ‘true’ or ‘valid’. Through years of research, however, I have become more and more convinced that the speaker’s evaluative stance, i.e., DESIRABLE/UNDESIRABLE plays a crucial role in understanding many instances of conditional reasoning in everyday life. Take, for example, the familiar adult speech to young children such as, “if you touch it, you’ll get burned.” In a recent crosslinguistic study (Clancy, Akatsuka, and Strauss; 1997), we examined how Japanese, Korean, and American parents use conditionals when talking to very young children of less than three years of age. Based on a total of 84 hours of spontaneous discourse data, we discovered that what we called “predictive” conditionals such as the above was one of the three most common semantic types of conditionals in all three languages. According to our analysis, “predictive” conditionals express the speaker’s prediction, “UNDESIRABLE-leads-to-UNDESIRABLE” and the speaker’s stance, “THAT’S UNDESIRABLE, SO I DON’T WANT IT TO HAPPEN.” Since the types of functions served by “predictive” conditionals, e.g., warnings, threats, and prohibitions, exemplify the earliest type of discourse to which we humans are exposed in the process of socialization and language acquisition, it is reasonable to assume that the natural logic of UNDESIRABLE-leads-to-UNDESIRABLE in conditional reasoning is universally acquired by children of any language and any cultural background at a very early age. The major goal of the present study is to share with the reader for the first time a full picture of my theory of desirability in conditional reasoning. I have had the great fortune of having access to the data collected in Japanese, Korean, and English by Iwasaki, Kawanishi, and Strauss, respectively, just following the 1994 Northridge earthquake near Los Angeles. In this study, I hope to show how and to what degree a qualitative analysis of cross-linguistic data can shed light on the understanding of the inherent relationship between the speaker and counterfactuality in everyday life. The paper will consist of two major divisions, Section 2 and Section 3. Section 2 is a discussion on what originally motivated my interest on conditionals in general. It will provide a brief overview of the literature and philosophy on conditionals which impacted my thinking at the time. I would like the reader to understand how and why I have become an advocate of the “desirability hypothesis” in conditional reasoning, and what exactly this

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 197 hypothesis is. Section 3 is an examination of counterfactual reasoning in everyday life. It will be divided into two smaller sections, 3.1 and 3.2. In 3.1, I will introduce Fauconnier’s (1985) theory of mental spaces as a more current view on counterfactual conditional reasoning and will systematically demonstrate why and how it fails to account for usages of counterfactuals in everyday situations. In 3.2, I will examine instances of counterfactual conditionals excerpted from the aforementioned 1994 LA earthquake data and will contrast these with Fauconnier’s made-up examples. A working hypothesis will then be presented to answer questions such as “When and why do we typically use counterfactuals?” or “What are we doing when we invoke counterfactuals into daily conversation?”

2.

Background

2.1 The First Encounter I first became interested in conditionals in about 1975, when I was starting out as an Assistant Professor at the Department of Linguistics at the University of Chicago and at the same time audited a course in mathematical linguistics. Many linguists at that time were using symbolic logic to formally represent linguistic semantics — by “formally” I mean in the sense of clearly and concisely. In that class, I learned that the analysis of conditionals was traditionally regarded as being within the territory of mathematical logicians and philosophers rather than that of linguists. I also learned that many brilliant people had been trying to come up with some analysis of the logical meaning of conditionals for a long time, and yet there was (and even still now is) very little agreement about the underlying semantics of conditionals. As a result of that experience, I started to think that in Japanese, two- and three- year olds can understand and use conditional forms such as Mama, itcha iya ‘Mommy, don’t go’ (literally ‘if you go, I’ll hate it’) and puzzled over the following question: Since young children understand and use these types of utterances right away, how could conditionals be as complicated as they were made out to be? I gradually started to think that perhaps the first premise of many scholars is wrong — the premise that people try to represent the meaning of natural language conditionals using mathematical logic. In the domain of mathematics, there is no ‘speaker’, no ‘listener’, and no evolution

198

NORIKO AKATSUKA

of time. The notion of true and false in mathematical logic does not need the existence of people. If 2 + 2 is 4, 2 + 2 will always equal 4 even after every human being disappears from earth. It is an impersonal, eternal truth. I began to think then that perhaps conditionals belong to an entirely different domain — the domain where what counts is the speaker, the listener, and the progression of time. Of course, initially this was only my gut feeling and for a long time I didn’t know how to go about substantiating this new premise. I then realized that the examples in logic textbooks are all in English, and that I should look at my own language. For example, in modern Japanese, the antecedent clause in konya shujin ga kaette ki-tara tazunemashoo ‘If/When my husband comes back tonight, I’ll ask him’ can express either ‘if’ or ‘when’. In order to accurately determine whether the meaning is a conditional or a temporal one, one must consider what the speaker is thinking about with respect to her husband coming home, that is, whether it involves a certainty or an uncertainty. If it is just a regular business day, he would be expected to come back at the end of the day. But if he is out of town on business and the wife does not know whether he will be coming tonight or tomorrow, the reading would be ‘if’. I also learned at that time that the Korean conditional marker -myen behaves in exactly the same way as -tara, particularly with respect to the possible alternation in meaning between ‘if’ and ‘when’. 2.2 Benveniste (1971) Around the same time that I began to get seriously interested in conditionals, I first read Benveniste’s (1971: 223) famous article, “Subjectivity in Language”, which starts with the question, “If LANGUAGE is, as they say, the instrument of communication, to what does it owe this property?” I was quite moved when he questioned the fact that people have equated language to a ‘communicative tool’, and then asked the question that if language is simply a ‘tool’ how do we explain where all of these properties come from? If language is a tool or an instrument, you could throw it away and replace it with a new one. However, there is no way to separate language from human beings. In Benveniste’s words, “Language is in the nature of man, and he did not fabricate it” (pp. 223-224). He also said that the real nature of language only shows up in discourse — “Many notions in linguistics, perhaps even in psychology, will appear in a different light if one reestablishes them within

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 199 the framework of discourse” (p. 230). Most importantly, Beneveniste characterized the speaker as being the ‘subject’ of consciousness. This type of perspective on the relationship between language and the speaker made me begin to understand how modern Japanese and Korean conditionals work, for example. That is, one basically needs to look into the consciousness of the speaker to see whether the utterance is a conditional or a temporal. One must know whether the speaker considers the issue at hand as being within the realm of certainty or uncertainty at the very moment of utterance. 2.3 The Initial Puzzle Even though the natural language conditional has typically been compared with the mathematical conditional, there are actually very few scholars who subscribe to the view that the natural language “if” behaves exactly like“⊃”. One well-known exception is, of course, Grice (1975). Recall that Grice’s famous maxims of conversation and his concept of conversational implicatures were originally conceived to defend his theoretical position that the natural language connectives such as “if”, “and”, and “not” are semantically equal to their counterparts in classical logic. Around the time that I seriously started to look at the natural language conditional, Grice’s ideas became available through publication (1975) for the first time and quickly gained popularity among linguists. As a result, there were a good number of “orthodox” Griceans — certainly more than there are now. The following is one of the puzzles that I faced at the early stages of my work on conditionals (e.g., 1983). Refer to the classical truth table represented in Table 1. If the value of both p and q are (T), then the evaluation of the whole conditional statement is also (T) as is evident in the first line. Similarly, if the value of both p and q are (F), the evaluation of the whole is (T), as is evident in the last line: Table 1. Classical Truth Table p

q

p⊃q

True True False False

True False True False

True False True True

200

NORIKO AKATSUKA

Focusing on the last line, one will note that a particular type of puzzle emerges — a puzzle which involves a contrast of two types of conditionals. For the first type, imagine a situation where an American mother is crying at her daughter’s funeral, saying: “If I hadn’t given her the car keys, this accident wouldn’t have happened.” If we adopt the truth table analysis, the antecedent, ‘if I hadn’t given her the car keys’ is F(alse), since the mother had given her daughter the keys, and the consequent clause, ‘this accident wouldn’t have happened’ is also F, since it actually did happen; thus the whole utterance is evaluated as T. This corresponds to the fourth line of the classical truth table (i.e., F F = T). Now compare this sentence with the following conditional: “If this sandwich was made this morning, you’re Shirley Temple,” excerpted from the movie “Gaslight”. At the time that this movie was made, Shirley Temple was one of the most popular child stars in America. One crucial contextual element here is that this conditional is uttered by an angry male customer to a middle-aged, unattractive woman working behind the counter of a diner. What the speaker is saying is that ‘this sandwich can’t possibly have been made this morning’ — it was dry and tasted bad, and he was unhappy about it. The truth-value of the first clause is F, and of course the interlocutor is not Shirley Temple, so that is F also. With the value of both p and q being F, the value of the whole conditional statement is evaluated as T (i.e., F F = T). Now let’s compare the first conditional with this one. Both conditionals are of the type F F = T, but the first one truly expresses the speaker’s heartfelt sorrow and regret, while the second one expresses the speaker’s cynical and nasty attitude towards his interlocutor, implying something like “I don’t believe such nonsense, you old hag”. The mental attitudes of the two speakers in each conditional are totally different. Given this, what does the logical analysis of F F = T say about the inherent difference of emotion associated with the two conditionals? Absolutely nothing. Furthermore, the truth-value approach fails to tell us anything about why the first conditional can begin a conversation, while the second one cannot. The “if” clause of the second type of conditional must always be a repetition of what the previous speaker has just asserted; it cannot originate in the speaker’s own mind. The formula F F = T clearly does not tell us anything at all about these differences. This is precisely what inspired me to choose conditionals as my focus of study. There seemed to be an entire domain of interaction and emotion that simply could not be accounted for, or even noticed, by the

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 201 traditional philosophical approach — and this domain is the domain of desirability.

3.

The Desirability Hypothesis

In recent years, I have been suggesting that natural language conditionals are an important device for encoding the speaker’s evaluative stance of desirability (e.g., Akatsuka, 1997; Clancy, Akatsuka, and Strauss, 1997; Akatsuka and Sohn 1994, Mays 1994). Table 2 below juxtaposes my desirability table with the traditional truth table: Table 2. Akatsuka Desirability Table vs. Classical Truth Table DESIRABILITY TABLE (Akatsuka)

TRUTH TABLE (classical logic)

p

q

if p then q

p

q

p⊃q

Desirable _______ _______ Undesirable

Desirable _______ _______ Undesirable

Desirable _______ _______ Undesirable

True True False False

True False True False

True False True True

I first arrived at the desirability table hypothesis when I realized that a statement such as the following can function as either a “promise” or “threat”: “If you eat my cookies, I’ll whip you” (Akatsuka 1991). For most of us who consider being whipped as undesirable, a natural interpretation of this statement would be a threat. However, it also could be a genuine promise if the speaker knows that the listener enjoys being whipped. Thus, in the first reading, the speaker wants the addressee to eat the cookies, while in the second, s/he does not. Clearly, based on this type of reasoning, speakers are using a contingency relationship between the antecedent and the consequent of natural language conditionals to express their desirability stance, as follows: DESIRABLE leads to DESIRABLE or UNDESIRABLE leads to UNDESIRABLE

202

NORIKO AKATSUKA

I believe that the natural logic working in our everyday reasoning is extremely simple, and that in many uses of conditionals, the logic is actually “Desirable leads to Desirable” and “Undesirable leads to Undesirable”. The first type of conditional expresses the speaker’s attitude “THAT’S DESIRABLE, SO I WANT IT TO HAPPEN.” Conversely, “Undesirable leads to Undesirable” conditionals express the speaker’s attitude “THAT’S UNDESIRABLE, SO I DON’T WANT IT TO HAPPEN.” The most dramatic, empirical support for the desirability table hypothesis has come from my recent joint research with Pat Clancy, of UCSB, and Susan Strauss, of UCLA, involving the types of conditionals used by parents to their children (Clancy, Akatsuka, and Strauss 1997). In this study we investigated how Japanese, Korean, and American parents use conditionals when talking to very young children and analyzed the types of conditionals the parents use. Examples of the type of questions we asked ourselves are: What kind of input using conditionals are Japanese, Korean, and American children of one-, two, and three- years of age exposed to? Could it really be the case that the conditional sentences children are hearing do, in fact, relate to issues of true vs. false, or is it more an issue of DESIRABLE (i.e., desirable-leads-todesirable) or UNDESIRABLE (i.e., undesirable-leads-to-undesirable)? Since the children in all three language samples were very young, these conversations were particularly rich in such speech acts as commands, prohibitions, and granting permission. In order to analyze desirability in the datasets of the three languages in an objective fashion, we first identified the frequently recurring types of events, actions, and states that can be reliably judged as desirable or undesirable on independent grounds. These objective criteria were refined and collapsed into six categories: safety, health, social acceptability, success, emotional wellbeing, and positive evaluation. On the basis of these criteria, we coded the content of every conditional clause in our data as “desirable”, “undesirable,” or “neutral.” Table 3 shows that the majority of conditionals in our data in all three languages (59% in Japanese, 71% in Korean, and 58% in English) follow the logic of Desirable-leads-to-Desirable or Undesirable-leads-toUndesirable. Only 4% of the conditionals in Korean and English, and 11% in Japanese reverse the expected relationship between desirable events/desirable outcomes and undesirable events/undesirable outcomes; and of these anomalous conditionals, 43% are not actually counterexamples to this natural logic, since they

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 203 are concessives. Concessive conditionals, by definition, exhibit a counter-toexpectation relationship between antecedent and consequent, as in the following examples: (1)

Japanese: (Child wants to sit on mother’s lap to eat) M: dakko shinaku-temo taberu desho ‘even if I don’t hold you, you can eat, can’t you?’

(2)

Korean:

(Child turns on TV and tells everyone to look; her mother tells her that the program she’s interested in isn’t on) M: pwa to maynnal kuke eps-canha ‘even if we look, that’s never on.’

In concessive conditionals, undesirable antecedents (not sitting on mother’s lap) leads to desirable consequences (being able to eat) while desirable antecedents (watching TV) lead to undesirable consequents (not seeing the desired program). Since it is the inherent nature of concessives to reverse the usual desirability relations found in conditionals, they can be seen as the exception that proves the rule. Table 3. Desirability Analysis of All Conditional Sentences (Clancy, Akatsuka, and Strauss 1997) Antecedent

Consequent

Japanese

Korean

English

desirable undesirable desirable undesirable

desirable undesirable undesirable desirable

(N=149) .25 .34 .01 .10 .30

(N=131) .18 .53 .02 .02 .25

(N=129) .29 .29 .008 .03 .38

neutral

3.

Counterfactuals and Desirability

Recall the counterfactual conditional expressing the mother’s grief that we discussed in Section 2.3. We can now understand that the antecedent clause, ‘If I hadn’t given her the car keys’ is not a matter of true vs. false, but rather giving her the car keys is either a ‘desirable’ action or an ‘undesirable’ action.

204

NORIKO AKATSUKA

Giving her the car keys turned out to be an undesirable thing; the desirable situation would have been not giving them to her. The counterfactual conditional, “If I hadn’t given her the car keys, this accident wouldn’t have happened’ uttered by a lamenting mother in any language expresses the logic of “desirable-leads-to-desirable” and the speaker’s attitude, “That’s desirable, so I want it to happen”. Why, then, does it express the mother’s deep sorrow and regret? It is because she KNOWS that what she wants is absolutely unattainable.2 In 1985, I claimed that conditionals are identifiable not by their syntactic forms, but by the speaker’s attitude at any given point within the epistemic scale in Table 4. Here it is shown that the speaker’s evaluation of the realizability of p ranges in value from zero (i.e., counterfactuals) to one (i.e., realis). Table 4. The epistemic scale (Akatsuka 1985) REALIS

IRREALIS 1

know (exist x)

0 get to know (exist x) newly-learned information

not know (exist x)

know not (exist x) counterfactuality

3.1 Critique of Fauconnier’s (1985) Account of Counterfactuality In the recent cognitive semantics movement, Fauconnier (1985) has attempted to provide an account of counterfactuality in the framework of his theory of mental spaces. I will demonstrate that Fauconnier’s vision of counterfactuality is inadequate in shedding light on everyday counterfactual reasoning, essentially because this perspective remains under the strong influence of the tradition of logic. First, consider example (3), appearing in the original as (F52). Fauconnier presents this example as an everyday occurrence of a counterfactual context. (3)

[originally appearing as (F52)] A1: If Napoleon had been Alexander’s son, he would have won the battle of Waterloo. B1: But he would have died long before that.

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 205 A2 : Well, suppose he lived a very long life, without ever aging, or that Alexander was resurrected in Corsica in the eighteenth century... Now, let me ask the readers to re-read this made-up conversation very carefully and see whether they can come up with any clear idea of what speaker A is saying by uttering (A1). Is the speaker saying that Napoleon was not as great a military leader as Alexander? If this is the case, why doesn’t he say so more directly, i.e., “If Napoleon had been Alexander” rather than Alexander’s son? And what about speaker B’s reaction to (A1)? Is (B1) a joke? And if it is a joke, why is speaker A proposing such ‘absurd’ conditions as an immunity to aging and the possibility of resurrection for a mortal man in (A2)? It would appear here that A is trying to make (A1) an instance of ‘valid’ reasoning. Does this mean, then, that by uttering (A1) speaker A is more interested in the imaginary father-son relationship between Alexander and Napoleon than in Napoleon’s military prowess itself? I must say that I am totally puzzled and unable to deduce the speakers’ intentions in any of the lines in the above excerpt. One could probably claim that it is primarily due to the lack of prior discourse context that we find it difficult to comprehend the speakers’ intentions in (3). However, I would strongly disagree with that line of thinking. Instead, I would plainly argue that contrary to Fauconnier’s claim, (3) cannot possibly be a reflection of ordinary usages of counterfactuals. I doubt that any conversation even remotely resembling this one would ever occur in spontaneous discourse; and regarding the ability to deduce the speaker’s intention in using counterfactual conditionals in spontaneous discourse, I will argue that the user’s intention is, on the contrary, normally quite obvious, not just to the interlocutor, but to the general audience as well. This point will be demonstrated in detail in the examination of authentic discourse data such as our 1994 LA earthquake data. Crucially, just like logicians and philosophers’ analyses of counterfactuals, Fauconnier’s theory is not at all interested in the question “when a speaker uses a counterfactual conditional, what is s/he really doing?” This point becomes much clearer when we examine Fauconnier’s concept of ‘spacebuilders’. According to Fauconnier, various linguistic elements such as “if” and “wish” as well as “not” and “prevent” are considered to be “spacebuilders”. As such, they are supposed to be capable of setting up a counterfactual mental space which functions in such a way that “some relation (sic)”,

206

NORIKO AKATSUKA

which does not hold in the parent space (i.e., the speaker’s belief world), may be satisfied in the counterfactual space. In this light, consider the utterances listed in (4), which appear in the original as (F3-F6) (emphasis original). (4)

[appearing in the original as F3, F4, F5, and F6] (F3) If Lucky had won, I would be rich. I would have moved to Tahiti. (F4) I wish Lucky had won. I would be rich. (F5) Fortunately, the fire did not cross the highway. My house would have been destroyed. (F6) Luckily, the fire was prevented from crossing the highway. My house would have been destroyed.

Note that the speakers of (F3)-(F4) sound unhappy because what they wanted to happen did not happen. In contrast, the speakers of (F5) - (F6) sound extremely happy. They are virtually saying, “Boy! Am I fortunate/lucky! My house wasn’t destroyed!” However, in Fauconnier’s theory of spaces, the speaker’s mental attitude is virtually of no relevance whatsoever. Fauconnier simply states that in all four of the above utterances, “the first sentence can be understood to set up a counterfactual space (incompatible with the origin), and the second one expresses some relation satisfied in that counterfactual space (pp. 109-110).” Fauconnier totally ignores the significance of the attitudinal adverbs “fortunately” and “luckily” in the first sentences in (F5) and (F6) and claims that counterfactuality in the second sentence is lexically imposed by the words “not” and “prevent” because these words are strong space-builders, much stronger than “if” and “wish” in (F3) and (F4) respectively. Fauconnier does, however, recognize the fact that “if” and “wish” do not always function as successful space-builders, as illustrated in (5), originally appearing as (F10): (5)

[originally appeared as (F10)] I wish you would help me tomorrow. (Compatible with your actually helping)

In sharp contrast, however, when we examine the 1994 LA earthquake

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 207 data in the next section, it will become crystal-clear that it is not the lexical power of negatives such as “not” and “prevent” that has ‘imposed’ the counterfactuality in (F5) and (F6); rather, as will be demonstrated, it is the speaker who has intentionally invoked counterfactuals into the discourse for specific purposes, e.g., complaining or rejoicing. To illustrate this point in preview, notice that the omission of the attitudinal adverbs from Fauconnier’s examples result in an unnatural occurrence of counterfactuals in the second sentence, as illustrated in (6) below: (6)

a. [from (F5)] ??The fire did not cross the highway. My house would have been destroyed b. [from (F6)] ?? The fire was prevented from crossing the highway. My house would have been destroyed.

While it is clear that the notion of desirability plays absolutely no role in Fauconnier’s theoretical framework, we have seen that Fauconnier’s own examples unwittingly serve as testimony to the critical importance of the consideration of the speaker’s attitude to an adequate theory of counterfactual reasoning in everyday life. 3.2 A Case Study from the 1994 Los Angeles Earthquake Data 3.2.1 Background On the morning of Monday, January 17, 1994, an unusually strong earthquake hit the Los Angeles area at approximately 4:30 a.m. It was a holiday, not a regular working day. When the earthquake hit, most people were at home, still fast asleep. Thanks to this, the casualties were miraculously small. About five weeks after the earthquake, audio-and video-taped narratives about the earthquake experiences were collected from UCLA students who are native speakers of Japanese, Korean, and American English. These data were collected by Iwasaki (Japanese), Kawanishi (Korean) and Strauss (English). The students were total strangers until they met for the first time at the data collection site. The students were paired up and were asked to talk to each other for approximately 20-25 minutes about their personal experiences during the earthquake. The student pairs were left alone in the room which was equipped with a video camera and supplemental audio recording device. For the purpose of this study, I will restrict the analysis to one particular excerpt from the Japanese conversational data.

208

NORIKO AKATSUKA

3.2.2 Co-construction of Counterfactuals in Japanese Data In the following, I will focus on one Japanese excerpt which followed a 15 minute exchange of the two speakers’ respective experiences about things such as where they were at the time of earthquake, how frightening it was, and what kind of damage they suffered (relatively small). This excerpt is shown in (7): The structure of (7) is clearly divided into 3 parts: (a) the opening, (b) coconstruction of counterfactuals ‘if not p, not q’ by the two speakers, where p and q stand for REALITY (e.g., p = “the earthquake occurred at 4:30 a.m.” and q = “casualties were very small”), and (c) the closing. Note that the speakers are saying essentially the same thing in the opening and closing. In the opening, speaker H says “Asa de yokatta (We were lucky it was in the morning )” and in the closing, speaker A repeats exactly the same thing, “Asa de yokatta”. (7)

a. Opening H:

demo nee asa de yokatta desu yo ne? gozen yoji han de? A: atashi mo omotta-n desu yo.

b. “if not p, not q”

H: A: H: A: H: A: H: A: H: A: H: A: H: A: H:

c. Closing A:

moshi nee kore ga nee imagoro = kore ga nee, yuugata no: ima toka ne rasshuji no goji toka ne ha ha ha ha soo soo soo yuugata no rasshuji no goji toka: ha haha haha ha ano toshokan nanka de mada gakusei nanka ir < iru jikan dattara: iru san yoji toka: moo atama kara saa hon ga ochitekite: aa zaa sugokatta desu nee. haiuei furiiuei nanka mo zenbu nee zenbu sundanshichattara moo kanbotsushichatte: shisha nanka . juubai gurai n natta deshoo nee?

nn atashi mo omotta-n su. asa de yokattn -n daroo na tte H: nnn

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 209 A: H: A: H:

minna neteru toki datta kara honto hukoo chuu no saiwai desu yo: nn =nn nn

[English version of (7)] a. Opening H:

but, see, we were lucky it happened in the morning, at four thirty, right? A: I thought about that, too.

b. “if not p, H: not q” A: H: A: H: A: H: A: H: A: H: A: H: A: H: c. Closing A: H: A: H: A: H:

If it had happened around now (in the late afternoon) (it) had happened in the evening now or something at rush hour, like five o’clock ha ha ha ha right, right, right. rush hour — 5:00 in the evening ha haha haha ha well like at the library, if it was the time the students were still there like three or four o’clock or well, on their heads, the books would fall, and Oh, like “crash” — it would’ve been terrible. the highway, freeway, all of them. All of them If they had been cut then they would have collapsed and The death toll would’ve been 10 times more, right? So, I thought too that we were lucky it happened in the morning. Yeah. everybody was sleeping at that time, so Right, it was the only blessing within this disaster. Yeah. =Yeah, yeah.

210

NORIKO AKATSUKA

What is so remarkable about these Japanese data is that its abstract structure is quite similar to that of Fauconnier’s examples in (4), (particularly (F5) and (F6)), even though Fauconnier’s examples are monologic and (7) is a dialogue. One important difference is that in the latter there is no occurrence whatsoever of the words “not” or “prevent”. In this light, I would like to propose here that contrary to Fauconnier’s claim, what ‘imposes’ a long stretch of co-construction of counterfactuals by the two speakers is the initial exchange of “Asa de yokatta desu yo ne? (Weren’t we lucky it happened in the morning!)” and “atashi mo omotta n desu yo (I thought about it, too!)”. This is precisely the role of the sentence initial English attitudinal adverbs, “fortunately” and “luckily” in (F5) and (F6). What are the two speakers doing by co-constructing “if not p, then not q” here? Notice that all of the co-constructed counterfactuals reflect the logic of Undesirable-leads-to-Undesirable. In sharp contrast to Fauconnier’s ‘absurd’ conditions, all co-constructed “if not p” components here represent a variety of situations which could have actually happened and could have resulted in much greater casualties. The two speakers have shared an unusually undesirable experience, but they could come out of it relatively unharmed, and they are congratulating themselves by conveying the message “it could’ve been much worse.” Recently, Strauss and Kawanishi (1996) have used the 1994 LA earthquake data for their pioneering study on assessment strategies in Japanese, Korean and American English. One of their findings is that Japanese speakers often participate in co-construction activities to establish common ground, contrary to the claim made by others, e.g., Ono and Yoshida (1996), that speakers of Japanese do not ‘finish each other’s sentences’. In fact, example (7) used here is excerpted from one of Strauss and Kawanishi’s examples illustrating collaborative finishes in Japanese. What deserves our special attention is that counterfactuals are used here to establish emotional solidarity between the two speakers. Throughout the 1994 LA earthquake data, Undesirable-leads-to-Undesirable counterfactuals typically appear regardless of which language the speakers use. Based on this, I hypothesize that the following is a proto-typical usage of counterfactual conditionals in everyday life. Ultimately, this hypothesis will have to be examined in the light of cognitive psychology. Hypothesis: When we humans undergo some unusual experiences, we tend to reappraise/appreciate REALITY in terms of desirability in comparison/contrast with hypothetical situations which actually could have happened.

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 211 This unusual experience does not have to be something undesirable. It can be an extremely pleasant one, such as the successful completion of a difficult task. As such, it is easy to find counterfactuals expressing the author’s gratitude in the acknowledgments sections of many different types of writings.3 Example (8) is a clear example taken from a popular paperback on the Vietnam War: (8)

Dear America could not have been completed without their dedication, insights, and commitment to the intrinsic value of these writings. (Edelman, ed. Dear America: Letters Home From Vietnam. 1985)

This type of counterfactual also expresses the logic of Undesirable-leads -to-Undesirable. The conditional as a whole hypothetically expresses the worst situation that the speaker can think of, namely, the non-existence of the very object which is the source of the present happiness. For this reason, the counterfactual can express a strong sense of gratitude by conveying the message that “you and you alone are responsible for bringing about the happy PRESENT moment.”

4.

Conclusion

I have tried to demonstrate how discourse-based research on conditionals has culminated in the construction of a theory of desirability in conditional reasoning. By doing so, I hope that I have also shown that fundamentally our study addresses itself to general questions in linguistic theory: 1. What is linguistic pragmatics? 2. Is semantics separate from pragmatics? 3. Why and how is language a reflection of the human mind? The research results overwhelmingly indicate that the most typical counterfactual conditionals occurring in everyday life are direct reflections of the deeply felt emotions of ordinary people. This is a concept that appears to be totally ignored by the rationalistic tradition of counterfactuals. I conclude by suggesting that an adequate theory of cognitive semantics will have to incorporate in some way a theory of desirability as delineated in the present study.

212

NORIKO AKATSUKA

Notes 1.

My special thanks go to Pat Clancy, Shoichi Iwasaki, and Susan Strauss for their valuable suggestions and criticisms at various developmental stages of this paper. In addition to the 6th J/K Conference at the University of Hawaii, earlier versions of this paper were presented at the following Japanese institutions: International Christian University, Tokyo; Kobe University; Nagoya University; Tohoku University; and Tsukuba University. This study was partially supported by a UCLA Academic Senate Research Grant and a Sasakawa Faculty Research Grant from the Center for Japanese Studies at UCLA.

2.

The reader is referred to my discussion of the epistemic scale and counterfactuality in Akatsuka (1985).

3.

“Gratitude counterfactuals” also appear often in the acknowledgement sections of MA theses and Ph.D. dissertaions.

References Akatsuka, Noriko 1983 “Conditionals”. Papers in Japanese Linguistics 9: 1-33. Akatsuka, Noriko 1985 “Conditionals and epistemic scale”. Language 61: 625-639. Akatsuka, Noriko 1991 “Dracula conditionals and discourse”. In C. Georgopoulos and R. Ishihara (eds), 25-37. 1994 “Negative conditionality, subjectification, and conditional reasoning”. In R. Dirven and A. Anthansiadou (eds). Akatsuka, Noriko (ed) 1994 Japanese/Korean Linguistics 4. Stanford: CSLI. Akatsuka, Noriko and Sung-Ock S. Sohn 1994 “Negative conditionality: The case of Japanese -tewa and Korean -taka”. In N. Akatsuka (ed), 203-220. Akatsuka, Noriko, Shoichi Iwasaki and Susan Strauss (eds) 1996 Japanese/Korean linguistics 5. Stanford: CSLI. Benveniste, Emile 1971 Problems in general linguistics. Coral Gables: University of Miami Press. Clancy, Patricia M., Noriko Akatsuka and Susan Strauss 1997 “Deontic modality and conditionality in adult-child discourse: A crosslinguistic study”. In A. Kamio (ed), 19-57. Cole, Peter and Jerry M. Morgan (eds) 1975 Syntax and semantics 3: Speech acts. New York: Academic Press. Dirven, Rene and Athanasiadou Anthansiadou (eds) 1994 On conditionals again. Amsterdam: John Benjamins.

TOWARDS A THEORY OF DESIRABILITY IN CONDITIONAL REASONING 213 Fauconnier, Giles 1985 Mental spaces. Cambridge, MA: MIT Press. Georgopoulos, Carol and Roberta Ishihara (eds) 1991 Interdisciplinary approaches to language: Essays in honor of S.-Y. Kuroda. Dordrecht/Boston/London: Kluwer Academic Publishers. Grice, H. Paul 1975 “Logic and conversation”. In P. Cole and J. L. Morgan (eds), 41-58. Haiman, John 1978 “Conditionals are topics”. Language 54: 564-89. Jackson, Frank (ed) 1991 Conditionals: Oxford readings in philosophy. New York:Oxford University Press. Kamio, Akio (ed) 1997 Directions in functional linguistics. Amsterdam: John Benjamins. Mayes, Patricia 1994 “Conditionals and the logic of desirability: An interview with Noriko Akatsuka”. In S. Strauss and B. Kreuter (eds), 449-461. Ono, Tsuyoshi and Eri Yoshida 1996 “A study of co-construction: We don’t finish each other’s sentences”. In N. Akatsuka, Iwasaki, and S. Strauss (eds), 115-130. Strauss, Susan and Yumiko Kawanishi 1996 “Assessment strategies in Japanese, Korean, and American English”. In N. Akatsuka, S. Iwasaki, and S. Strauss (eds), 149-166. Strauss, Susan and Betsy Kreuter (eds) 1994 Issues in applied linguistics 5. Department of Applied Linguistics, University of California, Los Angeles. Traugott, Elizabeth, Alice ter Meulen, Judy S. Reilly, and Charles A. Ferguson (eds) 1986 On conditionals. Cambridge: Cambridge University Press.

Texts Edelman, Bernard 1985 Dear America: Letters home from Vietnam. Pocket Books. Iwasaki, Shoichi 1994 Northridge Earthquake data.

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

215

Relative Tense and Absolute Tense in Adverbial Adjuncts in Japanese* Kiri Lee Lehigh University

1.

Introduction

In this study, I will investigate the behavior of Japanese tense morphemes in adverbial adjuncts, especially those which express temporal sequence, in relation to the lexical interpretation of adverbials and the matrix tense. Adverbial adjuncts discussed in this study are called “temporal connectives” following Hornstein (1990), since they are the type of adverbial which specifies the temporal sequence between a subordinate clause and a matrix clause. I will pay particular attention to the adjuncts headed with maeni ‘before’, atode ‘after’, and toki ‘when’. The sentences dealt with in this study are expressed in the following structures: (1)

[[[s ...Tense 2...] adjunct maeni/atode/toki] ...Tense 1...]

The contrast between the tense in maeni ‘before’/atode ‘after’ clauses and toki ‘when’ clauses is interesting, because in the maeni ‘before’ clauses, Tense 2 must obligatorily be non-past tense, and in the atode ‘after’clauses, Tense 2 must obligatorily be past tense. On the other hand, in the toki ‘when’ clause, Tense 2 can be either non-past tense or past tense, and that choice results in the different temporal interpretations. First let us observe sentences

* This paper is based on sections (primarily Chapter 3) of my Ph.D. dissertation (Lee 1993). I would like to thank Susumu Kuno, Akio Kamio and two anonymous reviewers for their helpful comments.

216

KIRI LEE

with the temporal connectives maeni and atode: (N.B. n/p and p are the abbreviated notations of non-past and past respectively). (2)

a. John-ga Mary kara tegami-o uketor-ru maeni, denwasu-ta.1 nom from letter-acc receive-n/p before telephone-p “John telephoned Mary before he received a letter from her.” b. John-ga Mary kara tegami-o uketor-ru maeni, nom from a letter-acc receive-n/p before denwasu-ru (daroo). telephone-n/p (probably) “John will telephone Mary before he receives a letter from her.”

(3)

a. John-ga Mary kara tegami-o uketor-ta atode, denwasu-ta. nom from letter-acc receive-p after telephone-p “John telephoned Mary after he received a letter from her.” b. John-ga Mary kara tegami-o uketor-ta atode, nom from a letter-acc receive-p after denwasu-ru (daroo). telephone-n/p (probably) “John will telephone Mary after he receives a letter from her.”

In the sentences in (2), the non-past morpheme ru is used in the subordinate clause with maeni ‘before’, and in the sentences in (3), the past morpheme ta is used in the subordinate clause with atode ‘after’, regardless of the matrix tense. Sentences (2a) and (2b) have past and non-past morphemes, respectively, in their matrix clauses. However, the subordinate clause has a non-past morpheme ru in the maeni clause of each sentence. The parallel phenomenon is observed in (3a) and (3b). In the atode clause, the past morpheme ta is used regardless of the matrix tense. In fact, it is ungrammatical to use the past morpheme ta in maeni clauses and the non-past morpheme ru in atode clauses in Japanese. Compare the sentences in (4) to those in (2), and the sentences in (5) to those in (3): (4)

a. *John-ga Mary kara tegami-o uketor-ta maeni, denwasu-ta. nom from letter-acc receive-p before telephone-p “John telephoned Mary before he received a letter from her.”

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

217

b. *John-ga Mary kara tegami-o uketor-ta maeni, nom from a letter-acc receive-p before denwasu-ru (daroo). telephone-n/p (probably) “John will telephone Mary before he receives a letter from her.” (5)

a. *John-ga Mary kara tegami-o uketor-ru atode, denwasu-ta. nom from letter-acc receive-n/p after telephone-p “John telephoned Mary after he received a letter from her.” b. *John-ga Mary kara tegami-o uketor-ru atode, nom from a letter-acc receive-n/p after denwasu-ru (daroo). telephone-n/p (probably) “John will telephone Mary after he receives a letter from her.”

On the other hand, the sentences with temporal connective toki ‘when’ allow both the non-past tense morpheme ru and the past tense morpheme ta. Observe the following sentences: (6)

a. John-wa nihon-e ik-ru toki, atarasii kamera-o kaw-ta. top Japan-to go-n/p when new camera-acc buy-p “John bought a new camera when he went to Japan.” b. John-wa nihon-e ik-ta toki, atarasii kamera-o kaw-ta. top Japan-to go-p when new camera-acc buy-p “John bought a new camera when he went to Japan.”

The only difference between the sentences in (6) is that the non-past morpheme ru is used in the subordinate clause in (a), and the past morpheme ta is used in the subordinate clause in (b). The English translations do not convey the difference in meaning between the two Japanese sentences. In fact, the English translations are ambiguous between two readings; one interpretation is that John bought a new camera prior to the trip, and the other one that he bought the camera in Japan. The Japanese sentences clarify this difference. Sentence (6a) with the non-past morpheme is construed as John’s purchasing the new camera prior to his trip, while sentence (6b) with the past morpheme is construed as John’s purchasing the new camera in Japan.

218

KIRI LEE

In the following sections, I will present the mechanism behind the fixed tense, namely the non-past ru in the maeni ‘before’ clause and the past ta in the adode ‘after’ clause by schematizing complex sentences of this type. Then I will examine sentences with the toki ‘when’ adjunct, and demonstrate how they allow both tense morphemes in relation to their temporal interpretations.

2.

Framework of This Study

Throughout this study, I will assume Reichenbach’s theory of tense, tense being the grammaticalized expression of point in time. Reichenbach (1947) proposed the theory that tense in natural language has a finite temporal structure. He describes a tense as a set of three points ordered by two relations. These three points are E for ‘the point of the event’, R for ‘the point of reference’, and S for ‘the point of speech’. The two relations are indicated by a comma and a line. The following is a brief description of how Reichenbach’s system works in English. The temporal structures for English simple past, present, and future tenses are given in the following: (7)

a. Simple Past: I saw John. R,E_S b. Simple Present: I see John. S,R,E c. Simple Future: I will see John. S_R,E

In (7a), (7b) and (7c), point R and point E are associated with a comma. This indicates that the point of Reference and the point of the Event are contemporaneous. The point of Reference is the point in time referred to in the sentence, and the point of the Event is the point in time when the event described in the sentence occurred. In (7a), the E and R are at the left side of S on the line. This indicates that R and E are temporally prior to S, the point of Speech, which occurred later. In (7b), S, R, and E are associated with a comma, indicating that the three points are contemporaneous. In (7c), R and E are to the right of S on the line, indicating that they occur in the future, after S. In each of the simple tenses in (7), R and E are contemporaneous, as

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

219

indicated by the comma. But this is not the case for the perfect tenses. Observe the sentences in (8) involving the perfect tense: (8)

a. Past Perfect: I had seen John. E_R_S b. Present Perfect: I have seen John. E_S,R c. Future Perfect: I will have seen John. S_E_R

Sentence (8a) means that the speaker saw John before a certain time in the past, referred to in R. E at the left of R on the line indicates that E occurred prior to R, and R at the left of S on the line indicates that R occurred prior to S. For (8b), the point of Reference is contemporaneous with the point of Speech but temporally future to the point of the Event. Sentence (8c) means that before a certain time in the future, the speaker will see John. This certain time in the future is the point of reference. E is temporally between S and R. The six tenses given in (7) and (8) are the basic tenses in English, and they are schematized in Reichenbach’s theory of tense as shown. I claim, following Reichenbach, that if two sentences have different temporal structures, they have different tenses in our framework. Let us now turn to our discussion of Japanese tense. There are only two tense morphemes in Japanese, i.e., past ta and non-past ru. Japanese does not have a specific tense morpheme to express future tense. Instead, the non-past morpheme can be construed as both present and future. All non-stative verbs in Japanese are temporally interpreted as future relative to the moment of speech when they have the non-past morpheme. The following sentence illustrates this. Observe: (9)

John-wa Tokyo-e ik-ru. top to go-n/p “John will go to Tokyo.”

The verb, ik-ru, in sentence (9) has the non-past morpheme ru and is temporally construed as simple future. The temporal structure for Japanese sentences such as (9) with the non-past morpheme is given in (10). Observe: (10) For (9): Simple Future: S_R,E

220

KIRI LEE

This structure is not suitable, however, for sentences with stative verbs, or non-stative verbs in the habitual reading, even though the syntactic structures are exactly the same as in sentence (9). Observe the following sentences: (11)

a. John-wa mainiti gakkoo-e ik-ru. top everyday school to go-n/p “John goes to school everyday.” b. Bill-wa okane-ga ir-ru. top money need-n/p “Bill needs money.” c. Tukue-no-ueni hon-ga ar-ru. desk-gen- top book-nom exist-n/p “A book exists on the desk.” (=There is a book on the desk.)

For the sentences in (11), a simple present temporal structure should be assigned, as in (12). (12) For (11): Simple present: S,R,E Generally, the unmarked temporal interpretation of non-stative verbs in the non-past form in the Japanese lexicon is that of future relative to the moment of speech, whereas the unmarked temporal interpretation of stative verbs is that of simple present in their non-past form. The habitual interpretation of the non-past morpheme ru is included in the latter class as exhibited in (11a). In addition, as a marked interpretation, stative verbs in the non-past form can have the temporal interpretation of future relative to the moment of speech, when the speaker assumes the event to take place definitely in the future, as observed in the following sentence: (13) Asita kaigi-ga ar-ru tomorrow meeting-nom exist-n/p “There will be a meeting tomorrow.” In Japanese, the present progressive and the perfect of result are expressed in the ‘te’ form of a verb, plus iru. Observe the sentences in (14): (14) a. John-wa tegami-o kak-te top letter-acc write-te “John is writing a letter.”

i-ru. be-n/p

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

221

b. John-wa kokyoo-ni kaer-te i-ru. top hometown return-te be-n/p “John is back in his hometown.” (14a) is present progressive, while (14b) is the perfect of result.2 In (14b), John’s returning to his hometown took place sometime in the past, and the state of being there is still true at the moment of speech. They are, grammatically, indistinguishable from the simple present because the non-past morpheme ru is used in both sentences. This raises a question: can we assign the simple present temporal structure to the sentences in (14)? Hornstein (1990) states that the difference between the simple present and present progressive is based on aspect, not tense. Comrie (1976) defines “aspects” as different ways of viewing the internal temporal constituents of a situation. What sentence (14a) means temporally is that, at the moment of speech, John’s writing of a letter is taking place. There is no reference to the time when this event began or will end. Therefore, the temporal interpretation of this sentence is exactly the same as that of simple present. We can assume that sentences in present progressive have the same temporal structure as simple present. In (14b), the event “John’s returning home” took place sometime in the past and the result of that event is still taking place at the moment of speech, i.e., John is still in his hometown. What we have to determine here is if the difference between the present progressive and the perfect of result is in tense or not. Notice that sentence (14b) does not offer any information about when the event began or will end, just as (14a) did not. What this sentence concerns itself with temporally is that the result of John’s returning is still taking place at the moment of speech. I claim that the difference between (14a) and (14b) is in aspect, rather than tense. Therefore, the correct temporal structure to assign to sentences with the te + i-ru is that of simple present, as shown in the following: (15) For (14): S,R,E Let us now turn to the past morpheme ta. One of the temporal interpretations the past morpheme ta exhibits is the simple past tense, as in example (16). (16) John-wa Boston-e ik-ta top to go-p “John went to Boston.”

222

KIRI LEE

The temporal structure of simple past is assigned to such sentences. Observe (17): (17) For (16): Simple past R,E_S The past morpheme ta is also known to express a perfective interpretation in Japanese. The sentence in (18) can be ambiguous between simple past and perfective. (18) Speaker A: Asagohan-o tabemasi-ta-ka? breakfast-acc eat-p-Q a. “Did you eat breakfast?” b. “Have you eaten breakfast?” How this sentence is construed depends entirely on context. A negative answer to this question clearly shows which interpretation was made of the ta morpheme in this sentence: (19) Speaker Ba: Iie, tabe-masen-desi-ta. no eat-neg-p “No, I did not eat it.” Bb: Iie, tabe-te i-masen. no eat-te be-neg-n/p “No, I haven’t eaten it.” The temporal structure of perfective is more appropriate for the (b) interpretation of (18), which means that a temporal structure of simple past cannot be assigned to sentences such as (18) just by judging from the past morpheme ta in the sentence. Following Reichenbach, I argue that the different interpretations of the basic tense depend on how the three points S, R, and E are linearly ordered and which points are associated with one another. In sentence (18a), the reference point is contemporaneous with E, while in (18b), the reference point is contemporaneous with S. Consider the following temporal structures: (20) For (18a): Simple past R,E_S For (18b): Perfective E_S,R Notice that the temporal structure assigned for (18a) is different from that of (18b). This means that the difference between (18a) and (18b) is one of tense,

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

223

not aspect. Recall the assumption in our framework that if two sentences have different temporal structures, they have different tenses.3 What the past tense morpheme ta does is specify the location of E temporally prior to S, but the location of R is specified by elements such as other lexical items, the context, and the stativity of the verb. For example, stative verbs in the past tense do not have a perfective reading, as illustrated in (21), and therefore R is always simultaneous with E rather than S. (21) John-wa nihongo ga wakar-ta. top Japanese nom understand-p “John understood Japanese.” In sum: the non-past morpheme ru places E temporally future relative to S, or contemporaneous with S. The past morpheme ta places E temporally prior to S. The location of R is determined by other factors such as lexical items, especially time adverbs, the context, and the stativity of verbs. As outlined in this section, the following are the four tenses for Japanese tense morphemes assumed throughout this study. (22) Tense and Temporal structures for Japanese For ru: S,R,E S_R,E For ta: R,E_S E_S,R

3.

simple present simple future simple past perfective

Function of Tense Morphemes in Japanese

There have been several observations made about the behavior of tense in subordinate clauses in Japanese (see e.g. Nakau 1976, Soga 1983, and Machida 1989). Each observer has a different approach to the phenomenon, but in general all have agreed that Japanese tense morphemes have some other function besides locating the event time around the moment of speech. This observation is crucial to understanding, for instance, how the non-past morpheme ru and the past morpheme ta are temporally interpreted in the maeni and atode clauses respectively. First, I will examine sentences with the maeni adjunct. Observe the sentences in (2) again, repeated here:

224

KIRI LEE (2)

a. John-ga Mary kara tegami-o uketor-ru maeni, denwasu-ta. nom from letter-acc receive-n/p before telephone-p “John telephoned Mary before he received a letter from her.” b. John-ga Mary kara tegami-o uketor-ru maeni, nom from a letter-acc receive-n/p before denwasu-ru (daroo). telephone-n/p (probably) “John will telephone Mary before he receives a letter from her.”

What sentence (2a) states is that John telephoned Mary before John’s action of receiving a letter from her took place. If this sentence is true, John’s telephoning Mary must be done temporally prior to John’s receiving a letter. Similarly in (2b), John’s telephoning Mary will take place before the act of John’s receiving a letter even occurs. This means that, for the sentences in (2), at the time when the matrix event takes place, the event in the subordinate clause must not have taken place. This shows that the non-past tense ru is interpreted future to the matrix event time, rather than to the moment of speech. Now let us examine how the past morpheme ta is interpreted in the atode adjunct in the sentences in (3), again repeated here: (3)

a. John-ga Mary kara tegami-o uketor-ta atode, denwasu-ta. nom from letter-acc receive-p after telephone-p “John telephoned Mary after he received a letter from her.” b. John-ga Mary kara tegami-o uketor-ta atode, nom from a letter-acc receive-p after denwasu-ru (daroo). telephone-n/p (probably) “John will telephone Mary after he receives a letter from her.”

What sentence (3a) expresses temporally is that at the time when John telephoned Mary, his receiving a letter must have already occurred. Similarly in (3b), at the time of John’s telephoning Mary, his receiving a letter must have already occurred. Therefore, for the sentences in (3), the event in the subordinate clause must have already taken place at the time the event in the matrix clause takes place. For the past tense ta in (3a) and (3b), it is inter-

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

225

preted prior to the matrix event time, rather than to the moment of speech. Thus, for both sentences (2) and (3), the subordinate clauses are temporally interpreted at the point in time expressed in the matrix clauses, not at the moment of speech. This time sequence can be dealt with by adopting the notions of absolute tense and relative tense. For what the sentences with maeni and atode specify is, after all, the time sequence. Comrie (1985) states the traditional definition of absolute tense and relative tense as follows: (23) Absolute tense includes as part of its meaning the present moment as deictic center; whereas relative tense refers to a tense which does not include as part of its meaning the present moment as deictic center. Recall that in (2a) and (2b), the event in the maeni adjunct must not take place until the event in the matrix clause does. This means, in our tense framework, that the subordinate E must be temporally future relative to the matrix E point. If we adopt the notion of relative tense in our theory, then what is happening here is that the tense in the subordinate clause in (2a) and (2b) takes the matrix E point as deictic center. Therefore, the deictic center for the temporal interpretation of the subordinate clause is the matrix event time, not the matrix speech time. Similarly, in (3a) and (3b), the event in the atode adjunct must take place prior to the event in the matrix clause. In our framework, the subordinate E point must be temporally prior to the matrix E point. Once we assume the notion of relative tense in (23), we can say that the tense in the subordinate clause in (3) takes the matrix E point as deictic center exactly as do the sentences in (2).4

4.

Temporal Structure and Anchor Time

The fact that the subordinate clause is temporally interpreted at the matrix E can be observed more easily if we schematize the sentences in (2) in the form of temporal structures introduced in the previous sections. Recall that in both sentences in (2), the subordinate clause has the non-past morpheme ru, which is obligatorily required by the temporal connective maeni ‘before’. As we demonstrated in the previous sections, the non-past morpheme ru is ambiguous between the present temporal interpretation and the future temporal

226

KIRI LEE

interpretation. Therefore, there are two possible temporal structures we should consider for the subordinate clause in (2); namely, S,R,E, and S_R,E. We will first consider the following two temporal structures for (2a). For convenience we will use E1, R1, and S1 for the three points in the matrix clause, and E2, R2, and S2 for those in the subordinate clause. (24) Temporal structure for (2a): John-ga Mary kara tegami-o uketor-ru maeni, denwasu-ta. “John telephoned Mary before he received a letter from her.” a. S,R,E for the subordinate clause Matrix clause

R1,E1_S1 | Subordinate clause S2,R2,E2 b. S_R,E for the subordinate clause Matrix clause

R1,E1_S1 | Subordinate clause S2_R2,E2

The matrix clause in (2a) has the past morpheme ta, and therefore, the temporal structure of simple past is assigned to the matrix clause in both (24a) and (24b). On the other hand, the subordinate clause has the non-past morpheme ru, which is ambiguous, and therefore, the temporal structure of simple present is assigned to the subordinate clause in (24a), and that of simple future is assigned to the subordinate clause in (24b). In (24a), the vertical line associating E1 and S2 shows that the S2 is anchored to E1, specifying that the matrix E1 is a deictic center for the subordinate clause. S2 is, therefore, contemporaneous with E1. This anchoring also makes E2 to be contemporaneous with E1, because E2 and S2 are contemporaneous in the simple present temporal structure. Thus, (24a) represents the interpretation that John’s receiving a letter and telephoning are simultaneous. This interpretation is not desirable, since the lexical interpretation of maeni ‘before’ requires the sequence of E1 and E2 to be E1 prior to E2. Hence, temporal structure (24a) is ruled out because of the contradiction it contains. On the other hand, in (24b), the line associating E1 and S2 shows that S2 is anchored to E1, having the matrix E1 as a deictic center for the temporal

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

227

interpretation of the subordinate clause. The structure clearly shows that the subordinate E2 point has not yet taken place at the matrix E1 point, since E2 is future relative to S2, which is contemporaneous with E1 by the anchoring line. This meets the lexical interpretation of maeni ‘before’, which obligatorily places E1 prior to E2. Notice also that the temporal relation between E2 and S1 is indeterminate. This allows the potential interpretation in which E2 has not taken place, i.e., John has not yet received a letter.5 (24b) then is the correct temporal structure for sentence (2a). Next, we will consider the following two temporal structures for sentence (2b): (25) Temporal structure for (2b): John-ga Mary kara tegami-o uketor-ru maeni, denwasu-ru (daroo). “John will telephone Mary before he receives a letter from her.” a. S,R,E for the subordinate clause Matrix clause S1_R1,E1 | Subordinate clause S2,R2,E2 b. S_R,E for the subordinate clause Matrix clause S1_R1,E1 | Subordinate clause S2_R2,E2 Both matrix and subordinate clauses have the non-past morpheme ru in (2b). The matrix verb in (2b) is a non-stative verb, and it is interpreted as simple future. Although it is possible to interpret (2b) as habitual, and assign the simple present temporal structure to its matrix clause, we will limit our discussion to the case in which the matrix clause has the simple future temporal interpretation; therefore, the temporal structure of simple future is assigned to it.6 In (25a), the temporal structure of simple present is assigned to the subordinate clause, while in (25b), that of simple future is assigned. In (25a), the subordinate S2 is anchored to the matrix E1, interpreting the subordinate clause at the matrix E1. In this structure, S2 is contemporaneous with E1, and E2 is also contemporaneous with E1. This is not a correct interpretation of (2b), however, because the temporal connective maeni ‘before’ in (2b) obligatorily places E1 prior to E2. Hence, the temporal structure in (25a) is

228

KIRI LEE

ruled out because its temporal relation of E1 and E2 is contradictory to the lexical interpretation of maeni. On the other hand, in (25b), the subordinate S2 is anchored to the matrix E1, specifying that the subordinate E2 is temporally future relative to E1. This meets the lexical interpretation of maeni ‘before’. Hence, the temporal structure (25b) is the correct temporal interpretation for sentence (2b). Notice that, in the subordinate clauses in the temporal structures in (24) and (25), S2 does not stand for the moment of speech, as S1 in the matrix clause does. Rather, S2 is a deictic point anchored to the matrix E1 within the sentence.7 This is possible only when we consider the tense morpheme in the subordinate clause as having relative tense. Recall that relative tense does not include the present moment as deictic center. Therefore, it is possible to anchor the subordinate S to the matrix E. Thus, we redefine the ‘point of speech’ as follows: (26) Anchor Time = Point of Speech by default For a Japanese maeni clause, we can safely assume that the subordinate speech time is obligatorily anchored to the matrix event time, and that this is possible because the tense morphemes in those subordinate clauses do not have absolute tense, but have relative tense. Furthermore, recall that the temporal structures in (24a) and (25a) are ruled out for (2a) and (2b) respectively. This suggests that the temporal structure of simple future, namely, S_R,E, must be assumed for the non-past morpheme ru in the maeni ‘before’ adjunct. This assumption is consistent with the lexical interpretation of maeni ‘before’. Since S2 to E1 anchoring is obligatory in sentences with the maeni ‘before’ adjunct, if we assign the simple present temporal structure S2,R2,E2, then E2 is always contemporaneous with E1 via this anchoring. However, the lexical interpretation of maeni ‘before’ always imposes the temporal relation of E1 prior to E2. Therefore, the simple present temporal structure for the maeni ‘before’ adjunct is automatically ruled out. On the other hand, if we assign the simple future temporal structure S2_R2,E2, then E2 is always future relative to E1 via S2 to E1 anchoring. And this meets the lexical interpretation of maeni ‘before’. Hence, we will assume the simple future temporal structure, namely S_R,E for the non-past morpheme ru in maeni adjuncts. Similar observations can be made for sentences with the atode clause, such as those in (3). The following (27) and (28) are the temporal structures

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

229

for (3a) and (3b), respectively: (27) Temporal structure for (3a): John-ga Mary-kara tegami-o uketor-ta atode, denwasu-ta. “John telephoned Mary after he received a letter from her.” Matrix clause R1,E1_S1 | Subordinate clause R2,E2_S2 (28) Temporal structure for (3b): John-ga Mary-kara tegami-o uketor-ta atode, denwasu-ru. “John will telephone Mary after he receives a letter from her.” Matrix clause S1_R1,E1 | Subordinate clause R2,E2_S2 In both (27) and (28), the temporal structure of simple past is assigned to the subordinate clause because of the past morpheme ta, which is obligatorily required by atode ‘after’. Recall that it was argued in the previous section that the past morpheme ta determines the temporal relation of E and S to be E prior to S. Since our concern regarding sentences such as (3) is the sequence of E1 and E2 in relation to S2, the location of R is not crucial. Therefore, it is safe to assume that the simple past temporal structure is assigned to atode ‘after’ adjuncts. In (27), the subordinate S2 is anchored to the matrix E1, making the matrix E1 the deictic center for the temporal interpretation of the subordinate clause. This anchoring places the subordinate E2 temporally prior to the matrix E1. Sentence (3a) interprets John’s receiving a letter from Mary as taking place prior to his telephoning her. The structure (27) bears out this interpretation. In (28), the subordinate S2 is anchored to the matrix E1, yielding the temporal interpretation where the subordinate E2 is temporally prior to the matrix E1. This meets the lexical intepretation of atode ‘after’. Notice, however, that the temporal relation between E2 and S1 is indeterminate. Therefore, there are three possible temporal relations between E2 and S1; namely, E2 prior to S1, E2 contemporaneous with S1, and S1 prior to E2. If E2 is prior to S1, or E2 is contemporaneous with S1, the only condition E1 must satisfy is that E1 is future relative to the moment of speech, since the temporal sequence of E1 and E2 is already determined to be E2 prior to E1 by

230

KIRI LEE

anchoring. If E2 already took place, or is taking place simultaneously to the moment of speech, pragmatically there is no need to specify the time sequence of E1 and E2 with the adjunct Mary-kara tegami-o uketor-ta atode, ‘after he receives a letter from Mary.’ Therefore, both the E2_S1 relation and the E2,S1 relation are ruled out. Thus, the temporal structure in (28) represents the correct interpretation for (3b). For a Japanese atode clause, like a maeni clause, the anchoring between the matrix event time and the subordinate speech time is obligatory. This fact can be demonstrated clearly by contrasting the temporal structures with and without obligatory anchoring between E2 and S1 in the following sentences. Observe sentences (29a) and (29b). (29) a. John-ga nihon-e kinoo tat-ru mae-ni, Mary-ni nom Japan-to yesterday leave-n/p before to denwasu-ta. telephone-p “John telephoned Mary before he left for Japan yesterday.” b. Mary-wa John-ga asita New York-ni tat-ta atode, top nom tomorrow to leave-p after kokyoo-ni kaer-ru. hometown-to return-n/p “Mary will go back to her hometown after John leaves for New York tomorrow.” Let us examine the temporal structures for both sentences in terms of anchoring between the matrix E and the subordinate S. Observe the following: (30) a. Temporal structure for (29a) with anchoring: John-ga nihon-e kinoo tat-ru mae-ni, Mary-ni denwasu-ta. “John telephoned Mary before he left for Japan yesterday.” Matrix clause Subordinate clause

R1,E1_S1 | S2_R2,E2 | kinoo

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

231

b. Temporal structure for (29a) without anchoring: John-ga nihon-e kinoo tat-ru mae-ni, Mary-ni denwasu-ta. “John telephoned Mary before he left for Japan yesterday.” * Matrix clause Subordinate clause

R1,E1_S1 S2_R2,E2 | kinoo

For (30), if the subordinate S2 is not anchored to the matrix E1, shown in (30b), the default interpretation takes place interpreting the subordinate S2 at the matrix S1, namely the moment of speech. In this structure, the sequence of the two events is correct, but the past oriented adverb kinoo is placed in the future relative to the moment of speech. Therefore, the sentence should be predicted as ungrammatical based on this temporal structure. On the other hand, by anchoring the subordinate S2 to the matrix E1, shown in (30a), the sequence of two events is temporally specified, but the temporal relationship between the matrix S1 and the subordinate E2 is indeterminate. Then, the past oriented adverb kinoo can modify the subordinate event without any problem. (31) a. Temporal structure for (29b) with anchoring: Mary-wa John-ga asita New York-ni tat-ta atode, kokyoo ni kaer-ru. “Mary will go back to her hometown after John leaves for New York tomorrow.” Matrix clause Subordinate clause

S1_R1,E1 | R2,E2_S2 | asita

b. Temporal structure for (29b) without anchoring: Mary-wa John-ga asita New York-ni tat-ta atode, kokyoo ni kaer-ru. “Mary will go back to her hometown after John leaves for New York tomorrow.”

232

KIRI LEE * Matrix clause Subordinate clause

S1_R1,E1 R2,E2_S2 | asita

(31) is a parallel case to (30). If the subordinate S2 is interpreted at the matrix S1, shown in (31b), the subordinate E2 would clearly be placed prior to the moment of speech, not allowing the subordinate E2 to be modified by the future oriented adverb asita. However, by anchoring the subordinate S2 to the matrix E1, shown in (31a), the temporal relationship between the subordinate E2 and the matrix S1 is undetermined, allowing asita to modify the subordinate event. Examination of the temporal structures in (30) and (31) clearly shows that it is obligatory to anchor the subordinate S to the matrix E in maeni and atode clauses. This is formalized as the following rule: (32) Obligatorily anchor S2 to E1 in the following syntactic structure: [s [adjunct [s... maeni/atode s]]......s] Furthermore, with the obligatory anchoring in (32), the fact that the past morpheme ta is not allowed in the maeni adjunct is explained automatically. With the past morpheme ta in the maeni clause, E2 is always placed prior to E1 via S2 to E1 anchoring. This is inconsistent with the lexical interpretation of maeni. Therefore, the past morpheme ta is ruled out in the maeni adjunct. Likewise, the fact that the non-past morpheme ru is not allowed in the atode adjunct is also explained. With ru in the atode clause, E1 is always placed prior to E2 via S2 to E1 anchoring, contrary to the lexical interpretation of atode. Hence, the non-past morpheme ru is ruled out in the atode adjunct.

5.

Another Temporal Connective, Toki ‘when’

In this section, I will look at sentences with another temporal connective, toki ‘when’, and examine whether the same type of anchoring between the matrix E and the subordinate S is obligatory. Recall that for the temporal connectives maeni ‘before’ and atode ‘after’, the non-past morpheme ru and the past morpheme ta are obligatorily required in these subordinate clauses respectively. Toki, on the other hand, is not such a case. Consider the sentences in (6)

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

233

again, repeated here: (6)

a. John-wa nihon-e ik-ru toki, atarasii kamera-o kaw-ta. top Japan-to go-n/p when new camera-acc buy-p “John bought a new camera when he went to Japan.” b. John-wa nihon-e ik-ta toki, atarasii kamera-o kaw-ta. top Japan-to go-p when new camera-acc buy-p “John bought a new camera when he went to Japan.”

As mentioned in Section 1, the only difference between the sentences in (6) is that the non-past morpheme ru is used in the subordinate clause in (a), and the past morpheme ta is used in the subordinate clause in (b). But this difference is very crucial to clarify the sequence of matrix and subordinate event time. Sentence (6a) with the non-past morpheme is construed as John’s purchasing the new camera prior to his trip, while sentence (6b) with the past morpheme is construed as John’s purchasing the new camera in Japan. This indicates that the lexical interpretation of the temporal connective toki allows two temporal relations of E1 and E2, namely, E1 prior to E2, and E2 prior to E1. Furthermore, observe the following sentence, which also contains the temporal connective toki ‘when’. (33) John-wa nihon-e ik-ta toki, kamera-o motte ik-ta. top Japan-to go-p when camera-acc take-p “When John went to Japan, he took a camera with him.” It is clear that in (33), John’s taking a camera to Japan cannot be either before or after going to Japan. The two events in the matrix and subordinate clauses must be contemporaneous. Notice that the subordinate clause in (33) is exactly the same as that in (6b). In (6b), the sequence of events is E2 prior to E1. On the other hand, in (33), E1 and E2 are construed as contemporaneous. This fact suggests that the lexical interpretation of toki ‘when’ allows another temporal relation between E1 and E2; namely E1 contemporaneous with E2. Thus, unlike sentences with the maeni ‘before’ or atode ‘after’ adjunct, sentences with a toki ‘when’ adjunct can obtain three temporal relations between E1 and E2; namely, E1 prior to E2, E1 contemporaneous to E2, and E1 future relative to E2. Let us here consider the lexical interpretation of toki ‘when’ itself. While maeni ‘before’ and atode ‘after’ specify the sequence of two events described in the matrix and subordinate clauses, toki specifies the matrix event time. In

234

KIRI LEE

other words, E2 provides the additional information to make more specific when it is that E1 takes place. Therefore, the matrix event and the subordinate event must not be temporally far from each other. When E1 is contemporaneous with E2 in a complex sentence with toki, two events overlap completely. However, when some sequence of two events is observed with the toki adjunct, as in the sentences in (6), the two events must be temporally very close; in fact, they must occur almost simultaneously. Therefore, it does not make sense to have the two events on different sides of the moment of speech in the temporal structure. Consider the following sentences: (34) a. Nihon-ni ik-ru-toki, menzeeten-de sake-o kaw-ta. Japan-to go-n/p when duty-free shop-at liquor-acc buy-p “When I went to Japan, I bought liquor at a duty-free shop.” b. *Nihon-ni ik-ru-toki, Empire State biru-ni nobor-ta. Japan-to go-n/p when Empire State Building climb-p “When I went to Japan, I climbed the Empire State Building.” c. Nihon-ni ik-ru-maeni, Empire State biru-ni nobor-ta. Japan-to go-n/p before Empire State Building climb-p “Before I went to Japan, I climbed the Empire State Building.” Both (34a) and (34b) have the same temporal structures. Although we can observe the sequence of E1prior to E2 in (34a), it is easily understood that E2 took place in relation to E1. Therefore, these two events are temporally very close as well as consistent in meaning. On the other hand, (34b) is not plausible.8 By comparing (34b) to (34c), it becomes obvious that if a sentence just indicates a definite sequence of two events, toki is not appropriate. When the temporal connective toki is used in Japanese, the two events are almost concurrent. Here, we will define the lexical interpretation of the temporal connective toki as follows: (35) The lexical interpretation of toki ‘when’: In complex sentences with a toki adjunct, the toki adjunct specifies the point in time of the matrix E1. The two events must be temporally very close to be considered almost concurrent. Thus, the two events have a general relation of E1≅E2. Within the relation of E1≅E2, the following three temporal relations of E1 and E2 are obtained; namely, E1_E2; E1,E2; and E2_E1.

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

235

What we have observed above is that both the non-past morpheme ru and the past morpheme ta appear in sentences with the toki ‘when’ adjunct. We also obtained the three temporal relations; namely E1 prior to E2, E1 contemporaneous with E2, and E1 future relative to E2. In addition, we examined the lexical interpretation of toki, and concluded that the two events must be very close temporally so that they can be considered almost concurrent. Therefore, even when we can obtain the sequence of E1 and E2 from the sentences, they must be temporally very close to each other. With the above observation in mind, let us consider the possible couplings of tense morphemes for complex sentences with the toki ‘when’ adjunct. Since both ru and ta can appear in the subordinate clause, there are four possible combinations of the morphemes. Observe the following temporal configurations: (36) Configuration (N.B. ru = n/p, ta = p) a. b. c. d.

[s [s...ru toki]...ru s] [s [s...ru toki]...ta s] [s [s...ta toki]...ru s] [s [s...ta toki]...ta s]

Recall that sentence (6a) is in configuration (36b), and sentences (6b) and (33) are in configuration (36d). Thus, (36b) can be interpreted as having the temporal relation of E1_E2, and (36d) as having the temporal relation of both E2_E1 and E1,E2. Now, let us examine the following sentences: (37) a. John-wa kondo New York-e ik-ru toki, kuruma-de ik-ru. top next time to go-n/p when car-by go-n/p “When John goes to New York next time, he will drive.” b. John-wa sensyuu New York-e ik-ta toki, kuruma-de ik-ta. top last week to go-p when car-by go-p “When John went to New York last week, he drove.” For the sentences in (37), the intended, and only, reading is that the matrix event and the subordinate event are contemporaneous, i.e, John’s going to New York takes place simultaneously to John’s driving there. In (37a), the non-past morpheme ru is used for both clauses. The action verb with ru in the matrix, coupled with the adverb kondo ‘next time’, places the two events

236

KIRI LEE

future relative to the moment of speech. In (37b), the past morpheme ta is used for both clauses. This, coupled with the adverb sensyuu ‘last week’, places the two events temporally prior to the moment of speech. Notice that sentence (37b) has the same configuration as (33); namely (36d). On the other hand, sentence (37a) has the configuration in (36a). Therefore, the contemporaneity of the two events is observed in both configurations (36a) and (36d). We will now examine if there are any other configurations which allow the contemporaneous reading of two events in complex sentences with toki ‘when’. Observe the following sentences in (38): (38) a. ?/vJohn-wa sensyuu New York-e ik-ru toki, kuruma-de top last week to go-n/p when car-by ik-ta.9 go-p “When John went to New York last week, he drove.” b. *John-wa kondo New York-e ik-ta toki, kuruma-de ik-ru. top next time to go-p when car-by go-n/p (Lit.) “When John went to New York next time, he will drive.” c. John-wa kondo New York-e ik-ta toki, Mary-to top next time to go-p when with syokuzisu-ru. have-n/p dinner “John will have dinner with Mary when he goes to New York next time.” d. John-wa kondo New York-e ik-ru toki, Mary-ni top next time to go-n/p when to renrakusu-ru. contact do-n/p “John will contact Mary when he goes to New York next time.” Sentence (38a) is acceptable only for the interpretation in which the matrix and the subordinate events are contemporaneous. On the other hand, sentence (38b) is completely unacceptable. Notice that (38c) is perfectly acceptable with the interpretation in which John’s having dinner with Mary takes place after he gets to New York. Also, in (38d), the most dominant interpretation is the one in which John contacts Mary prior to his going to New York. The

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

237

above observations indicate that the configuration (36b) has two readings. In addition to the reading of E1 prior to E2 previously shown in (6a), it has the contemporaneous reading between E1 and E2 as exhibited in sentence (38a). On the other hand, the configuration (36c) does not have the contemporaneous reading, as seen in (38b). The following sentence (39a) has the configuration shown in (36b), and sentence (39b) has the configuration shown in (36d). Examine the sentences: (39) a. Hikooki-ga Narita-ni tuk-ru toki, hidoku yure-ta. plane-nom in arrive-n/p when terribly bounce-p “When the plane arrived in Narita, it bounced terribly.” b. Hikooki-ga Narita-ni tuk-ta toki, hidoku yure-ta. plane-nom in arrive-p when terribly bounce-p “When the plane arrived in Narita, it bounced terribly.” In (39), both sentences attain contemporaneity between the matrix and subordinate events, i.e., the plane’s approach to Narita was bouncy. However, it is possible to interpret (39b) as indicating a sequence between the matrix and subordinate events, i.e., the plane bounced terribly after it landed in Narita. Based on the above observation, let us show the four possible configurations of the toki adjunct in (36) with their observed temporal interpretations. (40)

Configuration a. b. c. d.

[s [s...ru toki]...ru s] [s [s...ru toki]...ta s] [s [s...ta toki]...ru s] [s [s...ta toki]...ta s]

Temporal interpretation E1_E2, E1_E2, E2_E1, E2_E1,

E1=E2 E1=E2 *E1=E2 E1=E2

The configurations in (40a), (40b) and (40d) attain two interpretations each, i.e., E1 prior to E2, and E1 contemporaneous with E2 in (40a) and (40b); and E2 prior to E1, and E1 contemporaneous with E2 in (40d). On the other hand, the configuration in (40c) attains only the E2 prior to E1 interpretation as shown in sentences (38b) and (38c). Our task next is to account for all of the interpretations that the sentences with toki exhibit in (40), by assigning correct temporal structures to them.

238 6.

KIRI LEE To Anchor or not to Anchor in Toki Adjuncts

In this section, I will examine the temporal structures for sentences with toki ‘when’ to account for all the temporal interpretations shown in (40). First, the configuration in (40a) will be considered. Sentences with toki in this configuration can be interpreted as having contemporaneity between E1 and E2 or having the E1 prior to E2 sequence. Sentence (41a) attains the former reading, while sentence (41b) attains that latter reading in the following. (41) a. John-wa raisyuu New York-e ik-ru toki, kuruma-de ik-ru. top next week to go-n/p when car-by go-n/p “When John goes to New York next week, he will drive.” b. John-wa raisyuu New York-e ik-ru toki, Mary-ni top next week to go-n/p when to renrakusu-ru. contact do-n/p “John will contact Mary when he goes to New York next week.” Although they are syntactically identical, the temporal relation between E1 and E2 is different in (41a) and (41b). In order to limit our discussion to the simple future interpretation of the non-past morpheme ru, raisyuu ‘next week’ is added to the sentence to avoide a habitual interpretation. Therefore, we assign the temporal structure of simple future to the matrix clause. Now two issues must be addressed: one is which temporal structure, namely the temporal structure of simple present or that of simple future, should be assigned to the subordinate clause in this configuration. The other is whether or not obligatory anchoring should take place between S2 and E1 in these sentences. First, let us consider the following temporal structures with anchoring for (41). (42) The temporal structure for (41) with anchoring: a. S_R,E for subordinate clause. Matrix clause Subordinate clause

S1_R1,E1 | S2_R2,E2

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

239

b. S,R,E for subordinate clause. Matrix clause Subordinate clause

S1_R1,E1 | S2,R2,E2

In (42a) and (42b), the matrix clause has the temporal structure of the simple future. In (42a), the temporal structure of simple future is assigned to the subordinate clause. S2 is interpreted at E1 by anchoring, and the E1_E2 relation is attained via this anchoring. Notice that this structure correctly represents the interpretation of sentence (41b). On the other hand, in (42b), the temporal structure of simple present is assigned to the subordinate clause. This structure attains the E1=E2 relation via S2 to E1 anchoring. This is the right interpretation for (41a). Now, we will compare the structures in (42) to (43), in which anchoring S2 to E1 does not take place. (43) The temporal structure for (41) without anchoring: a. S_R,E for subordinate clause. Matrix clause Subordinate clause

S1_R1,E1 S2_R2,E2

b. S,R,E for subordinate clause. Matrix clause Subordinate clause

S1_R1,E1 S2,R2,E2

In (43), anchoring between S2 and E1 does not take place, and the subordinate clause is construed at the matrix S by default. What (43a) indicates is that both E1 and E2 are temporally future relative to the moment of speech, and the temporal relation of E1 and E2 is undetermined. Recall, however, that the lexical interpretation of toki ‘when’ puts two events as almost concurrent, thus ensuring the E1=E2 interpretation. This is the correct structure for (41a). On the other hand, the temporal structure in (43b) is ill-formed. S2 is construed at S1 by default, and as a result, E2 is contemporaneous to S1. Notice that, since the matrix clause has a temporal structure of simple future, the sentence as a whole must have a future interpretation. Therefore, it is pragmatically impossible to construe the subordinate E2 at the matrix S when a sentence has a future interpretation as a whole. Thus, the temporal structure in (43b) must be rejected.

240

KIRI LEE

As we have seen above in (42) and (43), assignment of both the temporal structures S,R,E and S_R,E to the non-past morpheme ru in the toki ‘when’ clause is crucial to accounting for the two temporal readings obtained in the configuration (40a). When a matrix verb has a non-past morpheme ru, a context and/or lexical items such as asita ‘tomorrow’, raisyuu ‘next week’, mainiti ‘everyday’, and itumo ‘always’, can disambiguate between a future reading and a habitual reading. However, a non-past morpheme ru in the toki ‘when’ adjunct seems transparent to this distinction. Thus, we assume that the non-past morpheme ru in the toki ‘when’ clause allows two temporal structures, namely, S,R,E and S_R,E, regardless of the temporal interpretation of the matrix tense. We also attained the correct results with and without anchoring between E1 and S2 for (41). Let us further compare the structures with and without anchoring to see if this holds in all the other configurations with toki ‘when’. Consider the configuration in (40b), repeated here: (40) b.

Configuration [s [s...ru toki]...ta s]

Temporal interpretation E1_E2, E1=E2

Sentences (6a), (38a), and (39a) are in this configuration. (6a) has the reading of E1_E2, while (38a) and (39a) have the contemporaneous reading of E1=E2. They are repeated here as (44a), (44b), and (44c) respectively: (44) a. John-wa nihon-e ik-ru toki, atarasii kamera-o kaw-ta. top Japan-to go-n/p when new camera-acc buy-p “John bought a new camera when he went to Japan.” b. ?/vJohn-wa sensyuu New York-e ik-ru toki, kuruma-de top last week to go-n/p when car-by ik-ta. go-p “When John went to New York last week, he drove.” c. Hikooki-ga narita-ni tuk-ru toki, hidoku yure-ta. plane-nom in arrive-n/p when terribly bounce-p “When the plane arrived in Narita, it bounced terribly.” Consider the following temporal structures with anchoring for the sentences in (44).

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

241

(45) Temporal structures for (44) with anchoring: a. Matrix clause Subordinate clause b. Matrix clause Subordinate clause

R1,E1_S1 | S2_R2,E2 R1,E1_S1 | S2,R2,E2

In (45a), the temporal structure of simple past is assigned to the matrix clause, and that of simple future is assigned to the subordinate clause. S2 is anchored to E1. Since S2 is temporally prior to E2, E1 is also prior to E2. Although the temporal relation of S1 and E2 is indetermined, the lexical interpretation of toki ‘when’ ensures to place E1 and E2 temporally very close. Thus, this temporal structure represents the temporal interpretation of (44a) correctly. On the other hand, in (45b), the temporal structure of simple present is assigned to the subordinate clause. S2 is construed at E1 by anchoring. This enables E2 to be concurrent with E1. This is a desirable structure for the sentences in (44b) and (44c) because they have the contemporaneous interpretation of E1=E2.10 Let us now consider the temporal structures for (44) without anchoring: (46) Temporal structures for (44) without anchoring: a. Matrix clause Subordinate clause

R1,E1_S1 S2_R2,E2

b. Matrix clause Subordinate clause

R1,E1_S1 S2,R2,E2

In (46a), the temporal structure of simple future is assigned to the subordinate clause. No anchoring takes place, and thus S2 is construed at S1 by default. Although the E1_E2 relation is attained in this structure, the lexical interpretation of toki rules it out, because it is implausible to consider E1 and E2 to be close enough to be almost concurrent in the E1_S1_E2 relation. In (46b), the temporal structure of simple present is assigned to the subordinate clause. S2 is construed at S1 by default. Since E2 is contemporaneous with S2, we will attain the temporal relation of E1 prior to E2. Notice that this temporal structure shows the E1_S1,E2 relation, indicating that the matrix event took place in the past and that the subordinate event is taking

242

KIRI LEE

place at the moment of speech. The temporal structure (46b) is thus ruled out by the lexical interpretation of toki ‘when’. We will continue our observation on the remaining configurations in (40). The configuration in (40c) is repeated here: (40)

Configuration c. [s [s...ta toki]...ru s]

Temporal interpretation E2_E1, *E1=E2

In this configuration, only the relation of E2_E1 is obtained. The sentences in (38b) demonstrated that it is impossible to attain the contemporaneity of E1 and E2, and the sentence in (38c) demonstrated that the sequence of E2 prior to E1 is attained. We will repeat (38b) and (38c) as (47a) and (47b) respectively: (47) a. *John-wa kondo New York-e ik-ta toki, kuruma-de ik-ru. top next time to go-p when car-by go-n/p (Lit.) “When John went to New York next time, he will drive.” b. John-wa kondo New York-e ik-ta toki, Mary-to top next time to go-p when with syokuzisu-ru. have-n/p dinner “John will have dinner with Mary when he goes to New York next time.” Since the past morpheme ta is contained in the subordinate clause, we will assign the temporal structure of simple past to the subordinate clause. The non-past morpheme ru and the adverb kondo ‘next time’ ensure the simple future reading for the matrix clause; therefore, we will assign the temporal structure of simple future to the matrix clause. Consider the following temporal structures with and without anchoring between E1 and S2. (48) a. Temporal structure for (47) with anchoring: Matrix clause Subordinate clause

S1_R1,E1 | R2,E2_S2

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

243

b. Temporal structure for (47) without anchoring: Matrix clause Subordinate clause

S1_R1,E1 R2,E2_S2

In (48a), S2 is anchored to E1. Via this anchoring, E2 is prior to E1. And this is a correct interpretation for the sentence (47b). In (48b), S2 is construed at S1 by default. E2 is prior to S2; S2 is contemporaneous with S1; and S1 is prior to E1. Therefore, E2 is prior to E1. This seems correct for the interpretation of (47b); however, notice that E2 prior to S1 means that E2 is in the past, and E1 future relative to S1 means that E1 is in the future. It is impossible to consider these two events in (48b) temporally close enought to be concurrent events. Therefore, (48b) is ruled out by the lexical interpretation of toki ‘when’. Thus, the temporal structure of (48a) is the right one for sentence (47b). Notice, furthermore, that in both structures in (48), it is impossible to attain the contemporaneous relation between E1 and E2. And this is a desirable result, since configuration (40c) does not obtain such a reading, as sentence (47a) shows. Finally, we will consider configuration (40d), which is repeated here: (40)

Configuration d. [s [s...ta toki]...ta s]

Temporal interpretation E2_E1, E1=E2

We examined sentences (6b), (33), (37b) and (39b) in this configuration. They are repeated here as (49a), (49b), (49c), and (49d) respectively: (49) a. John-wa nihon-e ik-ta toki, atarasii kamera-o kaw-ta. top Japan-to go-p when new camera-acc buy-p “John bought a new camera when he went to Japan.” b. John-wa nihon-e ik-ta toki, kamera-o motte ik-ta. top Japan-to go-p when camera-acc take-p “When John went to Japan, he took a camera with him.” c. John-wa sensyuu New York-e ik-ta toki, kuruma-de ik-ta. top last week to go-p when car-by go-p “When John went to New York last week, he drove.” d. Hikooki-ga Narita-ni tuk-ta toki, hidoku yure-ta. plane-nom in arrive-p when terribly bounce-p “When the plane arrived in Narita, it bounced terribly.”

244

KIRI LEE

Recall that (49a) has the reading of E2 prior to E1, (49b) and (49c) have the contemporaneous reading of E1=E2, and finally (49d) is ambiguous for both readings. Since all the clauses in (49) have the past morpheme ta, the temporal structure of simple past is assigned to them. Consider the following two temporal structures: (50) a. Temporal structure for (49) with anchoring: Matrix clause Subordinate clause

R1,E1_S1 | R2,E2_S2

b. Temporal structure for (49) without anchoring: Matrix clause Subordinate clause

R1,E1_S1 R2,E2_S2

In (50a), the sequence of E1 and E2 is specified by the S2 to E1 anchoring. Therefore, there is no room for E1 and E2 to be interpreted contemporaneously. This is desirable for (49a) where the temporal relation of E2 prior to E1 is obtained. On the other hand, in (50b), S2 is interpreted at S1. All that the temporal structure indicates is that E1 and E2 are temporally prior to the moment of speech. Therefore, the temporal structure in (50b) allows E1 and E2 to be contemporaneous, and their contemporaneity is ensured by the lexical interpretaion of toki ‘when’, because two events must be almost concurring in sentences with the toki adjunct. This structure represents correctly the contemporaneous reading of E1 and E2 in (49b), (49c), and (49d). In sum: what we have demonstrated in this section is that the toki adjunct is interpreted both at the matrix event time and at the moment of speech. And the lexical interpretation of toki ‘when’ filters out temporal structures in which the two events are temporally far from each other. Therefore, the following generalization is made: (51) Construe S2 at both E1 and S1 in the following syntactic structure: [s [adjunct [s... toki s]]....s] The lexical interpretation of toki ‘when’ given in (35) is repeated here: (35) The lexical interpretation of toki ‘when’: In complex sentences with a toki ‘when’ adjunct, the toki ‘when’

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

245

adjunct specifies the point in time of the matrix E1. The two events must be temporally close enough to be considered almost concurrent. Thus, the two events have a general relation of E1≅E2. Within the relation of E1≅E2, the following three temporal relations of E1 and E2 are obtained; namely, E1_E2; E1,E2; and E2_E1. Recall that it is obligatory to have the non-past morpheme ru in the maeni ‘before’ adjunct, and the past morpheme ta in the atode ‘after’ adjunct; and that the anchoring between E1 and S2 is obligatory in those adjuncts. In contrast, in the toki ‘when’ adjunct, both tense morphemes appear in its clause, and the anchoring between E1 and S2 is not obligatory.

7.

Implications of this Study towards the Theory of Tense Interaction

In this study, we have observed the Japanese tense mechanism in complex sentences with temporal connectives maeni, atode, and toki, and demonstrated that the subordinate clauses headed with maeni and atode in these sentences take the event time in the matrix clauses as a deictic center, while those with toki can take either the matrix event time or the matrix speech time as a deictic center, allowing temporal ambiguity. In this section, I will outline Hornstein’s work on tense in English complex sentences and explore some implications of the present study towards the theory of tense in general. Chapter 2 of Hornstein (1990) discusses extensively the English counterparts of complex sentences I have discussed in this study. Hornstein observes the following sentences, and concludes that the temporal interpretive properties of multi-tensed sentences are not alone responsible for the unacceptability of these sentences. Observe: (52) a. *John comes after Harry arrived. b. *John came before Harry arrives. There is nothing logically wrong with the sentences in (52). In (52a), the lexical interpretation of after places John’s coming temporally future to Harry’s arrival. Therefore, the sentences must have the interpretation in which the matrix event is future relative to the subordinate event. Since the subordinate clause is simple past and the matrix clause is simple present, the subordinate event is temporally prior to the moment of speech and the matrix clause is future relative or contemporaneous to the moment of speech. There-

246

KIRI LEE

fore, the matrix event is future relative to the subordinate event, meeting the requirements of the lexical interpretation of after. Therefore, logically, the sentence should be acceptable. Likewise, in (52b), the lexical interpretation of before places John’s coming temporally prior to Harry’s arrival. This means that the sentence must express the temporal interpretation in which the matrix event is prior to the subordinate event. Since the subordinate clause is simple present and the matrix clause is simple past, the matrix event is temporally prior to the subordinate event. Therefore, again, there should be nothing wrong with the sentence logically. However, these sentences are unacceptable. With the above observation, Hornstein proposes some rules and constraints to restrict arbitary coupling of matrix and subordinate tenses so that sentences such as those in (52) would be ruled out.11 But what concerns us here is that, for complex sentences with temporal connectives in English, it is assumed that the subordinate clause is interpreted at the moment of speech by a default rule. In Chapter 4, Hornstein (1990) has also pointed out the ambiguity in a sentence such as John heard that Mary was pregnant, where Mary’s pregnancy may or may not any longer hold at the moment of speech, and claims that this ambiguity is attributed to either the S2 to E1 anchoring or the default reading. Observe the following diagrams: (53) a. Matrix clause Subordinate clause b. Matrix clause Subordinate clause

R1,E1,_S1 | S2,R2,E2 R1,E1_S1 S2,R2,E2

In (53), the temporal structure of simple past is assigned to the matrix clause and that of simple present is assigned to the subordinate clause. This is because Hornstein claims that the sentence John heard that Mary was pregnant has a direct quote counterpart such as John heard, “Mary is pregnant.” In (53a), S2 is anchored to E1 interpreting the subordinate clause at the matrix event time. E2 is concurring with E1. This structure represents the reading that Mary’s pregnancy does not hold at the moment of speech. On the other hand, in (53b), S2 is interpreted at S1 as default, therefore leading to the interpretation of the subordinate event as contemporaneous to the moment of speech, i.e., that Mary’s pregnancy holds at the moment of speech.

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

247

As shown in Hornstein’s theory, for complex sentences with temporal connectives in English, S2 is interpreted at S1 as default, while in reported speech sentences, S2 is interpreted at both S1 and E1. On the other hand, in Japanese, the S2 to E1 anchoring is obligatory in the sentences with maeni ‘before’ and atode ‘after’, indicating that the deictic center for the temporal interpretation of the subordinate clause is the event time in the matrix clause. In a sentence with the temporal connective toki, ‘when’, the subordinate clause can be interpreted at two temporal points, namely at the matrix E and at the moment of speech as default. Notice that, in both English and Japanese, when a sentence exhibits a temporal ambiguity, it allows its subordinate clause to be construed at both the matrix event time and the moment of speech. Although more languages need to be investigated, it seems promising to speculate that, in any given language, when ambiguity is observed in the temporal interpretation of complex sentences, there are two possible anchoring times in the matrix clause, namely, E1 and S1.

Notes

1.

In order to make the tense clear in the example sentences, throughout this study, the nonpast tense morpheme ru is used as a morpho-phonemic representation for [ru] or [u], and the past tense morpheme ta as a morpho-phonemic representation for [ta] and [da].

2.

The terminology “the perfect of result” is from Comrie (1976).

3.

For a more extensive discussion on the past tense morpheme ta in Japanese, see Lee (1993).

4.

This observation has been made and applied also to the temporal relation between a relative clause and a matrix clause and other adverbial clauses. See e.g. Teramura (1984), Sunagawa (1986), and Mihara (1992).

5.

This interpretation is discussed in detail in Chapter 4 of Lee (1993).

6.

When the matrix clause has a habitual interpretation with the non-past morpheme ru, the temporal structure of simple present is assigned. Observe the following sentence: (I)

John-wa mainiti zimusyo-o de-ru maeni, uti-ni kanarazu top everyday office-acc leave-n/p before home-to surely denwasu-ru. telephone-n/p “John telephones home without fail before he leaves his office everyday.”

The subordinate clause has ru, and therefore, the following two temporal structures are considered:

248

KIRI LEE (iia) Matrix clause Subordinate clause (iib) Matrix clause Subordinate clause

S1,E1,R1 | S2,E2,R2 S1,R1,E1 | S2_E2,R2

In (iia), the temporal structure of simple present is assigned, while in (iib), that of simple future is assigned. In (iia), S2 is anchored to E1. Since E2 is contemporaneous with S2, it is also contemporaneous with E1. The contemporaneity of E1 and E2 is contradictory to the lexical interpretation of maeni ‘before’. Hence (iia) is ruled out. On the other hand, in (iib), E1 is clearly prior to E2 via S2 to E1 anchoring. Therefore, this structure is desirable. This observation indicates that the temporal structure of simple future is obligatorily assigned to the subordinate clause with maeni, even when its matrix clause with the non-past morpheme ru is construed as a habitual action. 7.

Hornstein (1990) states that when he reorganizes the three points S, R, and E in Reichenbach’s tense theory, S is a deictic element anchored within the discourse situation, and anchored to the moment of speech by default if there is no restriction in the discourse situation. However, he does not redefine ‘Point of Speech’ as ‘Anchor Time’.

8.

Sentence (34b) is acceptable if there is a specific context which connects the two events in relation to each other. For instance, the following sentence seems acceptable: Boston-kara New York keiyu-de nihon-ni ik-ru toki, zikan-ga ar-ta node from via Japan-to go-n/p when time-nom exisit-p since Empire State biru-ni nobor-ta. building climb-p “When I went to Japan from Boston via New York, I climbed the Empire State Building since I had some time.”

9.

Judgement regarding the acceptability of sentence (38a) varies among native speakers of Japanese. Although some speakers accept the sentence, others feel that it is less perfect compared to the following sentence: (i)

John-wa sensyuu New York-e ik-ta toki, kuruma-de ik-ta. top last week to go-p when car-by go-p “When John went to New York last week, he drove.”

The above sentence clearly has the contemporaneous reading between E1 and E2. We will discuss more on this difference in judgement later in the current section. 10.

The following sentence has exactly the same temporal structure as the ones discussed in (44). It is however completely unacceptable. Observe: i)

*Kyonen Canada-o tabisu-ru toki, pasupooto-o otos-ta. last year acc travel-n/p when passport-acc drop-p “When I travelled in Canada last year, I lost my passport.”

Compare the sentence above to the following sentence: ii) Kyonen Canada-e ik-ru toki, pasupooto-o otos-ta. last year Canada-to go-n/p when passport-acc drop-p “When I went to Canada, I lost my passport.”

RELATIVE TENSE AND ABSOLUTE TENSE IN JAPANESE

249

In the sentence in (ii), the subordinate verb is ik-ru ‘go’, and the sentence is acceptable for the reading that losing the passport occurred in the process of getting to Canada. The verb tabisu-ru ‘travel’ in (i) is a Continuative verb in Kindaichi’s classification (1976); the English terms are from Jacobsen (1981 and 1992). Continuative verbs acquire the progressive meaning in the te-iru form. On the other hand, the verb ik-ru ‘go’ is an Instantaneous verb that expresses, in the te-iru form, the interpretation that refers to a state resulting from an event that has occurred at some point in the past. Therefore, I attribute the unacceptability of sentence (i) to the lexical interpretation of the verb tabisu-ru ‘travel’. It seems that the te-iru form of a Continuative verb must be used in the toki ‘when’ adjunct when the matrix event takes place after the subordinate event begins. Compare sentence (i) to (iii) and (iv): (iii) Kyonen Canada-o tabisu-ru toki, pasupooto-o koosinsu-ta. last year acc travel-n/p when passport-acc renew-p “When I travelled in Canada last year, I renewed my passport.” (iv) Kyonen Canada-o tabisite i-ru toki, pasupooto-o otos-ta. last year acc traveling be when passport-acc drop-p “When I was travelling in Canada, I lost my passport.” The reading attained in (iii) is that renewing the passport took place before the trip. 11.

The rule Hornstein formulated to account for the unacceptability of the sentences such as in (52) is given below in (i), and this rule applies to syntactic configurations such as in (ii). (i)

RTC (rule for temporal connectives): In (52), write the Basic Tense Structure (BTS hereafter) of TNS2 under the BTS of TNS1. Associate the S points. Associate the R points by moving R2 to R1, placing E2 accordingly.

(ii)

[s... TNS1...[adjunct TC[s...TNS2...]]], where TC is a temporal connective, e.g., before or after.

In his theory, the anchoring of two R points is crucial to rule out unacceptable sentences. See Chapter 2 of Hornstein (1990) for details.

References Comrie, Bernard 1976 Aspect. Cambridge: Cambridge University Press. Comrie, Bernard 1985 Tense. Cambridge: Cambridge University Press. Hornstein, Norbert 1990 As time goes by: Tense and universal grammar. Cambridge, MA: MIT Press. Jacobsen, Wesley M. 1981 Transitivity in the Japanese verbal system. Dissertation, University of Chicago.

250

KIRI LEE

Jacobsen, Wesley M. 1992 The Transitive structure of events in Japanese. Tokyo: Kuroshio Shuppan. Kindaichi, Haruhiko 1950 “Kokugo dooshi no ichibunrui”. Reprinted in H. Kindaichi (ed), 27-62. Kindaichi Haruhiko (ed) 1976 Nihongo dooshi-no asupekuto. Tokyo: Mugi Shobokan. Lee, Kiri 1993 Tense interaction in complex sentences in Japanese and English. Dissertation, Harvard University. Machida, Ken 1989 Nihongo no zisei to asupekuto. Tokyo: Aruku Press. Mihara, Kenichi 1992 Zisei kaisyaku to tougo gensyoo. Tokyo: Kuroshio Shuppan. Nakau, Minoru 1976 “Tense, aspect, and modality”. In M. Shibatani (ed), 421-481. Reichenbach, Hans 1947 Elements of symbolic logic. Macmillan. (Reprinted in 1966 by Free Press.) Shibatani Masayoshi 1976 Syntax and semantics 5: Japanese generative grammar. New York: Academic Press, Soga, Matsuo 1983 Tense and aspect in modern colloquial Japanese. Vancouver: University of British Colombia Press. Sunagawa, Yuriko 1986 Suru, sita, siteiru. Tokyo: Kuroshio Shuppan. Teramura, Hideo 1984 Nihongo no sintakusu to imi, Vol 2. Tokyo: Kuroshio Shuppan.

SPECIFIC NP IN SCOPE

251

Specific NP in Scope Becky Kennedy Harvard University Extension School and Franciscan Children’s Hospital

1.

Introduction

In this paper I look at the phenomenon of a specific NP in scope. To understand the notion of specific NP, consider the following examples from Karttunen (1976): (1)

Bill didn’t see a misprint.

(2)

a. There is a misprint which Bill didn’t see. b. Bill saw no misprints.

(2a) reflects the specific interpretation for the indefinite NP a misprint in (1), whereas (2b) reflects the nonspecific interpretation. The specific NP in interpretation (2a) refers to an entity that the speaker has in mind, a notion often invoked in defining specificity. It has often been stated as well that the specific interpretation requires that the NP in question take widest possible scope (cf. Heim 1982); thus the indefinite a misprint takes wide scope in (2a) and is fronted in that paraphrase. In spite of these accepted intuitions regarding specificity, the nature of the phenomenon has been disputed: The association of specificity with a binary semantic feature has been questioned, as has been the semantic nature of specificity (e.g., Kripke 1977; Ludlow and Neale 1991). Karttunen took the position that specificity is an effect of scope and rejected the notion of a semantic feature corresponding to specificity, arguing that what appears to be a specific NP is simply an NP bearing wide scope over one but not necessarily every operator.

252

BECKY KENNEDY

I follow Karttunen here regarding this last point, insofar as I argue for a domain of specificity, such that an NP can be interpreted as specific yet bear narrow scope with respect to some operator in the sentence. I maintain, however, that there is indeed a binary semantic ambiguity. Because I take the position that a specific NP can be in scope, I distinguish between a referential NP — an NP, such as a name, that does not interact scopally with other elements — and a specific NP. I suggest that behaviors that have been explained in terms of a referential/quantificational ambiguity are better accounted for as effects of specificity in an NP. Specificity is seen as determined by speaker intent rather than by discourse contingencies or shared speaker/ hearer interpretation; this view differs from that of other analyses of specificity, notably Enç (1991) and Ludlow and Neale (1991). In section 2, I discuss the semantic ambiguity hypothesis and adduce various phenomena the hypothesis has been held to explain. In section 3, specificity is discussed. Section 4 introduces the notion of specificity in scope: I look at certain problems for the ambiguity hypothesis and argue that an ambiguity account of specificity in which the association with referentiality is abandoned will account for problematic data. In section 5, I develop the proposal that a specific NP can occur in scope, such that it has a domain of specificity; and I show that certain island-extraction phenomena that were not fully accounted for by invoking the referentiality/nonreferentiality of the extracted NP can be explained by utilizing the notion of domain of specificity. In section 6, I look at indirect specificity in indirect discourse contexts; section 7 presents conclusions.

2.

The Semantic Ambiguity Hypothesis

Consider (3)-(4), noting the ambiguity in both examples. In (3), from Fodor and Sag (1982), the indefinite subject can refer to a particular student that the speaker has in mind, or it can characterize a student whose identity is unspecified. In (4), from Donnellan (1966), the definite subject NP can likewise refer to a particular individual, or it can characterize Smith’s murderer, whoever he was. (3)

A student in the syntax class cheated on the final exam.

(4)

The man who murdered Smith is insane.

SPECIFIC NP IN SCOPE

253

It was argued in both cases that the interpretation of the subject NP reflected true ambiguity rather than vagueness; this type of ambiguity, moreover, has been observed to correlate systematically with certain syntactic and semantic effects. Donnellan treated the ambiguity in definite NPs like the subject of (4) as a referential/attributive alternation and discussed it in terms of situation of utterance: The interpretation of the NP is selected by the context in which it occurs and will covary with the circumstances of utterance and speaker intent. The attributive interpretation was later treated by Fodor and Sag as a quantificational interpretation of the NP in question, because an NP with attributive interpretation seems to display behavior characteristics of a quantified expression. Kripke (1977), however, suggested that Donnellan’s ambiguity hypothesis was not necessary. Like Karttunen, he argued that certain phenomena accounted for by an ambiguity approach could be better explained in terms of scope; and he appealed to a notion of speaker’s reference to argue that referential use need not entail a semantically referential interpretation of a definite. Ludlow and Neale (1991) extended Kripke’s approach to indefinites, arguing for a unitary quantificational analysis of indefinites and an appeal to pragmatics to account for the intuition of specificity. From a methodological perspective, the proponent of an ambiguity hypothesis must show that an ambiguity account is necessary. In the next subsection, I consider the scopal data invoked in support of referential/quantificational ambiguity in NPs. 2.1 Fodor and Sag: Referential Indefinites Fodor and Sag (1982) adduced crucial scopal phenomena that they explained by means of an ambiguity analysis of indefinite NPs, noting that an indefinite (and putatively quantificational) NP behaves in some contexts like a referential NP. They considered the behavior of indefinite NPs in island-extraction and VP-deletion structures, looking, for instance, at the following sentence: (5)

John overheard the rumor that a student of mine had been called before the dean.

An indefinite phrase like a student of mine has often been analyzed as an existential quantified phrase or QP (although cf. Heim 1982; Kamp 1981). Yet, Fodor and Sag observed, it is possible to assign to that NP in (5) a scopal value much wider than that ordinarily held by a QP in such an environment.

254

BECKY KENNEDY

The scope of a quantifier embedded within a complex NP is generally restricted to the complex NP itself, which functions as a scopal island; thus the restricted scope of each of my students in (6): (6)

John overheard the rumor that each of my students had been called before the dean.

(6) cannot mean: For each of my students (x), John overheard the rumor that x had been called before the dean. Yet (5) can be interpreted with the indefinite NP receiving maximally wide scope: For a student of mine (x), John overheard the rumor that x had been called before the dean. Fodor and Sag explained this observation by proposing that a N’ is not exceptional among QPs in its lack of respect for scopal islands; instead, indefinites have an additional referential interpretation, under which a student of mine appears to take maximally wide scope but is in fact behaving like a referring expression (like a name) and therefore does not interact scopally at all, so that the referential interpretation is distinct from a widest-scope quantificational interpretation. (The question whether a widest scope reading for an indefinite entails referential interpretation is considered later in greater detail, in the context of Abusch 1994, and Ludlow and Neale 1991; cf. also Kaplan 1989; Kripke 1972, 1977.) A binary lexical ambiguity in the indefinite determiner translates, they argued, into a referential/quantificational ambiguity in indefinite NPs. Fodor and Sag suggested that this analysis might be extended to definite determiners as well as the cardinal determiners, some, and many: determiners that introduce what has been termed the class of weak NPs (Barwise and Cooper 1981).1 2.2 Extraction from NP: An Ambiguity Analysis We see a further connection between Donnellan’s study of ambiguity in definites and Fodor and Sag’s ambiguity analysis of indefinites when we observe environments in which extraction is attempted from definite and indefinite NPs. Extraction from NP has been discussed by a number of scholars (cf., for instance, Bach and Horn 1976; Diesing 1992; Erteschik-Shir 1973; Kuno 1987), under a variety of proposals. Kuno, in particular, demonstrated the failure of a purely syntactic explanation and presented a nonsyntactic account of the phenomenon, showing that extraction of X from [NP Prep X] is possible only when X is a list head and NP is an attribute of X.

SPECIFIC NP IN SCOPE

255

Fiengo and Higginbotham (1981) contrasted examples like those in (7) with examples like (8a-b): (7)

a. John bought the picture of everyone. b. ??Who did John buy the picture of t ?

(8)

a. John bought a picture of everyone. b. Who did John buy a picture of t ?

Fiengo and Higginbotham noticed that a WH-term and a QP can be extracted scopally from the indefinite picture NP in (8) but not from the definite in (7); this contrast was attributed to a distinction in Specificity, a binary lexical feature whose setting depended in value on factors like singularity, definiteness, and presence of a proper-NP subject of NP. Under the Specificity Condition, a Specific NP was opaque to extraction. Fiengo (1987), however, noted possible extraction from comparative definites: (9)

a. John bought the biggest picture of everyone. b. Who did John buy the biggest picture of t ?

In Fiengo and Higginbotham’s terms, the definite picture NPs in (9a-b) permit extraction because they are [-Specific]. This account can be recast, however, as an extension to definites of Fodor and Sag’s semantic ambiguity: A definite NP interpreted quantificationally can interact scopally with an embedded WH-phrase or QP. Thus (a) semantic ambiguity would apply to definites as well (as proposed by Donnellan) and (b) extraction possibilities not only of NP but also from NP would be seen to covary with the NP’s semantic interpretation. In further support of this approach, note enhanced extractability from the picture NPs containing contrastive stress marking or a restricted relative clause in (10a-b) and (11a): (10) a. John bought the GOOD picture of everyone./ Who did John buy the GOOD picture of t? b. John bought the PICTURE of everyone./ Who did John buy the PICTURE of t? (11) a. John chose the picture that Bill had chosen of everyone. b. John chose the picture, which Bill had chosen, of everyone. The restrictive/nonrestrictive relative clause distinction was associated with

256

BECKY KENNEDY

Donnellan’s referential/attributive distinction. It is possible that the nonrestrictive relative induces the referential interpretation by introducing the presupposition that the object denoted by the NP has the property expressed; comparatives, contrastives, and restrictively modified NPs, in contrast, designate set membership without indicating a referent. This last discussion of extraction from NP adds two facets to our investigation of semantic ambiguity in definites and indefinites. First, we note that Fiengo and Higginbotham explained extraction phenomena in terms of the semantic ambiguity of [+Specificity]. Second, we derive another behavioral correlate to interpretation. Fodor and Sag’s referential interpretation permitted an NP to escape from a scopal island; their quantificational interpretation can be associated with possible extraction from NP.2

3.

Specificity

3.1 Defining Specificity We have accrued, thus far, a number of attributes of a specific NP. A specific NP is often described as referring to an entity that the speaker has in mind. Fodor and Sag’s specific NP has referential interpretation, can escape an island, and appears to take widest scope; Fiengo and Higginbotham’s Specific NP blocks extraction of an internal QP. Both these behaviors are explained by the immunity of a specific NP to scopal interaction. Two other characteristics of a specific NP have been described in the literature. First, there is an existential presupposition associated with a specific NP: If a speaker has a referent in mind in uttering a specific NP, the existence of a referent satisfying the description in the NP is presupposed. (Cf. Enç 1991. Note, however, that Diesing 1992, characterizing quantificational indefinites as presuppositional, used presupposition in a somewhat different way to refer to an existential presupposition associated with the set of entities to which the quantifier is applied; cf. Milsark 1974, and note 2.) Consider (12): (12) a. Some Spaniard seemed critical of John’s position. b. Some Spaniard seemed to be critical of John’s position. The subject NP in (12a) is necessarily specific, whereas the subject NP in

SPECIFIC NP IN SCOPE

257

(12b) can also be nonspecific (cf. the analysis by Williams 1983, of examples like (12a-b) in terms of possible scopal value for the subject NP). One test for the presupposition of existence (and thus for specificity) proceeds as follows. If existence is not presupposed, its denial will produce a paradox: The speaker will have asserted existence and then denied it, presenting the hearer with two contradictory pieces of information. But if existence is presupposed, its denial will create anomaly: The sentence will be uninterpretable, insofar as a sentence’s presupposition must hold in order for it to be uttered felicitously and assigned a truth value. Now consider (13a-b): (13) a. Some Spaniard seemed critical of John’s position; yet there was no Spaniard present. b. Some Spaniard seemed to be critical of John’s position; yet there was no Spaniard present. In uttering (13b), the speaker presents a paradox: Existence of a Spaniard is asserted by means of the (potentially) nonspecific some Spaniard, and then that existence is denied. But (13a) is anomalous: Existence of a Spaniard is presupposed under the specific some Spaniard, and then that existence is denied. If the denial of existence produces a paradox, the NP is nonspecific; if it produces anomaly, the NP is interpreted as specific. Another characteristic of a specific NP concerns its relationship to the clausal verb (the verb in its immediate clause). Guéron (1981) argued that a nonreferential indefinite (an incomplete constituent, in her analysis) must be coindexed with V for interpretation. Observe, in that regard, that a nonspecific NP bears narrow scope with respect to the local verb, whereas a specific NP remains beyond V’s scope (cf. Enç 1991). Consider (14): (14) a. John wants a donkey. b. John needs a donkey. c. John must buy a book. The indefinite object NP in (14a-b) escapes from the scope of want, need, and buy under a specific interpretation but can remain in the scope of sentencelevel must in (14c). The evidence I have considered thus far regarding the behavior of a specific NP suggests that such an NP is distinguished by the fact that, like a name or a demonstrative, it has acquired a constant value. (Cf., in that regard, Zimmerman’s 1993 observation that only namelike quantifiers do not scopally

258

BECKY KENNEDY

interact.) It therefore cannot enter into scopal relations because it cannot covary in value with a QP, in the examples we have used; in those examples, moreover, its value is known to the speaker. Consider (15): (15) An NP X is specific iff there exists some process by which X acquires a specifiable value for the speaker. With respect to the examples we have considered, the process in (15) is a logophoric process by virtue of which X is associated with an individual that the speaker has in mind. Kuno (1987) defined logophoric as pertaining to the speaker/hearer and demonstrated extensively the importance of logophoricity in a range of language phenomena. Compare the specific NP defined in (15) with that defined in, for instance, Ludlow and Neale (1991), where specificity in indefinite descriptions is defined in terms of use rather than interpretation: The proposition in which the indefinite occurs involves a belief about a particular individual described by the indefinite; yet the speaker does not intend to communicate that his belief is about that individual. Specific use, under that analysis, implicates the speaker-hearer interaction; an NP can be strongly or weakly specific, depending on the degree to which the hearer can deduce that the speaker’s belief involves an individual. Specific use, in this sense, contrasts with the specific interpretation characterized in (15), where the communicative consequence is irrelevant. By (15), an NP is specific iff it holds a constant value for the speaker, regardless of the hearer’s deductions. Indeed, a specific indefinite necessarily has a covert value. 3.2 Enç: Specificity and Discourse Anaphoricity Certain analyses associate with specificity some degree of discourse anaphoricity, although not the anaphoricity of definiteness. Enç’s (1991) analysis typifies this approach; here I compare her notion of specificity to that proposed in (15). Enç’s specific NP, unlike the Specific NP of Fiengo and Higginbotham or the referential NP of Fodor and Sag, is not defined by its scopal behavior. Specificity, in Enç’s analysis, does not affect sentence truth conditions (as it would if specificity influenced the range of scopal interpretations and thus the range of conditions under which a sentence was true). Instead, Enç argued that specificity contributes to meaning by virtue of links entailed between NP

SPECIFIC NP IN SCOPE

259

denotation and the domain of discourse, through membership in or selection by an established subset of discourse referents. In this respect, Enç’s specific NP resembles Pesetsky’s (1987) D-linked (discourse-linked) NP — an NP that has a necessary link to an established discourse set. Definiteness, in Enç’s view, strongly links an NP to its discourse antecedent through identity, whereas specificity weakly links an NP to discourse referents by means of an associative function. Thus a definite NP is always specific, but an indefinite NP may be specific or nonspecific. All strong QPs are held to be specific because they quantify over a set of individuals already in the discourse domain, thus defining a subset of a discourse-anaphoric set (cf. Diesing’s presupposed set of entities). Enç’s partitioning of NP types as specific or nonspecific follows the lines drawn by case assignment in Turkish; her specific NPs are identified in Turkish by overt case morphology. Enç’s specificity distinction, whereby strong QPs, names, and definite descriptions are always specific and indefinites are ambiguous with respect to specificity, allowed her to account straightforwardly for the distribution of NPs in existential sentences: (16) a. b. c. d. e. f.

There is a cow in the backyard. There are many cows in the backyard. There are fifteen cows in the backyard. *There is Elsie in the backyard. *There is the cow in the backyard. *There is every cow in the backyard.

Enç held nonspecificity in an NP to be both necessary and sufficient for thereinsertion. She proposed that a nonspecific NP does not presuppose existence and therefore can appear in an existential construction, which is used to assert existence. Thus names, definites, and strong QPs, all specific under Enç’s analysis, are ungrammatical in (16d-f). Indefinites and cardinal NPs can appear in existential constructions (16a-c) on the nonspecific reading. We see later that my analysis and Enç’s share the assumption that a specific NP can be in scope. Note, however, that the two analyses describe different phenomena. I take Enç’s specific NP to be defined as in (17); compare my (15): (17) An NP X is specific iff there exists some function f that links the denotation of X to previously established referents in the domain of discourse.

260

BECKY KENNEDY (15) An NP X is specific iff there exists some process by which X acquires a specifiable value for the speaker.

The process referred to in (15) has been associated with speaker intent: The speaker has a particular value in mind under the specific interpretation in (3)(4) (although we later consider examples in which this is not the case). Yet that value need not (under (15)) have either a strong or a weak discourse link. (15) permits a specific indefinite to hold a covert value that is accessible only to the speaker and is not recoverable from discourse content; under (17), however, f must link a specific indefinite X to the discourse domain. To see the consequences of this distinction, consider the following: (18) A: The students in Introductory Syntax are reading Radford. B: Did a student from the class borrow the library’s copy? B’: Could I find out whether a student from the class borrowed the library’s copy? Under definition (17) of specificity, the indefinite NP a student from the class is interpreted in (18B) as a specific NP because it can be said that its referent is included in a familiar discourse set: the set introduced in (18A) as the students in Introductory Syntax. The denotation of that NP is therefore weakly linked to previously established referents in the discourse domain. Under (15), in contrast, the same NP cannot be specific. (15) requires that the NP acquire a specifiable value; yet any value in (18B) would be covert, accessible only to speaker B. Because the hearer A has no access to that value, he does not have enough information to answer the question (18B) under the specific interpretation (assuming (15)) for the subject NP. The question is therefore ill-formed under the specific interpretation; the only available interpretation under (15) is that in which the subject NP is nonspecific (denoting ‘some student or other’). Under either (17) or (15), however, the specific interpretation for a student from the class is permitted in (18B’). This is because the indefinite in (18B’) occurs within an indirect rather than a direct question, so that the hearer need not have access to the speaker’s covert value for the NP. The speaker may have in mind a particular student, but the hearer need not know the identity of that student in order to answer the matrix question (Could I find out...?). Thus in (18B) — but not in (18B’) — (17) and (15) support contrasting predictions regarding specificity.

SPECIFIC NP IN SCOPE

261

A second difference, suggested earlier, between Enç’s specificity and that defined by (15) is that Enç’s analysis did not include a nonspecific definite. This is because an NP denotation that is strongly linked to a discourse antecedent (via definiteness) is also weakly linked and thus specific. Under (15), however, Donnellan’s attributive definite descriptions are nonspecific NPs. Because (15) does not define a specific NP in terms of a discourse link, definiteness does not preclude nonspecificity. A third difference lies in the partitioning of NP types in terms of specificity. In section 2, we saw that Fodor and Sag adduced extractability from a scopal island and Fiengo and Higginbotham adduced opacity to extraction of an internal QP or WH-phrase as behavioral concomitants to specific interpretation. A strong QP will not show these characteristic behaviors. By (15), strong QPs are not specific; names are specific; and definites, indefinites, and cardinal NPs are ambiguous. Enç, however, analyzed strong QPs as specific, by the criterion (17). (Although it is only the universal QP in Turkish that requires accusative case, a morphological concomitant to specificity in Turkish, Enç generalized specificity to other strong QPs.) Enç generalized as well that only a nonspecific NP could appear in existential constructions; strong QPs, names, and definite NPs — her class of obligatorily specific NPs — were clearly excluded.3 I suggest, however, that what is crucial in the existential construction is not that an NP be nonspecific but instead that it be capable of either specific or nonspecific interpretation. An NP that undergoes there-insertion must be interpreted as nonspecific; thus, in a context like Fodor and Sag’s (5), it can take only narrow scope. Thus the indefinite in the there-context of (19a) does not show the island-escaping behavior of the indefinite in (19b) under a specific interpretation (cf. similar examples in Abusch 1994): (19) a. John overheard the rumor that there was a student of mine in the dean’s office. b. John overheard the rumor that a student of mine was in the dean’s office. Yet ambiguous indefinites can introduce individual discourse referents into the discourse for later reference, in the sense of Karttunen (1976): (20) There is a book on the table. It is about phonology. An ambiguous NP (unlike a name) can be used nonspecifically to establish

262

BECKY KENNEDY

existence before that existence is presupposed; it can also (unlike a strong QP) establish a discourse referent, signified by anaphoric it in (20). (See Ludlow and Neale 1991, for further discussion of indefinites and their anaphoric pronouns.) Enç also considered extractability of an internal NP: (21) a. *Who did John read every story about t ? b. Who did John make many movies about t ? Enç argued that strong QPs (as in (21a)) block WH-extraction, whereas weak NPs (as in (21b)) permit it, suggesting (by Fiengo and Higginbotham’s criterion) that strong QPs are specific and weak QPs, nonspecific. Consider, however, the following counterexamples: (22) a. Who has John read every single story about t ? b. Who has John saved every picture of t ? c. I’ve saved every paper from every student I’ve ever had. The grammaticality of the WH extraction in (22a-b) and the natural widescope reading of every student I’ve ever had in (22c) indicate that strongly quantified picture NPs need not be islands for extraction. (It has been pointed out to me that the universally quantified NPs in (22) from which extraction occurs all contain a focus.) In sum, (15) and (17) make different predictions, reflecting a basic difference not only with respect to the defining features of specificity but also with regard to the phenomenon under investigation. Recall that a similar contrast was observed in section 3.1 with respect to Ludlow and Neale’s conception of specificity, insofar as their defining feature for specificity (or specific use) was that the speaker had singular grounds, but no intention of communicating a singular proposition. It is unclear, however, how speaker’s grounds would render a sentence like (18B) unacceptable under the specific interpretation. In the approach presented here, speaker intent is crucial insofar as it is infelicitous to utter (18B) intending specific interpretation for the indefinite NP, because the question expressed will be unanswerable. Under Enç’s definition, the indefinite in (18B) is not specific: Discourse linking is critical, with shared awareness between speaker and hearer regarding the discourse status of a specific NP’s referent. In section 2, I noted that arguments for a semantic ambiguity approach must adduce semantic distinctions that could not otherwise be explained. We

SPECIFIC NP IN SCOPE

263

saw that the truth conditions of a sentence were affected by specificity, insofar as specificity assignment played a role in scopal interpretation. In the next section, however, I adduce certain problems faced by the ambiguity hypothesis. I suggest (following Enç and others) that specificity can occur in scope, with the resulting dissociation of specificity and referentiality permitting us to retain the insights of Fodor and Sag.

4.

Specific NP in Scope

4.1 Donkey Sentences: A Problem for the Ambiguity Hypothesis Although the referential/quantificational ambiguity hypothesis allows us to account for distinctions in NP behavior in terms of an intuitively meaningful interpretative distinction, Heim (1982) outlined a problem presented by data like the following: (23) a. Every man who owns a donkeyi beats iti. b. Every man who loves his motheri remembers heri birthday. The antecedent NPs in the donkey sentences (23a-b) take narrow scope with respect to every and have nonreferential interpretation, in the sense that the speaker cannot be said to have an individual in mind. Under Fodor and Sag’s analysis, an NP in scope has nonreferential interpretation and is thus a quantificational NP. The pronouns it/her in (23a-b), moreover, appear to have bound-variable interpretation, as predicted under the ambiguity hypothesis, because they are pronouns coindexed with a QP. The pronouns, however, lie beyond the c-command syntactic domains of their quantified antecedent NPs, contra Reinhart’s (1983) restriction on bound-variable anaphora. A referential interpretation of each antecedent would preclude its taking narrow scope with respect to every man; yet a quantificational analysis cannot account for the anaphora: A pronoun with bound-variable interpretation lies beyond the binding domain of its antecedent QP. Heim’s solution (see also Kamp 1981) is that definite and indefinite NPs like a donkey and his mother are not quantified expressions but instead open formulae including a free variable and a predicate; a donkey translates as donkey(x). An indefinite or novel definite (a definite with no textual antecedent) like his mother in (23b) becomes bound by a higher quantifier in order for

264

BECKY KENNEDY

its free variable to acquire a value assignment. Reinhart (1987) observed further that this analysis applies not only to definites and indefinites: Any weak NP can become bound by a higher quantifier in a donkey sentence, allowing for donkey anaphora (although note that this analysis does not hold for no, classified as weak by Barwise and Cooper 1981). The weak/strong distinction in QPs separates cardinal (weak) determiners like many from the proportional (strong) determiners like every and most. Thus we can get donkey anaphora in (24a) but not in (24b):4 (24) a. Every man who despises many donkeysi beats themi. b. Every man who despises most donkeysi beats themi. Fodor and Sag based their ambiguity hypothesis on the island-escaping behavior of indefinites and suggested an extension of their analysis to the class of cardinal NPs. We have just seen, however, that these NPs share certain characteristics with both referential and quantitative NPs when in donkey contexts like (23a-b). The ambiguity hypothesis explains the behavior of a cardinal NP in an example like (5); yet the behavior of cardinal NPs in donkey contexts, where referential- and quantificational-type behaviors cooccur, confounds that hypothesis. Fodor and Sag argued that the indefinite in (5) can escape from the scopal island precisely because it is nonquantificational; such an indefinite, they argued, cannot be in scope in (25): (25) Each teacher overheard the rumor that a student of mine had been called before the dean. Fodor and Sag observed that a student of mine cannot both escape from the complex-NP island in (25) and take narrow scope with respect to each teacher. That is, (25) does not permit the interpretation under which each teacher overheard the rumor about a different, specific student. In that case, the indefinite would show both referential and quantificational characteristics. The interpretative difference between the indefinite a student of mine in (25) and the indefinite a donkey in (23a) is the difference that motivated the ambiguity hypothesis for indefinites: A student of mine in (25) does not (they argued) interact scopally under the referential (island-escaping) interpretation, whereas a donkey does interact scopally in (23a). If we assume that susceptibility to scopal interpretation is incompatible with referentiality, we can use that susceptibility to distinguish referential and quantificational interpretation in cardinal NPs.

SPECIFIC NP IN SCOPE

265

But we have also seen that the ambiguity hypothesis fails to handle antecedent NPs in donkey anaphora. Here the same NPs show nonquantificational behavior in examples like (23a); yet they show QP-like behavior in their narrow scope. Heim’s solution to this puzzle was that donkey NPs are neither referential nor quantificational; yet where does this class of counterexamples to the ambiguity hypothesis leave specificity?5 Can we account for an intuition of specificity, with its links to referential-like behavior, and not assume a referential/quantificational ambiguity? 4.2 Higginbotham’s Observation There are other problems associated with the identification of referentiality and specificity. Higginbotham (1987) pointed out that the assumption that a specific NP always designates an individual that the speaker has in mind is undermined by (26): (26) a. I met with a certain student of mine today. b. George said that he met with a certain student of his today. Higginbotham observed that his friend George might say (26a) to him, and then he himself might report (26b) to a third party. Both a certain student of mine in (26a) and a certain student of his in (26b) have a specific connotation. Yet the speaker in (26b) cannot be said to have a certain student in mind; he may not even know the student in question. Thus a speaker cannot be said to have in mind a referent that is unknown to him but can specify without referring directly (cf. also Hintikka 1986). 4.3 Intermediate Readings in the Fodor-Sag Structures Another problem for Fodor and Sag’s ambiguity analysis involves the possibility of intermediate readings for sentences like (25). Examples that permit readings under which an indefinite escapes an island but does not take maximal scope have appeared in Abusch (1994), Kennedy (1993), King (1988), Ludlow and Neale (1991), Ruys (1992); see also Kratzer’s example cited in Diesing (1992). Thus consider the following: (27) a. Twice, John overheard the rumor that some student was cheating.

266

BECKY KENNEDY b. Each teacher overheard the rumor that a different student had been called before the dean.

(27a-b), like (25), permit the narrow-scope reading (‘some student or other’) and the widest-scope reading (‘a particular student’) for the indefinite embedded in the complex-NP island. Both examples, however, also permit the intermediate reading that is excluded for the indefinite in (25). Under this reading, the indefinite escapes its island but remains within the scope of a higher quantified phrase (twice in (27a), each teacher in (27b)). (28), likewise, is one of many examples provided in Abusch (1994): (28) At most four committee members resisted a proposal that a candidate be turned down. (28) allows the intermediate-scope reading For at most four committee members, there is a candidate x such that the member resisted the proposal that x be turned down. Fodor and Sag acknowledged that examples like (25) might not be convincing for some speakers and suggested other types of islands: Thus the intermediate reading is unavailable in their sentential-subject-island and ifclause-island examples (29)-(30): (29) Each teacher thinks that for a student I know to be called before the dean would be preposterous. (30) If a student in the syntax class cheats on the exam, every professor will be fired. In (29), each teacher cannot have a different but particular student in mind; (30) cannot mean that for each professor there is a different but specific student who will cause the professor to be fired. Consider, however, the following examples containing sentential-subject or if-clause islands in which the intermediate reading is permitted: (31) Most teachers have at some point concluded that for some poor student to graduate would constitute a miracle. (32) If some German family had not provided an attic or basement hideout, most survivors would have perished in the Holocaust. (31) permits the reading under which each teacher has reached a pessimistic conclusion about a different, particular student; (32) permits the reading

SPECIFIC NP IN SCOPE

267

under which, for each survivor, a different, particular German family provided a hideout. Just as in the donkey sentences, then, an NP in examples (27), (28), (31), and (32) shows behaviors characteristic of both a QP (it interacts scopally) and a referential NP (it escapes a scopal island). Why should the intermediate scope interpretation hold in some instances and not in others? Lexical and pragmatic explanations can be adduced: Each and different in (27b) enhance a distributive reading, such that each teacher distributes over a different student; twice as an event quantifier in (27a) likewise enhances the distributive interpretation, as does at some point in (31). Diesing (1992) noted Kratzer’s observation that the intermediate reading can be pragmatically forced in certain contexts: (33) Each writer overheard the rumor that she didn’t write a book she wrote. A book she wrote must have scope beyond the complex rumor NP; yet the indefinite in (33) is in the scope of each writer. I have also been advised by a reader that the reference to the speaker in (25) makes the specific (referential, apparent widest scope) interpretation, under which the speaker has an individual in mind, most prominent; thus the following example in which the intermediate reading is available: (34) Each teacher was confronted with the claim that a student had been called before the dean. In Fodor and Sag’s examples, on the other hand, the pairing of teachers and students that would lead to a distributive reading is discouraged. Such a pairing is pragmatically less likely in (25) than in (27b); in (29), likewise, a pairing is pragmatically discouraged, in part because a student I know is most likely to refer to a particular (single) student that the speaker has in mind. In (32), an attic or a basement ensures that more than one instance of hiding is involved; thus the interpretation under which most survivors distributes over some German family. The if-clause structure in (30) most strongly discourages the intermediate scopal reading for the embedded indefinite; Ludlow and Neale (1991) held the consequent clause to be a scope island for the universal QP in such an example, precluding the intermediate scope reading for the indefinite in the antecedent. I suggest, however, that the intermediate reading is absent because a conditional generally discourages an interpretation of

268

BECKY KENNEDY

distribution within hypothetical events. The counterfactual in (32) suggests that the necessary distribution was in discrete events that had in fact occurred and supports the intermediate reading. In sum, a number of scholars have observed that the intermediate scopal interpretation that Fodor and Sag argued against does indeed occur. Although the factors that enhance it are lexical and pragmatic, those that block it in Fodor and Sag’s examples are pragmatic as well. The vulnerability of that reading to pragmatic constraints could mean that the intermediate scope interpretation is possible only when pragmatics forces us to violate grammar. On the other hand, it could mean that the intermediate scope interpretation is in principle grammatical; but given the unlikelihood that a wide-scope QP will distribute over an NP that is contained in a scopal island, pragmatics must support such an arrangement. I assume the second position, for the following reason. If an NP is either referential or quantificational, there is no middle ground: Either it is or it is not immune to scopal interaction. If it is immune to scopal interaction and thus escapes islands, it cannot be in scope; and pragmatic support cannot facilitate an intermediate scopal reading. If it is not immune to scopal interaction, it should not show the distinctive islandescaping behavior, especially in the case of an absolute island like that in (32). But if the island-escaping NP is not necessarily referential but can instead be a specific NP in scope, the significant role of pragmatics in scopal possibilities is expected. 4.4 Do We Need Specificity? In addition to her numerous intermediate-scope examples, Abusch (1994) provided a semantics that permits a correct derivation for that reading. Her position, like that of Fodor and Sag, was that indefinite descriptions differ from genuine quantifiers in their scopal properties: Indefinites escape islands, whereas quantifiers do not. Unlike Fodor and Sag, however, she presented a unified analysis of indefinites in the Lewis-Kamp-Heim tradition, as free variables without quantificational force that get bound by local quantifiers. Crucially, her analysis includes a mechanism whereby the restriction on an indefinite gets carried up in the course of interpretation and is interpreted at the level at which the variable is quantified. A quantifier’s scope is indicated by LF-movement; subjacency will then predict a sensitivity to islands in that movement and in quantifier scope. An indefinite, in contrast, is given scope

SPECIFIC NP IN SCOPE

269

by indexation to a quantifier; because movement is not involved, the scope of an indefinite is not constrained by subjacency. In this way, Abusch related scopal differences between quantifiers and indefinites to a difference in the source of the quantificational force: the quantificational determiner, in the case of a quantifier; a coindexed existential quantifier, in the case of the indefinite. Demonstrating that the intermediate scopal reading did in fact occur, Abusch presented data and an analysis that argue not only against a referential/quantificational ambiguity analysis of indefinites but also against the invocation of any ambiguity analysis to explain island-escaping behavior of indefinites. Heim, we saw in section 4.1, had argued that an indefinite is capable of both referential and quantifierlike behavior in a donkey context; and Abusch extended Heim’s analysis of indefinites by arguing that the contrast between quantifiers and indefinites in island sensitivity can be explained under Heim’s account of indefinites. Abusch reserved judgment on the existence of referential indefinites, because of one sentence type: that represented by Fodor and Sag’s (30), with a sentence-initial if-clause that functions as an island for an embedded indefinite. My example (32), however, does permit the intermediate interpretation, suggesting that the ambiguity hypothesis is no longer necessary. But what about the intuition of specificity? Abusch’s approach is reminiscent of that of Karttunen (1976), in that what might be perceived as a semantic effect of referentiality or specificity was analyzed instead as a consequence of the operations of a scope mechanism. An unresolved issue, however, is that of the distinction made by Fodor and Sag between the maximal scope reading and the referential reading of an indefinite. Ludlow and Neale (1991) discussed, in that regard, the following example, remarking that “the wide scope reading... cannot be emulated by a reading in which the indefinite is treated referentially” (Ludlow and Neale 1991: 185): (35) John thinks a student of mine cheated. The authors adduced a case in which the speaker is told that a student of his is such that John, not knowing that he is the speaker’s student, believes that he cheated. The speaker, not apprised of the identity of the student, could utter (35) truthfully on the wide scope reading; yet a student of mine cannot be interpreted (or is not being used) referentially. Ludlow and Neale distinguished between a general proposition, which involves the wide scope read-

270

BECKY KENNEDY

ing, and a singular proposition, which involves the referential interpretation. We saw earlier that it is the referential rather than the wide scope reading that is available in an example like (5), under the island-escaping interpretation for the indefinite. The distinction between the referential reading and the maximal scope reading is paralleled by the distinction between an in-scope, nonspecific interpretation and an in-scope, specific interpretation in an example like (36): (36) Every profession rewards those who acquire some skills. (36) permits an in-scope, nonspecific interpretation for some skills, under which the narrow-scope indefinite signifies ‘some skills or other’. But under an intermediate, island-escaping interpretation, some skills must be specific: For every profession there are certain relevant skills. The island-escaping indefinite, in other words, is always specific; and this is in fact the result predicted by Fodor and Sag’s analysis, although those authors argued that the intermediate interpretation is simply unavailable. Fodor and Sag attributed an apparent maximal scope reading of an islandescaping indefinite to absence of scopal interaction, so that the identity of truth conditions between the maximal scope reading and the referential reading belies a semantic difference. Abusch did not need an ambiguity analysis, arguing that it was the source of their quantificational force rather than an ability to assume referential interpretation that permitted indefinites to escape islands. The contrast between maximal scope and referential interpretation discussed by Ludlow and Neale, however, argues for the existence of a distinct referential interpretation that is associated with island-escaping behavior in an indefinite. (36), moreover, shows that a similarly distinct specific interpretation is associated with the intermediate, island-escaping interpretation for an indefinite. I therefore take the position that cardinal NPs are ambiguous with respect to specificity. I argue against the association between referentiality and specificity and in favor of the analysis of referentiality as a subcase of specificity, such that a specific NP can be in scope. Thus an island-escaping NP has referential interpretation when it takes widest scope but is a specific NP in scope in (37), under the intermediate interpretation: (37) a. Every businessman who hears the rumor that a competitori is opening up shop tries to convince himi to leave.

SPECIFIC NP IN SCOPE

271

b. Everyone who hears about Bill dating a womani tries to warn heri about him. Compare these examples with (38), from May (1985): (38) Your shooting at every owner of a donkeyi frightened iti. May adduced (38) to show that the higher quantifier binding the donkey antecedent must c-command the donkey pronoun in order for a donkey interpretation to take place. The gerund in (38) acts as a scopal island for the universally quantified phrase. The narrow scope indefinites in (37a) and (37b) are embedded in a complex-NP island and a gerund island; yet in both cases the indefinite is able to escape the island in order to antecede the donkey pronoun occurring later in the sentence. (Recall from section 4.1 that under the Heim 1982, or the Reinhart 1987, analysis of donkey anaphora, the donkey antecedent must escape from the island containing it in order to be bound by the universal quantifier in (37a-b).) The indefinites in (37a-b) escape their islands yet remain in scope. 4.5 Hintikka’s (1986) Epistemic Operator (15) suggests that a specific NP has acquired a specifiable value for the speaker; when that NP is also referential, this occurs via a logophoric process that associates the specific NP with an individual that the speaker has in mind. What of the specific NP in scope? Hintikka (1986) argued (contra Hornstein 1984) that a certain need not take wider scope than other quantifiers. Thus the indefinite in (39) can take narrow scope: (39) Each husband had forgotten a certain date — his wife’s birthday. Hintikka characterized a certain as a quantifier that does not necessarily take priority over other quantifiers but that does have priority over epistemic operators. His sentence (40a) is thus interpreted as in (40b): (40) a. I know that Richard is dating a certain woman. b. ( x)(x is a woman & KI (Richard is dating x)) Read KI in (40b) as ‘I know that’. Hintikka suggested further that a sentence like (41) has an implicit epistemic operator, such that (41) is also interpreted as in (40b):

272

BECKY KENNEDY (41) Richard is dating a certain woman.

If a certain takes logical priority over an epistemic operator but need not take scope over another quantifier, the scopal ordering in an example like his (42) is not linear: (42) I know that every true Englishman adores a certain woman. A certain in (42) bears scope over the epistemic operator but can remain in the scope of the universal quantifier, as in Hintikka’s linear higher-order formulation: (43) ( f ) KI ( y) (y is a true Englishman y adores f(y)). The function quantifier in (43) has priority over the epistemic quantifier, but the relative scopal positions of a certain and the universal quantifier in (42) are captured in terms of a function relating the two quantified terms. We can schematize the orderings for a certain out of scope and in scope as follows: (44) a certain — K (45) f — K — Q Notice that a certain has scope over the epistemic operator in (44), whereas in (45) a certain is in the scope of a quantifier that is in turn in the scope of the epistemic operator. What maintains the priority of a certain over the epistemic operator in (45) is Hintikka’s function quantifier, which has scope over K and fixes the value of a certain relative to that of Q. Let us now see whether Hintikka’s treatment of a certain can be extended to the account of a specific NP suggested in (15). His description of the force of a certain as indicating an individual whose identity is known but not necessarily divulged is similar to the characterization here of a specific NP as identifiable by the speaker but not necessarily by the hearer — and not necessarily linked to previously established discourse referents. The logical priority of a specific NP over an epistemic operator explains the apparent widest scope interpretation of a specific NP out of scope; in addition, it helps us understand the intermediate reading for sentences like (28) or (37a-b). Hintikka’s formulation for this kind of example, illustrated in (43), explains the apparent paradox under which an island-escaping NP is putatively inert with respect to scope, yet is in the scope of a universal quantifier. In (43), the

SPECIFIC NP IN SCOPE

273

function whereby the value of the indefinite is subordinated to that of the universal quantifier is expressed as the verifying value of a function variable bound by a wide scope existential. Within the scope of that existential, however, lies the epistemic operator with respect to which the indefinite apparently takes logical priority. If the specific interpretation of an indefinite can be attributed to its priority with respect to an epistemic operator, then that interpretation should hold within the scope of the epistemic operator but not beyond the scope of the existential function quantifier — not beyond the scope of the universal quantifier in whose scope the indefinite lies. Thus the indefinite can escape an island that lies in the scope of the universal quantifier, as in (37a-b). The process by which an NP acquires a specifiable value for the speaker has been described as logophoric, and its consequence is recorded in schemata (44)-(45). An NP is accorded priority over an operator K — a logophoric operator indexed to the speaker. A specific NP out of scope is associated with an individual the speaker has in mind; an NP in scope gains priority over an operator K, in turn subordinated to f relating NP and Q. (45) schematizes an additional aspect of specificity: the boundedness of the domain of specificity. Note a contrast between (37a-b) and (38): The indefinites in (37a-b) are able to escape an island containing them in order to antecede a donkey pronoun — as long as that island is in the scope of Q. In (38), in contrast, the indefinite cannot remain in the scope of Q and at the same time escape from the gerund island in order to antecede the donkey pronoun. The domain within which an NP in the scope of Q can display the island-escaping behavior typical of a specific NP is delimited by Q. (45) represents the priority of NP over K as the prioritization of f over K. NP may escape an island in order to take priority over K but remains, as a function of Q, in the scope of Q; this state of affairs is represented linearly by subordinating K to f. In this section, we have looked at specific NPs in scope, focusing on the distinction between referentiality and specificity and considering the availability of the intermediate reading ruled out by Fodor and Sag. Abusch analyzed the semantics of that reading and concluded that an ambiguity hypothesis was unnecessary, yet I adduced justification for maintaining the semantic notion of specificity. Hintikka’s analysis of a certain as logically prior to an epistemic operator illuminated the behavior of a specific NP in scope. We now consider further consequences of the analysis of specificity as occurring both in and out of scope, as well as bounded within a domain.

274 5.

BECKY KENNEDY Domain of Specificity and the Interpretation of Questions

In this section, we consider examples in which the notion of domain of specificity permits an explanation of certain scopal interactions that cannot be explained in terms of the quantificational/referential distinction. I adduce a number of examples involving WH-movement, assuming, on the basis of data like the following, that a WH-term is ambiguous with respect to specificity: (46) A: Who did John marry? B: a. John married Leona. b. John married a lawyer. Two possible answers for (46A) — one that provides a value for the questioned term, and one that characterizes the questioned term without supplying a value — suggest that who can be understood as either specific or nonspecific. Here, again, I diverge from the D-linking approach, under which the relevant distinction is not between two interpretations for who but between a term like who, which is not D-linked, and a term like which boy, which is. (Cf. Bolinger 1978, for early observation of D-linking phenomena; the reader is referred as well to Comorovski 1988, who argued that WH-words like who can be D-linked. Recall that Pesetsky’s 1987, D-linked NP — Enç’s specific NP — is weakly linked to an established discourse set.) Under the assumption that a WH-term can be specific, WH-movement examples are of particular interest because the island-escaping behavior of a WH-NP has visible consequences at S-structure. Cinque (1990) followed Pesetsky in his analysis of ambiguity in WHterms. Pesetsky argued that a distinction in D-linking accounts for a difference in the behavior of in situ WH-phrases, proposing that only non-D-linked WH-phrases in situ will move at LF, whereas D-linked WH-phrases in situ are interpreted via binding by a question operator. Only a non-D-linked WHphrase should therefore exhibit sensitivity to constraints on movement. Cinque extended this analysis to the case of long WH-extraction across a weak island, arguing that such extraction can occur only when the extracted term is, in Pesetsky’s sense, D-linked or referential. In Cinque’s account, this is because only a phrase bearing a referential theta-role can be involved in binding relations. The original trace of a referentially indexed NP can be properly governed via long-distance binding, but a chain of antecedentgovernment relations formed by successive-cyclic movement is the only option for a nonreferential NP.

SPECIFIC NP IN SCOPE

275

Cinque’s distinction between a referential and a nonreferential WHword resembles the Fodor-Sag distinction. He argued that a WH-word in the scope of a quantifier must be interpreted nonreferentially and utilized an example of WH-extraction over an island: (47) (?) Quanti pazienti non ricordi se lui avesse visitato? How many patients don’t you remember whether he had visited?’ Cinque asserted that (47) permits the extraction of the WH-phrase from the WH-island because the WH-phrase does not interact scopally with a QP and can therefore (as in Fodor and Sag’s analysis) bear referential interpretation. In (25), this means that the indefinite cannot interact scopally with the universal QP. In an example like (47), the WH-phrase can be extracted only if it does not interact scopally with a QP. Consider, now, some other examples involving weak island violations: (27) a. Twice, John overheard the rumor that some student was cheating. (48) Each man braced himself before meeting some woman. (49) Each man wondered whether John had invited some woman. (50) Who did each man invite? (51) Who did John brace himself before meeting? (52) Who did each man brace himself before meeting? (53) Who did each man wonder whether John had invited? We noted earlier an intermediate scope reading for (27a); now consider (48)(49). Under one interpretation for these examples, the indefinite NP within the before-phrase island or the WH-island has the widest-scope interpretation: There is one woman involved. Under a second interpretation, some woman has narrowest scope, signifying ‘some woman or other’. But, as in the case of (27a), there is an intermediate reading available for (48)-(49), under which each man in question has braced himself before meeting, or wonders whether John invited, a different but specific woman: The specific indefinite is in scope. Consider now the questions (50)-(53). Note first the family-of-questions interpretation available for (50), under which the universal QP has wider scope than WH, with a scopal interaction. (54) represents a possible answer, under this interpretation:

276

BECKY KENNEDY (54) John invited Mary, Bill invited Sue, and Joe invited Ann.

(51), unlike (50), does not permit the family-of-questions interpretation because it contains no QP with which the WH-term can interact scopally. Because the WH-phrase is not in the scope of a QP, it is interpreted referentially. And, as was the case for (47), Cinque’s analysis predicts possible WHmovement across the island in (51). In (52)-(53), as in (51), a WH-term has been extracted from an island; here, however, as in (50), the WH-term co-occurs with a QP. Under Cinque’s analysis, questions like (52)-(53) should not permit a family-of-questions interpretation: The WH-term would then need to interact with the QP and would be necessarily nonreferential, but Cinque asserted that a nonreferential WH-phrase cannot be extracted from an island. Thus (55): (55) ?Quanti pazienti ti chiedevi come ognuno dei medici riuscisse a visitare in un’ora? ‘How many patients did you wonder how every one of the doctors could visit in one hour?’ Cinque pointed out that (55), if it is acceptable, cannot bear a family-ofquestions interpretation because the WH-phrase in (55) must be interpreted referentially in order to escape the island containing it and therefore cannot interact scopally with the QP. (52) and (53), like (55), involve WH-phrases that have been extracted from weak islands; in both examples, interaction with the QP will putatively render the WH-phrase nonreferential, in Cinque’s sense. Yet (52) and (53), insofar as they are acceptable, do allow the family-of-questions interpretation. Under that interpretation, the questioner is asking about the particular individual that each person meets/wonders about. Thus (56) is a possible answer to (52): (56) John braced himself before meeting Abigail, Bill braced himself before meeting Hank, and Ralph braced himself before meeting Jack. How can we reconcile Cinque’s observations regarding (55) with the observation that a family-of-questions interpretation is available in (52)-(53)? In (52)-(53), as in (55), the WH-phrase fronted across an island should be referential; yet (52)-(53) permit a family-of-questions interpretation as if the WH-phrase were nonreferential.

SPECIFIC NP IN SCOPE

277

To see the difference between (55) and (52)-(53), compare (52)-(53) to the following: (57) Who did John brace himself when each man met? (58) Who did John wonder whether each man had invited? Notice that questions like (57)-(58), insofar as they are possible, do not allow the family-of-questions interpretation; thus (57), for instance, permits only an answer in which a single NP identifies who, as in (59): (59) John braced himself when each man met Abigail. Observe now that (57)-(58) resemble (55) rather than (52)-(53) in the following respect: The QP each man includes in its scope the extraction island in (52)-(53). In (55) and (57)-(58), however, the extraction island contains the relevant QP. Consider now the following observation regarding the domain of specificity for an NP: (60) A specific NP in the scope of Q is specific relative to the scopal domain of Q. The observation articulated in (60) was discussed in section 4.5 with reference to Hintikka’s model of the scopal relations between a certain, Q, and the epistemic operator K. We saw its effects instantiated in (37a-b) and are now considering another type of example in which the domain of specificity for an NP is delimited by the quantifier with respect to which the NP is in scope. In (52)-(53), who, bearing narrow scope with respect to the QP under the familyof-questions interpretation, bears specific interpretation relative to the matrix S, representing in those examples the scopal domain of Q. Who exhibits the island-escaping behavior characteristic of a specific NP in relation to material within the matrix S, the scopal domain of Q. In (55) and (57)-(58), on the other hand, if the WH-phrase bears narrow scope with respect to the QP (under the family-of-questions interpretation), it can bear specific interpretation only within the embedded S, the scopal domain of Q. The scopal domain of Q in each case is restricted by the island that dominates a QP. (Recall the observation with respect to (6) that a strong QP like each man respects scopal islands.) To summarize, I have discussed an effect of the phenomenon of domain of specificity, proposing that an NP in scope can bear specific interpretation only with respect to elements in the domain of the quantifier with respect to

278

BECKY KENNEDY

which it bears narrow scope. This phenomenon can be accounted for in terms of schema (45). Distinguishing the domain of specificity for WH-NPs in questions has permitted an account for the possibility of family-of-questions interpretation in certain examples and its impossibility in others. The WHquestion data are consistent as well with data presented by Abusch (1994) and other that demonstrate that an intermediate scopal reading is possible in the Fodor-and-Sag-type example: Thus an NP exhibiting island-escaping behaviors characteristic of specificity can also be in scope. A referential/quantificational ambiguity hypothesis accounts for the unavailability of the family-of-questions interpretation in Cinque’s (55) but does not explain why a family-of-questions interpretation is possible in questions like (52)-(53). It does not account for the contrast between (55) and (57)-(58), on the one hand, and (52)-(53), on the other, we need the additional notion of boundedness of island-escaping behavior, explained here in terms of the domain of specificity. (61) schematizes the relative orderings of WH, the island, and Q. In terms of a schema like (45), who in the scope of Q can be fronted syntactically beyond Q to take scope over an epistemic operator K and yet remain in the scope of Q by virtue of a higher-function quantifier that relates the value of who to that of Q. (61a) represents the relative positions of f, K, and Q. (61b) displays the position of the island in examples (52)-(53); (61c) shows the island’s position in (57)-(58): (61) a. f — K — Q b. f — K — Q — island c. K — island — Q Notice that a WH-term that escapes the island in schema (61b) can remain in the scope of Q. In (61c), in contrast, a WH-term that escapes its island to assume priority over K will no longer remain in the scope of Q with its value relativized to that of Q via a function f. Thus (57) and (58) do not allow family-of-questions interpretations but instead permit as an answer only a single NP identifying who. Again we see referentiality as a subcase of specificity: Who in these examples must be referential.

SPECIFIC NP IN SCOPE 6.

279

Indirect Specificity in Indirect Discourse Contexts

In this section I reconsider Higginbotham’s (1987) example (26): (26) a. I met with a certain student of mine today. b. George said that he met with a certain student of his today. In section 4.2, I drew on Higginbotham’s observations regarding (26) to argue for the dissociation of referentiality and specificity. I now explore the question whether the specificity in a certain student of his in (26b) is explained in the same way as is the specificity of NPs in scope, which likewise do not denote a referent that the speaker has in mind. If we analyzed say in (26b) as a discourse operator in whose scope a certain student of his lies, we could say that the indefinite NP in (26b) is in scope but specific within the domain of the discourse operator. This analysis would account for the observation that in (26b), there are two domains in which the indefinite can be interpreted as specific, yielding interpretations (62a) and (62b): (62) a. For a certain student of his, George said that he met with that student today. b. George said that, for a certain student of his, he met with that student today. In (62a), the speaker has a referent in mind; in (62b), George does. If a certain student of his in (26b) takes wider scope than the discourse operator, we have interpretation (62a), under which the indefinite denotes a referent in the mind of the primary speaker (the speaker of the utterance). If, on the other hand, a certain student of his is within the scope of the discourse operator, that indefinite can be specific only within the scopal domain of say, yielding (62b), in which the indefinite denotes a referent in the mind of the secondary speaker (George) but not the primary speaker. If this analysis were correct, it should be generally true that a certain N in an indirect discourse environment can be specific with respect to either the entire utterance or the discourse operator. Now consider (63a-e): (63) a. b. c. d. e.

George implied that he had insulted a certain student. George asked whether he had insulted a certain student. George didn’t say that he had insulted a certain student. George believed that he had insulted a certain student. George wondered whether he had insulted a certain student.

280

BECKY KENNEDY

Notice that in (63a-e), a certain student can denote a referent that the primary speaker has in mind. But in (63a-c), the indefinite cannot denote a referent that the secondary speaker has in mind, although this second interpretation is available in (63d-e). These facts undermine an analysis of an indefinite in an indirect discourse environment as associated with a domain of specificity delimited by an operator, insofar as the specificity effect is inconsistent across discourse operators and does not survive negation.6 I suggest that the specificity effect is generally interpreted as an effect of primary speaker intent. Certain verbs of indirect discourse, however, allow the primary speaker to communicate the secondary speaker’s intent regarding specificity, so that an NP in the verb’s complement is indirectly specific. Wonder (in (63e) behaves like believe in (63d), a verb of indirect discourse long known to introduce an effect of referential opacity, whereby the reference of an NP in its complement can reflect either the perception of the primary speaker or the belief of the secondary speaker (cf. Frege 1952; Quine 1960). It is natural that such a verb can be used to communicate secondary speaker intent regarding not only reference but also specificity. In (63a), however, imply necessarily represents only the interpretation of the primary speaker; we would therefore not expect that an effect of secondary speaker use would be communicated in the complement to imply. The reader is directed to Kuno (1972, 1987) for indepth analysis of the relationship between an indirect discourse structure and its underlying direct-discourse representation. Indirect specificity of the indefinite is blocked as well in the indirect question (63b). To see why this is so, consider the following direct question: (64) Did I insult a student? Notice that (64), like (18B), is an acceptable question only if a student is nonspecific. This is because a specific indefinite has a covert value: In an example like (64), in which the indefinite is not in scope, its value is known to the speaker but not to the hearer. The hearer cannot assess the truth value of the proposition I insulted a student, under the specific interpretation, without knowing the intended value of a student and therefore cannot answer the yesno question (64). An indirect question like (63b), then, cannot contain an indirectly specific indefinite NP, because the secondary speaker would have a covert value for the specific NP in an associated direct question like (64), and that question would be unanswerable. The only felicitous specific interpretation in (63b) is the direct specificity interpretation under which the primary

SPECIFIC NP IN SCOPE

281

speaker has a covert value for the specific indefinite; the covert value is then relevant not to a question but to the value of a proposition (the entire S). The negation of the matrix verb, as in (63c) (where the matrix verb itself is perceived as the target of Neg; see note 6), blocks indirect specificity as well. This result can be explained on pragmatic grounds: If the complement to said in (63c) is a string that was not uttered, there is no secondary speaker intent to report. As further evidence for the association between indirect specificity and secondary speaker intent, consider the existential presupposition that was argued in section 3 to accompany specificity. Recall that if the occurrence of a nonspecific NP was accompanied by a denial of the existence of a referent for that NP, a paradox resulted. But if a denial of the existence of a referent followed the occurrence of a specific NP, anomaly resulted. Now consider (65): (65) John believed that some Spaniard seemed critical of his position. Recall that the subject NP of seemed has a specific interpretation (cf. (12ab)).7 Some Spaniard in (65) can be interpreted as directly specific, denoting a referent in the mind of the primary speaker, or as indirectly specific, denoting a referent in the mind of the secondary speaker (John). Now observe the cancelation of the existential presupposition: (66) John believed that some Spaniard seemed critical of his position; yet there was no Spaniard present. If some Spaniard in (66) is directly specific, the discourse is anomalous: The speaker makes a statement involving a presupposition of existence that is canceled. But if some Spaniard is indirectly specific, the cancelation of the existential presupposition has no effect on the well-formedness of the sentence. Does this mean that an indirectly specific NP, unlike its directly specific counterpart, does not presuppose existence? I suggest that the indirectly specific NP bears the same existential presupposition as does a directly specific NP. But when a primary speaker is reporting on the speech act of a secondary speaker, he makes no commitment to the assumptions of the secondary speaker and can, without inconsistency, continue in his text to make this lack of commitment overt. The existential presupposition is a presupposition of the secondary speaker. We can again utilize Hintikka’s epistemic operator K: Recall that a

282

BECKY KENNEDY

directly specific NP was analyzed as taking wide scope over an epistemic operator. I characterized K as a logophoric operator, indexed to the speaker. Suppose that for certain verbs of indirect discourse, K can be indexed to either the primary or the secondary speaker. This means that (63d) will be represented as in (67a) or (67b): (67) a. ( x)(x is a student & KI (George believed that he had insulted x)). b. George believed that ( x)(x is a student & KG (he had insulted x)). (63b), in contrast, permits K to be indexed only to the primary speaker. Thus (68a) is a possible interpretation of (63b), but (68b) is not: (68) a. ( x)(x is a student & KI(George asked whether he had insulted x)). b. George asked whether ( x)(x is a student & KG (he had insulted x)). (68b) is ruled out on pragmatic grounds: It would be unreasonable for George to ask a question containing an NP (a certain student) with a covert value accessible to him but not to the hearer. (68a), on the other hand, does not include an NP with a covert value in George’s question: The matrix S includes an NP with a covert value for the primary speaker, but the force of that higher S is declarative rather than interrogative. In this section, I investigated the source of the intuition of specificity in Higginbotham’s (26b) and distinguished indirect specificity from specificity in scope. I argued that the dissociation between reference and specificity in that example is explained if specificity can be indirect, such that the primary speaker is communicating the intent of the secondary speaker in the complement to say. I examined the existential presupposition of an indirectly specific NP, concluding that its apparent cancelability is due to its association with the secondary speaker. Indirectly specific NPs were represented here, as were directly specific NPs both in and out of scope, as assuming logical priority over Hintikka’s epistemic operator. Specificity in scope was represented, following Hintikka, as a logical prioritization of a function over an epistemic operator (cf. (43)); indirect specificity was represented as the indexing of the epistemic operator to a secondary rather than a primary speaker, as in (67b). Selective co-occurrence of indirect specificity with certain verbs of indirect

SPECIFIC NP IN SCOPE

283

discourse that permit communication of the secondary speaker’s intent support the conclusion that specificity is a function of speaker intent rather than an effect of discourse linking or shared speaker-hearer knowledge.

7.

Conclusions

I have proposed that cardinal NPs (including definites and indefinites) are ambiguous between a specific and a nonspecific interpretation. I have treated specificity as an effect of speaker intent, differing from other accounts that considered critical a discourse set membership or shared speaker-hearer apprehension. I defined specificity as in (15): (15) An NP X is specific iff there exists some process by which X acquires a specifiable value for the speaker. I first considered certain reflexes of specificity: A specific NP escapes from a scopal island; and, as a picture NP, a specific NP behaves as an extraction island. These behaviors can be explained under the hypothesis that cardinal NPs bear either a referential or a quantificational interpretation. The immunity to scopal islands accompanies referential interpretation, such that the NP behaves like a name; and the capacity to take narrow scope with respect to an embedded NP can likewise be explained as an effect of quantificational interpretation. Abusch (1994), however, argued against the ambiguity hypothesis, showing that an island-escaping NP need not take maximal scope and explaining island-escaping behavior under Heim’s account of indefinites. I suggested, however, that an island-escaping NP has an interpretation that differs from that of an NP with maximal scope; this specific interpretation, moreover, occurs in the case of intermediate-scope island-escaping indefinites. The referentiality associated with apparent widest-scope, islandescaping indefinites is a subcase of specificity: A referential NP is a specific NP that is not in scope. I considered Hintikka’s (1986) analysis of a certain as bearing logical priority with respect to an epistemic operator, showing that this representation can be used to explain direct and indirect specificity effects. In addition, specificity and the notion of domain of specificity were invoked in an account of question interpretation, to explain data that were not accounted for under a referential/quantificational ambiguity analysis.

284

BECKY KENNEDY

Notes 1.

A strong determiner is defined by Barwise and Cooper as one that yields either a tautology or a contradiction in a proposition of the form Det X are X. A weak determiner, on the other hand, yields a contingent proposition — one that is contingently true or false — in the same format.

2.

Diesing (1992) argued for a somewhat different ambiguity split, distinguishing a cardinal and a presuppositional reading in indefinite NPs or weak NPs. She associated her presuppositional reading with the syntactic occurrence of quantifier raising (QR) and her cardinal reading with existential closure as the source of quantificational force. Specific indefinites were characterized, under her analysis, as presuppositional and therefore quantificational. She argued that extraction cannot occur from a presuppositional (quantificational) NP, thus recasting the Specificity Condition in terms of her cardinal/ presuppositional ambiguity. A reader has, however, adduced counterexamples to her proposal; thus consider (i)-(ii): (i) (ii)

a. Who did you see a picture of t ? b. *Who did you destroy a picture of t ? a. Which river did the flood destroy the banks of t ? b. Which actress did the lunatic destroy a picture of t ?

Diesing argued that extraction occurs in (ia) from a nonpresuppositional NP but cannot occur in (ib) because destruction of an object presupposes its existence. Her account would have difficulty, however, with examples like (iia-b), in which the picture NP is again introduced by the verb destroy. The account developed here derives different associations: Specificity and presupposition are associated with referential rather than quantificational interpretation (cf. section 3.1). The possibility of extraction from the NPs in (iia-b) can be explained under the present account in terms of the irrelevance of specifiable value for the banks in (iia) and a picture in (iib). More precise explanation of these provocative examples is not possible here but would profit from furthur examination, drawing on Kuno’s (1987) insights regarding extraction from NP. 3.

A reader reminded me of the many counterexamples to Enç’s generalization (and to the traditional definiteness restriction). Thus consider (i): (i)

a. There’s always John. b. There is every reason to study them. (Ward and Birner 1995)

4.

Many also has a proportional reading, under which it is a strong rather than a weak determiner. Notice, with respect to (24a) (and in support of Reinhart’s analysis) that the donkey anaphora interpretation for that example holds only under the nonproportional (weak) interpretation for many donkeys, which signifies ‘many (specifiable) donkeys’ in (24a) but cannot signify ‘a large proportion of the donkeys’.

5.

Note that Heim, in illuminating problems in the ambiguity hypothesis, did not question the existence of specific indefinites that exhibited exceptional behavior relative to scopal islands, such that “the specific reading for an indefinite entails the widest-scope existential reading” (Heim 1982: 221). She proposed, in the context of her analysis of indefinites as open formulae bound by a higher quantifier, that by a rule of construal, a specific indefinite is raised and attached to the text node, thus acquiring maximal scope.

SPECIFIC NP IN SCOPE 6.

Susumu Kuno (personal communication) pointed out that (63c) will permit a reading under which the indefinite represents a referent for the secondary speaker if the target of Neg is restricted to he, as in the following: (i)

7.

285

George didn’t say that he had insulted a certain student. He said that Tom had.

Consider, however, the following counterexamples to the claim that the embedded subject NP in an example like (65) must have a specific interpretation (or, in Williams’ 1983 terms, must take wider scope than the matrix verb): (i)

a. b. c. d. e.

Someone trained in both syntax and phonology seems perfect for this job. A well-written text seems indispensable for an introductory course. A bigger turkey seems preferable for this age group. The largest home seems best for the class party. Three guests seem sufficient this time.

References Abusch, Dorit 1994 “The scope of indefinites”. Natural Language Semantics 2: 83-135. Almog, Joseph, John Perry, and Howard Wettstein (eds) 1989 Themes from Kaplan. Oxford: Oxford University Press. Bach, Emmon and George Horn 1976 “Remarks on ‘Conditions on transformations’”. Linguistic Inquiry 7: 265299. Barwise, Jon and Robin Cooper 1981 “Generalized quantifiers and natural language”. Linguistics and Philosophy 4: 159-219. Bolinger, Dwight 1978 “Asking for more than one thing at a time”. In H. Hiz (ed), 107-150. Cinque, Guglielmo 1990 Types of A’-dependencies. Cambridge, MA: MIT Press. Comorovski, Ileana 1988 Discourse and the syntax of multiple WH-questions. Ph.D. dissertation, MIT, Cambridge, MA. Davidson, Donald and Gilbert H. Harman (eds) 1972 Semantics of natural language. Reidel: Foris. Diesing, Molly 1992 Indefinites. Cambridge, MA: MIT Press. Donnellan, Keith 1966 “Reference and definite descriptions”. The Philosophical Review 75: 281304. Enç, Mürvet 1991 “The semantics of specificity”. Linguistic Inquiry 22: 1-25. Erteschik-Shir, Nomi 1973 On the nature of island constraints. Ph.D. dissertation, MIT, Cambridge, MA.

286

BECKY KENNEDY

Fiengo, Robert 1987 “Definiteness, specificity, and familiarity”. Linguistic Inquiry 18: 163166. Fiengo, Robert and James Higginbotham 1981 “Opacity in NP”. Linguistic Analysis 7: 395-421. Fodor, Jerry A. and Ivan Sag 1982 “Referential and quantificational indefinites”. Linguistics and Philosophy 5: 355-398. Frege, Gottlob 1952 “Ueber Sinn and Bedeutung”. In P. Geach and M. Black (eds), 7-78. Geach, Paul and Max Black (eds) 1952 Translations from the philosophical writings of Gottlob Frege. Oxford: Blackwell. Groenendijk, Jeroen, T. Janssen, and Martin Stockhof (eds) 1981 Formal methods in the study of language. Amsterdam: Mathematical Center. Guéron, Jacqueline 1981 “Logical operators, complete constituents, and extraction transformations”. In R. May and J. Koster (eds), 43-86. Heim, Irene 1982 The semantics of definite and indefinite noun phrases. Ph.D. dissertation, GLSA, University of Massachusetts, Amherst. Higginbotham, James 1987 “Indefiniteness and predication”. In E. J. Reuland and A. G. B. ter Meulen (eds), 43-70. Hintikka, Jaakko 1986 “The semantics of a certain”. Linguistic Inquiry 17: 331-336. Hiz, Henry (ed) 1978 Questions. Dordrecht: Reidel. Hornstein, Norbert 1984 Logic as grammar: An approach to meaning in natural language. Cambridge, MA: MIT Press. Kamp, Hans 1981 “A theory of truth and semantic interpretation”. In J. Groenendijk, T. Janssen, and M. Stockhof (eds), Part 1, Tract 135. 277-322. Kaplan, David 1989 “Demonstratives”. In J. Almog, J. Perry, and H. Wettstein (eds), 481-566. Karttunen, Lauri 1976 “Discourse referents”. In J.D. McCawley (ed), 363-385. Kennedy, Rebecca 1993 Reference and focus in English. Ph.D. dissertation, Harvard University, Cambridge, MA. King, J. 1988 “Are indefinite descriptions ambiguous?”. Philosophical Studies 53: 417440.

SPECIFIC NP IN SCOPE Kripke, Saul 1972 Kripke, Saul 1977

287

“Naming and necessity”. In D. Davidson and G. H. Harman (eds), 253-355. “Speaker’s reference and semantic reference”. Midwest Studies in Philosophy 11: 255-276.

Kuno, Susumu 1972 “Pronominalization, reflexivization, and direct discourse”. Linguistic Inquiry 3: 161-195. Kuno, Susumu 1987 Functional syntax. Chicago: The University of Chicago Press. Ludlow, Peter and Stephen Neale 1991 “Indefinite descriptions: In defense of Russell”. Linguistics and Philosophy 14: 171-202. May, Robert 1985 Logical form. Cambridge, MA: MIT Press. May, Robert and Jan Koster (eds) 1981 Levels of syntactic representation. Dordrecht: Foris. McCawley, James D. (ed) 1976 Syntax and semantics 7: Notes from the linguistic underground. New York: Academic Press. Milsark, Gary 1974 Existential sentences in English. Ph.D. dissertation, MIT, Cambridge, MA. Pesetsky, David 1987 “Wh-in-situ: Movement and unselective binding”. In E. J. Reuland and A. G. B., ter Meulen (eds), 98-129. Quine, Willard van Orman 1960 Word and object. Cambridge, MA: MIT Press. Reinhart, Tanya 1983 Anaphora and semantic interpretation. London: Croom Helm. Reinhart, Tanya 1987 “Specifier and operator binding”. In E. J. Reuland and A. G. B. ter Meulen (eds), 130-167. Reuland, Eric and Alice G. B. ter Meulen (eds) 1987 The representation of (in)definiteness. Cambridge, MA: MIT Press. Ruys, Eduard 1992 The scope of indefinites. Ph.D. dissertation, OTS, Utrecht. Ward, Gregory and Betty Birner 1995 “Definiteness and the English existential”. Language 71: 722-742. Williams, Edwin 1983 “Against small clauses”. Linguistic Inquiry 14: 287-308. Zimmerman, Thomas Ede 1993 “Scopeless quantifiers and operators”. Journal of Philosophical Logic 22: 545-561.

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

289

Some Referential Properties of English It and That* Akio Kamio Dokkyo University

Margaret Thomas Boston College

1.

Introduction

This paper examines certain uses of English it and that. Although at first these words seems to be a pair of largely interchangeable anaphoric pronouns, we will see that they exhibit subtle functional and pragmatic differences, differences which are consistent across a range of spoken and written examples. Our primary focus is on instances of it and that which have clausal antecedents in the preceding context that they in some sense “stand for.”1 But parts of our analysis also extend to it and that with nominal antecedents. Observe the example in (1): (1)

She bought a blanket during her lunch hour and brought a. it b. that back with her to the office.

Here, the antecedent of the anaphoric term is the simple NP a blanket. Thus, (1) exemplifies nominal it and that. On the other hand, (2) illustrates propositional it and that: * We are indebted to Susumu Kuno for his comments on a much earlier and abridged version of this paper, which was presented at the 1989 Annual Meeting of the Linguistic Society of America held in Washington, DC. We are also grateful for the financial support of the Dokkyo International Cooperative Research Fund.

290

AKIO KAMIO AND MARGARET THOMAS (2)

Tom knew that Joanne wanted to sell the car, and a. it b. that bothered him considerably.

In (2), what bothered Tom was that Joanne wanted to sell the car: the antecedent of the propositional it or that appearing in the second conjunct is the sentential object of know, which is at some level clausal.2 The roles which anaphoric pronouns play in discourse are quite complex. Since it is impossible to discuss the full range of the properties of it and that here, we need to clarify which aspects of these terms we will deal with in this paper. First, linguists usually consider it to be an anaphoric pronoun and that a demonstrative which refers to objects and events deictically. However, that may also be used as an anaphoric pronoun as examples (1) and (2) show. This paper deals with it and that as anaphoric pronouns, although Section 1.3 adverts to cases in which they have no linguistic antecedents and therefore might be called deictic. Second, while it cannot carry stress, that may do so. There are at least two versions of that, stressed and unstressed. Consider the following example. (3)

My father tried making a kitchen table by himself. Tony told him that buying one would certainly be a lot easier than a. that. b. THAT.

“That” in (3a) denotes unstressed that, while upper-case “THAT” in (3b) denotes stressed that. Stressed THAT singles out the act of making a table as uniquely troublesome among possible means of obtaining a table, including the act of buying one. Thus, THAT is inherently exclusive, contrasting its referent with all other possibilities. But with the use of unstressed that in (3a), Tony’s comment is merely that the relative ease of buying a table exceeds that of making one. There is no implication that making a table occupies a distinctively difficult position among means of acquiring furniture. Whereas THAT is explicitly contrastive, unstressed that is neither exclusive nor nonexclusive. This difference between that and THAT can be observed consistently among examples in which either may appear. Henceforth we will disregard THAT in our discussion, setting aside the difference just pointed out.3 Third, both it and that may appear as determiners within an NP, as observed in (4):

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT (4)

291

a. The committee reconsidered its recommendation. b. The committeee failed to make that recommendation.

Our work does not address such uses of it and that.

2.

Prior Knowledge

2.1 It and That and Prior Knowledge Our first proposal about the different referential properties of it and that is that the referent of it must be what we call “prior knowledge” to the speaker. Prior knowledge is typically information which a speaker already has access to before it enters into the relevant conversational exchange. In contrast, the referent of that need not be prior knowledge to the speaker. Consider the dialogue in (5): (5)

[A rushes into the room excitedly] A: Guess what! I just won the lottery! B1: *It’s amazing! B2: That’s amazing!

We assume in (5) that speaker A’s statement is completely novel information to speaker B; in this context, the response in B1 is quite artificial, whereas B2 is a natural expression of surprise at hearing unexpected news. However, B1, It’s amazing!, could be an acceptable response to the same announcement, in a different context. Imagine that before A arrives home, B hears the names of today’s lottery winners read over the radio. In this case, A’s statement is prior knowledge to B at the point when A utters it. Then the discourse in (6) is natural and coherent: (6)

A: Guess what! I just won the lottery! B: (Yes,) it’s amazing! I heard about it on the radio, and I’ve invited everyone on the block to our house for a party!

A similar principle is at work in (7): (7)

A: Overnight parking on the street is prohibited in Brookline. B1: That’s absurd! B2: It’s absurd!

292

AKIO KAMIO AND MARGARET THOMAS

B1 may be the voice of an out-of-town visitor who is unfamiliar with parking regulations in Brookline, for whom A’s comment constitutes novel information. B2, however, may be a Brookline resident to whom this city ordinance is well known. This difference in the properties of it and that is not limited to instances where these words are subjects of a copula. In (8), it and that are direct objects: (8)

A: Fred arrived even later than Sally. B1: I know that. B2: I didn’t know that. B3: I know it. B4: *I didn’t know it.

Notice that that can be used to assert that B has prior knowledge of A’s statement (as in B1) or, to deny that B has prior knowledge of A’s statement (as in B2). It can be used to assert the speaker’s prior knowledge, as in B3. But there is something quite odd about B4, *I didn’t know it, in this context. This is because B4’s explicit denial of previous awareness that Fred arrived later than Sally conflicts with the requirement that the referent of it be part of the speaker’s prior knowledge. A similar paradigm emerges with the expression be aware of it / that, as shown in (9): (9)

A: Janice fired her secretary yesterday. B1: Yes, I’m aware of that. B2: Really? I wasn’t aware of that. B3: Yes, I’m aware of it. B4: *Really? I wasn’t aware of it.

Unlike know and be aware of, the verb remember (and also recall) can appear with it even when the verb is negated, as in (10): (10) Fred may have told me that he wanted to quit his job, but I don’t remember a. it. b. that. This is because what the speaker denies in (10) is the ability to remember what Fred said. But the speaker does not deny that the information that Fred wanted to quit his job was ever part of his or her knowledge. If the speaker claims never to have possessed this information, the appearance of it is less

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

293

acceptable, as in (11a): (11) Fred never told me that he wanted to quit his job, so naturally I don’t remember a. ?it. b. that. Our claim, then, is that felicitous use of it in (5) through (11) requires that the speaker have knowledge of the referent of it at some point before this information enters into the discourse, but that no such requirement is imposed on the use of that. Thus in (7), speaker B2’s use of it signals that he or she already knows that overnight parking on the street is prohibited in Brookline before A utters a statement to that effect. In fact, B2 may have known about this regulation for years. But prior knowledge can also be relatively recentlyacquired information. Consider example (12): (12) [Alice and Carl are long-term housemates, whose relationship has been troubled recently. Alice comes home one evening to confront Carl with some news.] “Carl, I have something important to tell you. Mark called me into his office this morning and said he wanted to give me Gino’s job. He made me a great offer and I accepted it. But of course I’ll have to move to San Francisco.” Carl stared at her in silence for a long moment. Then, forcing himself to speak calmly, he said softly, “I hope it will make you very happy, my dear.” Carl has no prior knowledge of Alice’s news before she utters the statement in (12), but he can use it to refer to her plan to leave him and move to San Francisco. Notice, however, that the narrative assigns to him a “long moment” to absorb Alice’s announcement, and that his response indicates a degree of acceptance of her decision. These facts work to legitimize the appearance of it. In effect, Carl has incorporated entirely novel, and we can assume, somewhat shocking news in a very short period, such that he can treat it as prior knowledge. Contrast the acceptability of it in (12) to its unnaturalness in (13): (13) [Same situation] Alice: “....But of course I’ll have to move to San Francisco.” Carl: “*It will be the end of our relationship!” In (13) Carl responds without reflection, without taking time to integrate the

294

AKIO KAMIO AND MARGARET THOMAS

referent of it, and so its use here is unusual. Notice that Carl can immediately respond with That will be the end of our relationship! since that can refer to entirely novel information. The point here is that prior knowledge can be relatively recently-learned information, provided that there is some opportunity for it to be integrated into the speaker’s store of knowledge. Once this requirement is met, a speaker can refer to such information with it. It is worth noting that neither it nor that requires a linguistic antecedent: the speaker of it is only required to have conceptual foreknowledge of the referent of it. The example in (14) shows that both it and that can take deictic antecedents, in the terms of Sag and Hankamer (1984).4 (14) [A and B are driving along a highway. Suddenly, their right front tire blows out.] A1: I knew a. it was going to happen someday! b. that A2: I had no idea a. *it would ever happen! b. that It in A1 is fine because the speaker declares that he or she had previously anticipated the event which has just occurred, whereas it in A2 is unnatural because here the speaker explicitly renounces any claim to prior knowledge. Thus the requirement that its referent constitute prior knowledge to the speaker seems to extend to deictic uses of it. An additional set of examples shows how syntactic and semantic factors bear on the identity of a piece of information as prior information, and thus how they are implicated in whether that piece of information can be referred to by it. Observe the differences among responses to A’s statement in (15): (15) A: Just 200 years ago, Japan was closed to all foreign trade. B1: ??/* It surprises me. B2: That surprises me. B3: It still surprises me. B4: It has always surprised me. The response in B1 is unacceptable since the referent of it is manifestly not prior knowledge to the speaker: the verb surprise indicates that A’s statement constitutes novel information to B. B2 is fine since the referent of that need not be prior knowledge to the speaker. The interesting cases are B3 and B4. In the former, the adverb still means that B has been surprised in the past, and

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

295

continues to feel surprised in the present. Therefore, the referent of it represents prior knowledge; B3 means “Yes, I know that fact and my reaction is sustained amazement.” Similarly, in B4 the present perfect tense plus the adverb always communicate a meaning compatible with the requirement that the referent of it be prior knowledge. The sentence means “At some unspecifed point in the past I first experienced surprise and that feeling is still relevant now.” Thus, A’s statement is clearly prior knowledge to the speaker in B4 and the appearance of it is legitimate. 2.2 Our Hypothesis So far we have seen that prior knowledge is typically information which the speaker brings to the discourse, although it can be recently-acquired information if there is some evidence that the speaker has assimilated it on the spot. We have also seen that the status of a piece of information as prior knowledge can be indicated syntactically in various ways, for example by the use of certain adverbs or verb tenses. These observations lead us to suggest that it and that serve as anaphors at different stages of discourse processing. Our proposal is that it is available to represent information which has been integrated into the speaker’s store of knowledge, whereas that can take as its antecedent information which has not been thoroughly processed. Akatsuka (1985) has pointed out several instances in which natural languages distinguish information which is in her terms “newly-learned.” These data about it and that suggest that English is sensitive to an additional stage in mental processing, an “already-learned” stage, at which level it serves as an anaphor. That can serve to indicate what is from the speaker’s point of view novel or newly-learned information (although it may also represent old information), whereas it refers to information which has already undergone some degree of integration into the speaker’s store of knowledge.

3.

Wide Reference

3.1 Wide and Narrow Reference If it is true that the antecedent of it can only be information already assimilated by the speaker, this makes sense of a second, perhaps subtler, semantic

296

AKIO KAMIO AND MARGARET THOMAS

and functional difference between it and that. We characterize this second difference as a difference in the breadth of reference of the two terms. Our claim is that that narrowly specifies its referent, whereas it refers broadly: while that points, it evokes. The difference is more perceptible with propositional than with nominal it and that. Consider (16): (16) Sonja was born out of wedlock, but I never revealed a. it to her. b. that Notice that the semantics of reveal require that the speaker have prior knowledge of the verb’s direct object, so that there is no contrast between (16a) and (16b) on this score. Nevertheless, it and that introduce a subtle difference between the two versions of this sentence. Our claim is that it in this context refers broadly to a set of related facts and events: it means something like “that Sonja was born illegitimately, and the whole story of her mother’s disastrous affair with the Prime Minister, the dangerous international intrigue which resulted from it, etc.” But that is much less evocative; (16b) means simply that the speaker never told Sonja that her parents were unmarried at the time of her birth. Returning to the example in (12), the wide reference property of it would be displayed if Carl were to respond to Alice’s announcement with I hope it will make you very happy, my dear (compared with I hope that will make you very happy, my dear). The referent of it in this context may be something like “the fact that this transfer represents an important promotion for you, that you will make a new life for yourself in San Francisco, that you will find new friends there, that this move will effect a radical change in our relationship, etc.” But if Carl says I hope that will make you very happy, he refers more narrowly to Alice’s decision; he may mean only I hope your moving to San Francisco will make you happy. The ability of it to refer widely rather than to point specifically at its antecedent as does that can in some cases lead to the two anaphors having quite different referents in similar contexts. Consider (17): (17) [A middle-aged father and his 25-year old daughter are observing a young man on the street below pushing a baby stroller with an infant inside, evidently his own child.] Father: Look at that! There’s some young guy taking care of his own kid! Daughter: You see a. that a lot around here, Dad. b. it

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

297

If the daughter responds with that in (17a), her statement means something like, “The sight of fathers tending their own children is a common one in this town.” That refers simply to the scene which she and her father have witnessed on the street below. If she uses it, the referent of the anaphor becomes more diffuse, and her comment could be taken to be a much more general statement about child-rearing practices in that community. That is, (17b) could mean, “It is not unusual in this town for fathers to assume various childcare responsibilities, including but not limited to taking a baby out for a walk in a stroller. Fathers also do such things as feed, bathe, and clean up after their own children. None of these activites are unusual here.” This difference reflects the wider referential properties of it compared with those of that. Additional support for our claim about the referential properties of it and that comes from the fact that that may be the subject of what Declerck (1984) calls an inverted WH-cleft sentence, whereas it may not. (18) Georgina’s plane is due to arrive on schedule. Or at least a. that b. *it is what they told me over the phone. The second sentence in (18) contains an inverted WH-cleft; note that it cannot serve as its subject.5 This fact is understandable in the light of our analysis of the differences between it and that, given Declerck’s characterization of sentences like (18). Declerck argues that “specificational sentences”, a class which includes inverted WH-clefts, possess three properties: (i) they involve a contrast; (ii) they are emphatic, and (iii) they involve exhaustiveness. These properties are incompatible with the characterization of it that we are developing in this paper. First, as (19) shows, it cannot serve as the target of a comparison: (19) Any policeman can run faster than a. that. b. *it. Second, we have mentioned that it cannot bear contrastive stress. Third, it may refer to its antecedent broadly, rather than entailing exhaustiveness. 3.2 Native Judgments It should be clear that the judgments on which we are building our characterization of it and that are subtle. But they are nevertheless accessible to many

298

AKIO KAMIO AND MARGARET THOMAS

speakers of English, including 11 native speakers who were asked to consider pairs of sentences like (16a) and (16b) and to describe any differences in meaning which they sensed in the referents of it versus that. Many of these informants indicated that they found the reference of it more diffuse compared to the narrow, specifying force of that. (20) reports some of their comments: (20) a. Native speakers’ characterizations of it in sentences like (16a) “It is more general [...], it implies more of a situation.” “It implies there are a lot of other situations and stuff that you are talking about too.” “It means the whole process...the whole project.” “[It] is sort of general [...]. It doesn’t pinpoint the fact like that does.” b. Native speakers’ characterizations of that in sentences like (16b) “That is more localized.” “That is more pointing, more clear in its reference [...], it points back to the previous thing.” “That is not as much of a springboard as it. That is more restricted.”

4.

A Model of the Referential Properties of It and That

4.1 A Processing View of It and That The property that it refers broadly and that narrowly makes sense in the light of our claim that it represents a speaker’s prior knowledge. Prior knowledge is already-known, already-processed information. In the act of being processed, this information is integrated into other relevant facts and events in the speaker’s central store of knowledge. Thus it, the anaphor for prior knowledge, represents its referent as an aspect of a connected whole. In contrast, that can refer to information which has been processed relatively shallowly, information less richly embedded in its context, less centrally-located in the speaker’s storehouse of knowledge. Therefore that narrowly singles out its referent, while it evokes the context along with the antecedent. Figure 1 represents this characterization of it and that, using the shape of a bottle as a metaphor for linguistic processing:

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

299

Figure 1. Representation of two levels of information processing and their relationship to the anaphoric pronouns it and that

That can refer to incoming, newly-learned information as it passes in through the neck of the bottle. The referent of that can be relatively narrow. In contrast, it can represent information which has already entered the bottle and has become connected with its relevant context, the way water entering a bottle mixes with fluid already present there. That is, it refers to some aspect of a speaker’s prior knowledge, and in doing so broadly evokes the context of that already-known, already-integrated information. 4.2 Two Kinds of ‘Exception’ to the Model However, there are two sets of exceptions to the association which Figure 1 makes between the properties of it and that and the nature of information processing. We have seen in examples (8) and (9) that that can also be used to refer to information which is already part of the speaker’s store of knowledge. Extending the metaphor of the bottle, it is as though that can represent information already present in the bottom of the bottle, but which is being

300

AKIO KAMIO AND MARGARET THOMAS

recycled back through the neck of the bottle. (21) illustrates such a use of that: (21) A: I just found out that two of the Kennedy brothers were assassinated in the 1960’s. B: That’s common knowledge. Here the referent of that in speaker B’s response is clearly prior knowledge from his or her vantage point. Therefore we can say that both totally new information and prior knowledge — at least prior knowledge which is being recycled through the speaker’s processing device — can be represented by the propositional anaphor that. The second set of exceptional data involves uses of it in circumstances where the referent of the anaphor does not appear to have been part of the speaker’s store of information prior to its introduction into the immediate discourse. We have already alluded to one such case in the discussion of (12), where Carl quickly assimilates Alice’s announcement that she plans to move to San Francisco. Here it is as though novel information enters through the top of the bottle in Figure 1, and reaching the bottom with exceptional speed, is very quickly integrated by the speaker. Another example where it is possible to refer to totally new information with it involves expressions like I’m glad to hear it or I’m sorry to hear it as in (22) or (23): (22) A: My daughter Paula has been offered an excellent job at a school in Virginia. B: Oh, I’m glad to hear it. (23) A: My dog was just bitten by a poisonous snake! B: I’m sorry to hear it. Will he be all right? We feel that the availability of it to refer to novel information in these contexts is due to a particular kind of social accommodation. Although speaker A’s statements in (22) or (23) may actually constitute novel information to speaker B, B unconsciously represents the referent of the pronoun as having been incorportated into his or her store of knowledge as a means of communicating sincere involvement in the news which speaker A announces. This constitutes a pragmatically-motivated falsification of the actual status of the referent of it, which serves to express B’s solidarity with A. That is, B uses it in (22) to refer to the news of Paula’s job in order to indicate his or her vicarious participation in A’s parental satisfaction in a daughter’s accom-

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

301

plishment. This property is limited to quasi-idiomatic expressions like I’m glad to hear it or I’m sorry to hear it as demonstrated by the unacceptability of *It’s wonderful news! or *Oh, is it true? as responses to speaker A in (22), or the unacceptability of *It’s terrible! or *I didn’t realize it in (23). An extreme case illustrating the pragmatic forces bearing on it and that appears in (24): (24) [A is attending a funeral, and approaches family members of the deceased to express his or her condolences.] A1: It’s tragic! A2: *That’s tragic! Both it and that refer to the death of the person for whom the funeral is being held. A1 is a normal expression of sympathy, in which the use of it communicates the speaker’s prior knowledge of the death of the person in question and furthermore communicates that the speaker has integrated that information into his or her store of knowledge. But notice that use of that in the same context is unusual, even bizarre. Although in general that can refer to either novel or prior knowledge, the social context in (24) requires that speaker announce his or her participation in the grief of the family members of the deceased. Use of that gives no indication that the speaker has integrated this information or that he or she participates in the feelings of individuals to whom the remark is addressed. Thus A2 is out of place at a funeral, and might even be taken as sarcastic rather than comforting. In this way, the pragmatic forces on language use interact with the properties of it and that to produce a complex pattern of distribution which is in part driven by the nature of information processing and in part by social exigencies.6

5.

It and That with Nominal Antecedents

5.1 It and That and Presupposed Information In this section, we extend our analysis from propositional it and that to instances where the two anaphors have simple nominal antecedents. Note the differences in the reference of it and that in (25) and (26):

302

AKIO KAMIO AND MARGARET THOMAS

(25) First put the vase on a table, then take a picture of a. it. [= vase] b. that. [= table + vase] (26) First put a vase on the table, then take a picture of a. it. [= table] b. that. [= table + vase] In these examples, the referent of it is consistently a noun with the definite determiner: not surprisingly, it is an “old information” NP, already entered into the discourse, that represents the speaker’s prior knowledge. On the other hand that in (25) and (26) can refer to the table-plus-vase configuration created de novo by these sentences themselves. A related observation about nominal it and that is developed by Schiffman (1985:216). She argues that use of it dominates in contexts where its referent is “more highly presupposing.” That is, “[it] functions thematically” while “[that] functions rhematically.” In (25) and (26), the referent of it is taken to be the previously-mentioned noun, marked with the definite article. The referent of that need not be prior knowledge and therefore that can refer to the vase and table together, a configuration whose existence is created (and not presupposed) by the utterance itself. A similar example is cited in a different context by Isard (1975:289– 290): (27) a. First square 19 and then cube it. b. First square 19 and then cube that. In (27a), it is 19 which is first squared and then subsequently cubed. In (27b) it is the square of 19 which is cubed. In the first example, it refers to a previously-mentioned quantity, namely “19,” whereas in (27b) that refers to the newly-created quantity, “the square of 19.” Thus the property of propositional it that it represents prior knowledge from the point of view of the speaker seems to function in at least some cases of nominal it as well. The ability of it to pick out a referent which is prior knowledge to the speaker interacts with the properties of factive verbs. Consider (28) and (29): (28) The authorities regretted the strike, but it was inevitable. (29) The authorities regretted the strike, but that was inevitable. In (28) it in the second conjunct means the strike, whereas in (29) that means

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

303

the authorities’ regret of the strike. That is, we interpret the presupposed factive object of regret to constitute prior knowledge to the speaker, while the assertion made in the first conjunct of the sentence is not prior knowledge and is therefore less accessible to reference with it. 5.2 Linde’s (1979) Analysis Linde (1979) investigates the factors which condition the appearance of it and that in discourse in a study which looks at how native speakers of English use the two terms in describing the spatial layout of an apartment. Most of her examples employ nominal rather than propositional it and that. Linde concludes that “Both it and that are used to refer to rooms, but they are distinguishable [in that] in the case of reference to a room inside the focus of attention, the pronoun chosen is it” (p. 347), whereas there is a “strong preference for that to accomplish reference to items outside the focus of attention” (p. 350).7 For example, Linde (1979:349) cites the following example of use of that. (30) At the end of the long hallway there’s a door which is the entrance to my room, which is about seven by eleven and has a window facing the door. Next to that is a room which is about the same depth, about eleven feet, but about twice a wide. And next to that is another room which is about the same size as the one next to it. The speaker in (30) describes a sequence of rooms in an apartment. As what Linde calls the “focus” of the discourse moves from one room to the next, the speaker refers back to earlier foci using that. Thus in the first sentence the focus is my room. In the second sentence, the focus shifts to an adjacent room, and since my room is now out of focus, it is referred to with that. The process iterates in the third sentence as the focus shifts again. Linde’s observations are consistent with our claim that it (both nominal it, and more sharply, propositional it) refer to information which is already integrated into the speaker’s knowledge, that is, information which is central to his or her body of knowledge. In our terms, that can refer to newly-acquired information, that is, information which either is outside of the speaker’s center of knowledge, or which has not yet been integrated into it. In Linde’s terms, that refers to information which is outside the current focus of attention.

304

AKIO KAMIO AND MARGARET THOMAS

Table 1. Summary of some referential properties of English it and that It

That

Represents speaker’s prior knowledge (“already-learned information”)

Need not represent prior knowledge

Refers widely

Points narrowly

Information central to speaker’s knowledge

Information peripheral to speaker’s knowledge

Linde: information in focus

Linde: information out of focus

Schiffman: thematic material

Schiffman: rhematic material

French il: represents old information (Coppieters)

French ce: represents new information (Coppieters)

5.3 Summary In this section we have discussed some of the properties of nominal as opposed to propositional it and that. Although much previous work focuses on nominal uses whereas our major concern is with propositional uses, we find that our characterization of the properties of it and that presented in Sections 2, 3, and 4 is consistent with the observations of Schiffman and Linde. Table 1 represents some of the ways in which it and that have been depicted, summarizing our observations as well as those of other writers. (Table 1 also anticipates the discussion in Section 6 of French il and ce).

6.

Anaphoric Pronouns in Other Languages

6.1 French This section addresses the referential properties of French pronouns in the light of our analysis of English it and that.8 Togo (1988) compares French il (along with elle, ils, and elles) and ce in copular sentences. (Il is traditionally identified as a third-person singular pronoun, frequently translated as English he or it; ce is identified as a demonstrative pronoun, translated variously as he,

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

305

she, it, this, or that.) Togo’s major conclusion (pp. 109–110) is that ce does not make reference to what an NP itself stands for. Rather, this is accomplished by reference with il. Ce may refer to a situation surrounding what il refers to, or to an event that occurs in that context, or to actions or mental activities relevant to the reference of il. In our terms, ce appears to have the property of wide reference, which in Section 3 we have associated with English it. (31) and (32) give some of Togo’s examples: (31) Ce carrefour, c’est toujours pareil. Tous les jours il y a ce intersection, it’s always similiar. All the days there is quelque chose. some thing ‘That intersection, it’s always the same. There’s always an accident there.’ (32) Racine, c’est sublime. Racine, ce is sublime ‘Racine, his work is sublime.’ In (31), ce does not narrowly refer to the actual intersection in question. Rather, it refers broadly to the intersection and to events which occur there repeatedly, namely traffic accidents. In (32), ce clearly refers to Racine the person and to his works, or to Racine as viewed through his works. Thus ce indicates not only a particular person, but information associated with that person. If il were to replace ce in (32), then the utterance becomes a comment strictly about Racine the person. In this way the narrow referential properties of French il parallel those of English that.9 Additional examples of the properties of ce and il are displayed in (33): (33) Ton sac rouge, a. il est épouvantable. b. *c’est your bag red, a. il is appalling b. ce is ‘Your red bag, it’s appalling.’ (34) Ton sac rouge avec ta robe orange, c’est épouvantable. your bag red with your dress orange, ce is appalling ‘Your red bag with your orange dress, it’s appalling.’

306

AKIO KAMIO AND MARGARET THOMAS

While il can appear in (33a), ce in the same context in (33b) is impossible. This is because the antecedent of the anaphoric item is a singular definite entity which can only be appropriately referred to by the pronoun of narrow reference, il. However, ce in (34) is entirely natural. This is because the referent of ce is not the single entity your bag, but the assembly of the bag with the dress, and the resulting color clash which their juxtaposition produces. This complex configuration of referents can most naturally be represented by the pronoun of wide reference, ce. Togo cites earlier work on il and ce by Coppieters (1975) and Declerck (1983) which claims that ce is used in “identificational sentences” (i.e. sentences of the form “X is Y” meaning something like “X = Y”) whereas il is used in “descriptive sentences” (also known as “specificational” or “characterizational sentences”, i.e. copular sentences which characterize X as having the property Y). Togo, however, finds a number of counter-examples to the hypothesis that ce is associated with identificational sentences and il with descriptive sentences: (35) a. Un homme honnête, c’est rare. a man honest ce is rare ‘Honest men, they are rare.’ b. *Un homme honnête, il est rare. a man honest il is rare ‘An honest man, he is rare.’ The clause after the comma in (35) is clearly descriptive. Nonetheless only ce can appear here, contrary to the notion that il appears in descriptive sentences. Togo accounts for the data in (35) by pointing out that the sentence makes a generic statement, referring to humans in general: ce is necessary, since ce alone has the ability (in our terms) to refer broadly. Thus Togo’s characterization of il and ce is to be preferred over earlier accounts of their distribution.10 In this section, we have suggested that the property of wide reference which we attributed to English it in Section 3 is also relevant to the distribution of French ce. But this property does not function identically in both languages. Compare the English sentences in (25) and (26), repeated here, with the French examples in (33) and (34): (25) First put the vase on a table, then take a picture of a. it. [= vase] b. that. [= table + vase]

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

307

(26) First put a vase on the table, then take a picture of a. it. [= table] b. that. [= table + vase] In (25) and (26), it refers to the NP marked as old information with the definite determiner, whereas that refers to the newly-created complex assembly of the vase plus the table. In contrast, in (34) it is ce (whose parallel with it we have been developing) which refers to the assembly of the bag plus the dress. There are two possible interpretations of this difference between it in English and ce in French. First, the consequences of the property of wide reference for anaphors in the two languages may simply be different. A more appealing possibility is that there may be an interaction between the wide reference and prior knowledge properties of anaphors. Coppieters (1975:23) classifies ce as a conveyor of old information and il of new information, a distinction in some ways similar to our concept of prior knowledge versus newly-acquired information. But it may be that the wide reference properties of ce dominate, leading to its association with the (wide reference) bag plus dress configuration in (34). In English, it may be that the prior knowledge properties of it dominate, leading to its association with the (old information) vase and not the vase plus table configuration in (25) and (26). 6.2 Japanese This subsection explores some data from Japanese relevant to the properties of English it and that. Japanese has two kinds of anaphoric pronouns: overt anaphors, which are phonetically realized; and null anaphors, which are assumed on syntactic and semantic grounds to exist but which are not phonetically realized. Our discussion begins with the class of overt anaphors, specifically the tri-partite set of demonstratives. As is well known, Japanese demonstratives constitute a system of three discrete bound morphemes, ko-, so-, and a-. Attached to other morphemes they form words which indicate proximal, intermediate, and distal relationships, as illustrated for example by kore (conventionally translated as ‘this’), sore (‘that’), and are (‘that over there’). Ko-, so-, and a-, in combination with other morphemes, can also serve as anaphoric pronouns with either propositional or nominal antecedents. What is notable is that ko-, so-, and a- exhibit the property of wide reference. Observe (36):

308

AKIO KAMIO AND MARGARET THOMAS (36) A: Yosida-kun, kyoo kaisya ni konai’n desu yo. Yosida, today company to come+not COMP is SFP11 Uti ni denwa sitemo, dare mo denai si. Home to telephone do+if no one even answer+not SFP Kesa wa hayaku kuru tte itteta This+morning TOP early come that saying+was no ni. despite12 ‘Yosida hasn’t come to his office. I called his home, but no one answered. Besides, he said he was coming in early this morning.’ B: Sore wa hen da ne. it TOP strange is SFP ‘It’s strange, isn’t it?’

Here, sore in B’s response refers broadly to the facts that Yosida hasn’t shown up at the office and that no one answered the phone at his house despite his plan to come to work early. Thus (36) illustrates the wide reference property of sore; the English gloss represents sore in B’s response as it. Interestingly, however, the wide reference of sore in the context of (36) disappears in the interaction with one function of the nominative particle ga. Observe (37): (37) A: [Same as in (36)]. B: Sore ga hen da ne. it NOM strange is SFP ‘That’s strange, isn’t it?’ A: E? Doko ga hennan’ desu ka? what? Where NOM strange+COMP is Q ‘What? What’s strange?’ If B responds as in (37) to A’s statement in (36), the “exhaustive listing” character of the nominative marker ga (as described in Kuno 1973) changes the referential properties of sore. B’s response in (37) means something like ‘Sore and only sore is strange’ and thus the reference of sore is made quite narrow. Speaker A is left to wonder which part of his or her statement is being referred to by B, and so may follow up by posing the question in (37). In this way, the wide reference properties of Japanese overt anaphors, like the properties of English it, are sensitive to contextual factors.

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

309

It should also be noted that the antecedent of the distal morpheme a- is subject to the requirement that its referent constitute prior knowledge (Kuno 1973). Consider (38): (38)

A: Tanaka-kun o sugu Amerika ni haken Tanaka ACC immediately USA to dispatch subeki da to iu koe mo aru soo desu. should be COMP say voice even exist hear is ‘I hear that someone is saying that we should send Tanaka to the US immediately.’ B1: Are wa okasii to omotte ita yo. that TOP odd that thinking was SFP ‘I thought it was odd.’ B2: *Sore wa okasii to omotte ita yo. that TOP odd that thinking was SFP ‘I was thinking that that was odd.’ B3: Sore wa okasii to omou yo. that TOP odd that think SFP ‘I think that that is odd.’

The information reported by speaker A, which is the referent of are in B’s response, must have been prior knowledge to B1 before it was introduced into the discourse. In fact, speaker A might naturally continue the exchange by exclaiming, “Oh, I see you already knew about it!” It should be stressed that the other two demonstrative morphemes ko- and so- differ from a- in this respect: neither of the former must meet the requirement that their antecedents constitute prior knowledge. In this sense ko- and so- parallel English that, whereas a- behaves more like English it. For example, B2 is unacceptable because the phrase omotte ita ‘was thinking’ indicates that the given information is prior knowledge to the speaker, while on the other hand sore indicates that the same information is novel to the speaker. B3 with sore is acceptable because the phrase omou ‘think’ indicates that the information is novel to the speaker. This contrast confirms that so- is parallel to English that.13 Null anaphors in Japanese, which have long been recognized to correspond to English overt anaphors, have quite different properties. In recent formal syntactic studies, null anaphors such as “[pro]” (in the sense of Chomsky 1986) can refer either to specific persons or can have arbitrary reference:

310

AKIO KAMIO AND MARGARET THOMAS (39) a. [pro] atte miru to, Yamada-san wa ii [pro] meeting try when, Yamada Mr./Ms. TOP nice hito da. person was ‘Meeting Mr./Ms. Yamada, he/she turned out to be a nice person.’14 b. [pro] sunde miru to, soko ga ki ni iru [pro] living try when, there NOM mind to enter mono da. SF is15 ‘Living in a place, we tend to like it there.’

In (39a), the referent of the null object of au (> atte ‘meet’) can only be interpreted as identical with Mr. / Ms. Yamada. Consequently, this is a case of specific (co-) reference of the null anaphor [pro]. In contrast, in (39b) the referent of the null complement of sumu (> sunde ‘live’) can be interpreted as being arbitrary. As long as the referent of [pro] is semantically compatible with the verb sumu, it can be anywhere. Thus, this is a case of arbitrary reference of [pro]. There are, however, cases which are intermediate between specific and arbitrary reference. That is, there are instances where [pro] refers not to a specific entity but rather broadly to a range of relevant objects or information. In (40), for example, [pro] seems to have wide (but not arbitrary) reference: (40) a. [pro] matte iru to, Hikari 129-goo ga [pro] waiting is when, Hikari 129 number NOM haitte kita. pulling came ‘When we were waiting, Hikari No. 129 was pulling in.’16 b. [pro] kangaete miru to, Haruko mo moo 30 da. [pro] thinking try when, Haruko even already 30 is ‘Come to think [of it], Haruko is already 30.’ c. [pro] kiite miru to, sono hito wa titi no [pro] listening try when, that person TOP father of mukasi no dookyuusei datta. long+ago of classmate was ‘Listening [to him], it turned out that the person was one of my father’s old classmates.’

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

311

What does the [pro] object of matsu (> matte ‘wait’) in (40a) refer to? At first glance it might seem to refer to the train named Hikari No. 129; but it is possible that the speaker is actually waiting for Hikari No. 130, or even for some entirely different kind of train. On the other hand, it is obvious that the speaker is not waiting for a ship or an airplane. Therefore the referent of [pro] ranges broadly over various kinds of trains, but not over other means of transportation. Similar analyses are available for (40b) and (40c). In the former case, the [pro] object of kangaeru ‘think’ includes various kinds of information relevant to Haruko’s age. In (40c), the [pro] object of kiku ‘listen’ includes all of the previous discourse, from which the speaker has deduced that the person in question is a classmate of his or her father. It is important to note that the referent of [pro] in examples like (40a) may not always be determined on the basis of the previous discourse. Consider (40a) embedded in the discourse in (41): (41) Subete no sigoto ga owatta no de, wareware wa all of work NOM finished because, we TOP kaeru koto ni sita. Massugu eki ni yuki, hoomu return -ing to decided straightaway station to go, platform de [pro] matte iru to, Hikari129- goo ga haitte on pro waiting is when, Hikari129 number NOM pulling kita. came ‘Since all of the work was finished, we decided to go home. Going directly to the station and waiting on the platform, Hikari No. 129 was pulling in.’ The discourse in (41) provides no clues as to the specific referent of [pro], although the words platform and station broadly limit the range of possible referents. Therefore the wide reference property of [pro] is intact even in a rich context such as that in (41). In this way, we find that Japanese null anaphors have wide reference properties in at least some cases. Summarizing Section 6, we have argued based on evidence from Togo (1988) that French ce has wide reference while il refers narrowly. The Japanese anaphoric morphemes ko-, so-, and a- do not always have wide reference properties, although they can refer widely in certain contexts. The distal morpheme a- differs from ko- or so- in that it requires prior knowledge of its antecedent on the part of the speaker. Finally, we have seen that

312

AKIO KAMIO AND MARGARET THOMAS

Japanese null anaphors in at least some contexts have wide reference properties.

7.

Conclusion

In conclusion, we make the following claims about the referential properties of English it and that: it refers broadly to prior information, information already known and already entered into the speaker’s central store of knowledge. That points narrowly, to incoming information that may be either novel or familiar, information which is some sense more peripherally located in the speaker’s knowledge. This suggests that English anaphors distinguish between on the one hand, already-processed information which a speaker has absorbed and integrated into its relevant context, and novel information which has been processed only to some shallow degree. Already-processed information is represented by propositional and nominal it, whereas that can refer to either type of information. Furthermore, we observed that the properties of English it and that seem to be displayed among anaphoric pronouns in at least some other languages. We also illustrated how the referential properties of it and that interact with certain pragmatic constraints. Thus our observations are consistent with work by Akatsuka and others who have claimed that linguistic structure reflects the steps by which humans process information.

Notes 1.

As is well known, English anaphors may precede or follow their “antecedents.” For the sake of simplicity, we will discuss only cases where it or that follow their antecedents, but the analysis presented here applies equally to instances where it or that may precede.

2.

It is sometimes difficult to determine whether the antecedent of an anaphoric pronoun is a clause or an NP. This is because almost all clauses can be nominalized by adding to them such forms as the fact that, the act of, etc. We will not be concerned with whether the antecedent of a propositional anaphor is syntactically an S (or an S-bar) or an S / S-bar dominated by an NP node. Instead, we will treat as a propositional anaphor any instance of it or that whose antecedent contains a clause at some level. The terms “nominal it / that” are then reserved for cases like (1) where the antecedent does not contain a clause at any level.

3.

There is another propositional anaphor, this, which is usually considered to constitute a

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

313

paradigmatic pair with that. In (i), any one of four anaphors may refer to the proposition contained in the first sentence. (i) The clerical workers have been threatening to strike over a dispute about working conditions. (a) That (b) THAT (c) This (d) THIS has been a major problem at the New Haven office. Differences between (ia) and (ib) are consistent with our proposals about how the referential properties that and THAT differ. Unstressed this, parallel with unstressed that, refers neither exclusively nor non-exclusively, while THIS, parallel with THAT, is exclusive. What distinguishes the two realizations of this from the two realizations of that is that the former pair implies that some variety of closeness exists between the speaker and the antecedent of the anaphoric term. For example, (ic) or (id) could be uttered by the manager of the New Haven office in conference with the company president. The manager is presumably close to the problems which plague the office. Use of this or THIS by the manager reveals this closeness in reporting to the president. If the president, having heard the manager’s report, utters the sentence in (i) while meeting with the board of directors, he or she would most naturally use either that or THAT, as long as the troubles of the New Haven office have remained at a distance from the president’s perspective. Therefore, we will say that both this and THIS involve a sense of closeness whereby the speaker represents himself or herself as having some kind of intimacy with or authority over the referent of the term. The representation of closeness is absent when a speaker uses either that or THAT. We do not discuss this versus that further. 4.

This observation contrasts with the claim of Frajzyngier and Jasperson (1991:43), that that is used to refer to elements in the domain of de dicto, whereas it refers to elements in the domain of de re. However, Frajzyngier and Jasperson’s association of that with de dicto referents is congruent with our proposals. As discussed in Section 3, information referred to by that lies at a shallower level of processing, at which point it still retains much of its verbatim character. Thus it is natural that that be available to refer to the overt linguistic representation of a given piece of information.

5.

This applies only to propositional it. It can appear in subject position of an inverted WHcleft if its referent is nominal rather than propositional. (ii)

Cheryl just got a job as an air-traffic controller. a. That was what she had always wanted to i. ii. b. It was what she had always wanted to i. ii.

do. be. */?? do. be.

In (iia), that can refer either to the proposition “(Someone) gets a job as an air-traffic controller” as when the cleft ends with do; or that can refer to the simple NP “an air-traffic controller” as when the cleft ends with be. On the other hand, (iib) shows that it as the subject of the inverted WH-cleft may refer to an NP, but only marginally to a proposition. We do not have an explanation of why the properties of propositional and nominal it diverge in this case. Our claim is only that there are certain properties shared by propositional and nominal it and that; we do not claim that they are alike in every way.

314

AKIO KAMIO AND MARGARET THOMAS

6.

As an additional counter-example, it occurs to us that there is no announcement, no matter how unexpected, to which a speaker cannot respond with I can’t believe it! or I don’t believe it! Going back to the example in (5), when A declares, Guess what! I just won the lottery! B can indeed respond with I can’t believe it! even without having had any prior knowledge of A’s announcement. This challenges our claim about it, since there is no obvious pragmatic motivation for exceptionally rapid processing of A’s announcement. However, notice that a range of other expressions still reject it in this context, such as *I didn’t know it or *It’s astonishing. There seems to be something unusual about the verb believe: perhaps the expressions I can’t believe it! and I don’t believe it! function as idioms, independent of the general requirements on felicitous use of it which we have described.

7.

Linde’s use of the term “focus” differs from the conventional sense. To Linde (1979:345–347), the “focus of attention” is a kind of pointer moving across the discourse, highlighting the central locus of attention at any given instant. Information which is “outside the focus of attention” is material not being pointed to at that moment.

8.

We are grateful to Yuji Togo for making us aware of his work. All the French examples in this section are taken from Togo (1988) but the English glosses are provided by us. The interpretation of Togo’s examples is also ours, although we draw on his analysis.

9.

An interesting difference between French and English is that French seems to make use of the wide referential properties of ce not only when it is an anaphoric pronoun as in (32), but also when it functions as a determiner as in (31). We do not have evidence that English it has wide reference in its role as a determiner.

10.

Many dictionaries (e.g. Cassell’s French Dictionary) provide this or that as the primary English translations of ce. This may need to be re-considered in the light of the fact that neither this nor that shares the wide reference property of ce.

11.

We employ the following abbreviations: ACC object marker; COMP complementizer; NOM subject marker; Q question marker; SFP sentence-final particle; SF sentence-final form; TOP topic marker.

12.

Yo roughly means ‘I tell you’; si means ‘also.’

13.

We are grateful to Ken-ichi Takami (personal communication) for providing us with (38B2) and (38B3).

14.

A more complete representation of the first clause in (39a) would be PRO pro atte miru to. That is, it should contain [PRO], a null subject, and [pro], a null object. Japanese is a null subject language and thus subject position is frequently not phonetically realized. But since null subjects are not the target of our analysis here, we omit them in the examples which follow.

15.

Mono here means roughly ‘It is the case that...’

16.

Matsu (> matte ‘wait for’) in Japanese is clearly transitive. Therefore it must have a direct object, which in the case of (39a) is a null pronoun, [pro].

SOME REFERENTIAL PROPERTIES OF ENGLISH IT AND THAT

315

References Akatsuka, Noriko 1985 “Conditionals and the epistemic scale”. Language 61: 625–639. Chomsky, Noam 1986 Knowledge of language: Its nature, origin, and use. New York: Praeger. Coppieters, René 1975 “The opposition between il and ce and the place of the adjective in French”. In S. Kuno (ed), 221–280. Declerck, Renaat 1983 “It is Mr. Y or he is Mr. Y?”. Lingua 59: 209–246. Declerck, Renaat 1984 “The pragmatics of it-clefts and WH-clefts”. Lingua 64: 251–289. Frajzyngier, Zygmunt and Robert Jasperson 1991 “That-clauses and other complements”. Lingua 83: 133–153. Givón, Talmy (ed) 1979 Syntax and semantics 12: Discourse and syntax. New York: Academic Press. Isard, Stephen 1975 “Changing the context”. In E. L. Keenan (ed), 287–296. Keenan, Edward L. (ed) 1975 Formal semantics of natural language. Cambridge: Cambridge University Press. Kuno, Susumu 1973 The structure of the Japanese language. Cambridge, MA: MIT Press. Kuno, Susumu (ed) 1975 Harvard studies in Syntax and Semantics 1. Cambridge, MA: Department of Linguistics, Harvard University. Linde, Charlotte 1979 “Focus of attention and the choice of pronouns in discourse”. In T. Givón (ed), 337–353. Sag, Ivan A. and Jorge Hankamer 1984 “Toward a theory of anaphoric processing”. Linguistics and Philosophy 7: 325–345. Schiffman, Rebecca J. 1985 Discourse constraints on it and that: A study of language in careercounseling interviews. Dissertation, The University of Chicago. Togo, Yuji 1988 “Mon frère, il est linguiste et le coupable, c’est lui”: Daimeishi il to ce no yohoo ni tsuite (On the use of the pronouns il and ce). Furansugo furansubungaku kenkyu (Research on French and French literature). 53: 102– 111.

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

317

The Role of Empathy in Sentence Production A Functional Analysis of Aphasic and Normal Elicited Narratives in Japanese and English Lise Menn, Akio Kamio, Makoto Hayashi, Ikuyo Fujita, Sumiko Sasanuma, and Larry Boles* University of Colorado, Dokkyo University, International University for Health and Welfare, and University of Hawaii

1.

Introduction

Linguists still tend, for the most part, to be uncertain as to the theoretical relevance of work with abnormal speakers (Menn and Obler 1982, Menn 1989). In this paper, we argue as follows: in the attempt to account for the error patterns of aphasic patients in our picture-description experiments, we have had some success through invoking a concept of empathy. We have been strongly influenced by Kuno’s work (Kuno and Kaburaki 1977, Kuno 1978, 1987); however, we have treated the concept of empathy as a psychological primitive — that is, as an attitude or state of mind towards an entity; specifi* We gratefully acknowledge the assistance of Audrey L. Holland and her research group at the Center for Neurogenic Language Disorders, University of Arizona, in affording us access to their patient population, and of Yasunori Morishima, who translated and carried out a preliminary analysis of the Japanese narratives. This work was partly supported by NIH grant PHSDC00730-02, the Japan Society for the Promotion of Science, and the Japan-U.S. Educational Commission (Fulbright Commission). Data were collected while Dr. Fujita was at the National Institute for Rehabilitation of the Disabled in Tokorozawa, Dr. Sasanuma was at the Tokyo Metropolitan Institute of Gerontology, and Dr. Boles was at the University of Arizona.

318

LISE MENN ET AL.

cally, an attitude of ‘identification with’ a participant in an event. That attitude may or may not have an overt linguistic manifestation; in addition, a speaker may have no particular empathic reaction to a given circumstance. Therefore, an utterance may have no empathic focus. This psychological/cognitive notion of ‘empathy’ has been helpful in describing the response patterns of both aphasic and normal control subjects. We can therefore argue that it is an empirically testable and psychologically real notion. However, attempts to analyze elicited narratives and manipulate ‘empathy’ experimentally have led us to realize how complex a notion it is. So in the conclusion of this paper, we discuss some problems that have arisen and our proposed research directions. In our more ambitious moments, we conceive of this proposed line of research as part of a more general program of experimental functional syntax.

2.

Background

2.1 ‘Empathy’, Speaker, and Hearer in a Cognitive Model of Sentence Production Current psycholinguistic models of language production (e.g., Levelt 1989) explicitly acknowledge that the selection of a particular syntactic/lexical form for the expression of a proposition begins with pre-linguistic cognitive processes involving the speaker’s evaluation of the addressee and the setting of the discourse. These processes include discourse tracking, which requires maintaining and updating a representation of the knowledge that is believed to be possessed by or available to one’s conversation partners. Such a representation is needed in order to know what the topic is, and to determine when to use a pronoun or one of many possible full noun phrases as referring expressions. Selection among forms of expression is also affected by sociolinguistic considerations of politeness, setting, group structure, etc.; evaluation of these must also be a cognitive process. This group of discourse factors is hearer-oriented; that is, they concern how the speaker takes the hearer’s needs into account. But the normal speaker is not a perfect rhetorician; natural discourse does not always afford smooth referent tracking and perfect coherence. Even if the speaker maintains a perfect model of the hearer’s mental state — which in itself seems unlikely —

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

319

he or she does not always utilize that model properly. Models of pre-linguistic choices for language processing therefore also need to include speaker-oriented factors: that is, factors which reflect the speaker’s impulses and which may interfere with the speaker’s evaluation of the hearer and the setting. Our recent research has focused on some of these factors which seem to be matters of ‘speaker’s impulse’, at least in rhetorically-naive, unedited spoken language. Empathy (in the psychological sense of identification with a person or creature involved in the situation under discussion) is one such factor; two others that are found in our data appear to be an entity’s motion or other exhibited power to cause an event, and its novelty/unexpectedness. In ordinary, rhetorically-unskilled speakers, the integration of these factors can be clumsy, leading to revisions and afterthoughts. For example, in English and Japanese, speakers often use a pronoun or other deictic word, and then supply the referent as a repair. Japanese conversation also contains OVS and other non-verb-final sentences in which the speaker gives new information, and then adds the old referent if he or she thinks the hearer might have difficulty in retrieving it (Ono and Suzuki 1992). These psychological factors which affect the speaker’s impulse — empathy, motion/causal potential, and novelty — may be subsumed under ‘salience’, and all three were manipulated to some extent as variables in the most important experimental work in this area to date, Sridhar (1989). Sridhar’s study of the relation of psychological salience to linguistic encoding in ten languages was based on order of mention in speakers’ descriptions of objects and events in single filmed episodes. The objects varied in position, motion, and humanness,1 and the languages varied in canonical word order (SOV, SVO). While the data showed many language-related complexities, the study showed (p. 223) that “Entities rendered salient by virtue of their intrinsic meaningfulness (e.g., humanness), or perceptual focus, tend to be expressed sentence-initially, at or near the beginning of the sentence in SVO languages, leading to the use of object-fronted topicalized sentences.” The empirically noted ‘impulse factors’ of empathy, power to cause events, and novelty are, then, elements of the situation which attract the speaker’s attention; in our data for Japanese as well as English, they tend to result in early explicit mention of the attention-getting entity. An evaluative predicate may also be mentioned early in an emotive utterance: Ono and Suzuki (1992: 439) claim that Japanese predicates are sometimes uttered before topics (without breaks in intonation or any other indications of being

320

LISE MENN ET AL.

afterthoughts/repairs) to express “the speaker’s urgency in expressing his/her inner feeling” in natural Japanese conversations: (1)

...yaa da na kono kokonatsu disgusting COP PTCL this coconut ‘Tastes awful this coconut drink.’

They therefore argue (p. 441) for “...the importance of affect as one of the motivations of word order variability in Japanese.” The linguistic data are, necessarily, complicated by the fact that the speaker’s impulse factors interact with the hearer-oriented factors already mentioned. The speaker should make sure that the hearer can integrate new information into his/her existing mental representation of the events being discussed; if re-evocation of old information is needed for this, the sufficiently skilled speaker postpones the mention of new information, as is well known (Chafe 1976). Work with aphasic patients who have reduced syntactic abilities shows that they mention new information (Bates and Wulfeck 1989) even when they can do little else; so do very young children (Greenfield and Smith 1976). This has typically been discussed in terms of attunement to the hearer’s needs (if any attention has been paid at all to the speaker-hearer distinction), but it is equally well accounted for in terms of the effect of novelty on salience to the speaker. The work of Karmiloff-Smith (1979, 1986) on young children’s pronoun use certainly supports a conservative estimate of children’s ability to anticipate hearers’ information needs and, thence, our distinction between speaker-based and hearer-based functional considerations. Consideration of experimental and natural discourse in speakers with limited skills makes it painfully evident that there must be a process of cognitive integration of speaker impulse with consideration for the hearer, which then leads to or interacts with syntactic selection. The final syntactic choice is additionally influenced by the relative availability of the various syntactic frames which might be suitable for conveying the information in question, probably by structures recently used (Bock and Loebell 1990), and of course by other language-specific factors. We envision this whole process of integration as taking place in a parallel-processing constraint-satisfaction model (Rumelhart, McClelland and PDP group, 1988), but this claim is independent of those above.

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

321

2.2 Empirical Work on Aphasia In recent years, a group of aphasia researchers led by Elinor Saffran and Myrna Schwartz (Saffran, Schwartz, and Marin 1980; Linebarger, Schwartz, and Saffran 1983) obtained several sets of findings which were difficult to reconcile both with each other and with ordinary clinical experience. First, they found that so-called ‘agrammatic’ aphasic patients — that is, patients whose speech is slow, usually poorly articulated, and linguistically characterized by simplified syntax, a high percentage of functor omissions, and use of morphologically unmarked forms — seemed to make word-order errors in describing the action of one animate figure on another (boy chases dog), the relative configuration of two inanimates (suitcase is behind chair), and, most especially, the action of an inanimate on an animate (ball hits boy). In contrast, there were few such errors in description of the action of an animate on an inanimate (boy hits ball). These errors suggested that the patients were unable to deal with the argument structure of a sentence in syntactic terms, and that their successes in the Animate-Action-Inanimate sentence type were due to a strategy of choosing the animate referent as subject. This strategy could only work above chance level in the ‘boy hits ball’ case, and would give below-chance results in the case of ‘ball hits boy’. However, these results were difficult to reconcile both with the absence of reported word-order errors in the free conversations of patients of this type, and with the second set of wholly unexpected findings of this same research group (Linebarger, Schwarz and Saffran 1983, since replicated and extended by Wulfeck 1988), that such patients had near-normal abilities in making many types of grammaticality judgements. Study of Japanese speakers’ difficulty in describing a particular cartoon sequence in the Standard Language Test of Aphasia used in Japan (Fig. 1 “Hat”) led us to consider the following possible explanation of the Saffran and Schwartz group’s results on ‘boy hit ball’ vs. ‘ball hit boy’: Agrammatic aphasic speakers have limited syntactic production resources, by definition. Perhaps they attempt to start sentences by mentioning the empathic focus first, but can basically only produce active-voice sentences and/or agent-first sentences. This fits in with experimental and narrative data showing that aphasics of all types tend to produce sentences with canonical word order and simple verb forms (Bates and Wulfeck 1989). Suppose, in particular, that the stimulus picture is ‘ball hits boy’, and that no animate entity other than the

322

LISE MENN ET AL.

Figure 1. “Hat,” from the Standard Language Test of Aphasia, Nirayama conference.

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

323

undergoer ‘boy’ is illustrated. In that case, the empathic focus is presumably the person affected (undergoer). The aphasic speaker, like the normal speaker, might therefore begin by mentioning the undergoer, but then, being unable either to passivize or to topicalize, would have only two choices: to abort the sentence or to continue with an erroneous ‘reversed’ production. In the opposite case of describing ‘boy hits ball’, however, there would be no word-order errors, as the narrator, after beginning with the empathic focus ‘boy’, could go on in the canonical word order, using active voice. As mentioned, Saffran, Schwartz, and Marin (1980) also reported reversal errors in agrammatic aphasics’ attempts to describe Animate-Animate and Inanimate-Inanimate interactions/configurations (‘the cat chased the dog’; ‘the suitcase is behind the chair’). When such pictures are presented in isolation, we assume that there is probably no empathic focus; the reported reversal errors could result from the patient having no empathic or other cognitive guidance as to which entity to start with. In a cognitive model of sentence production, this would cause the arousal of competing, equallyappropriate sentence frames which could blend and cause errors. The fact that reversal errors have not been reported from agrammatic narratives (Menn and Obler 1990: 1386) would be easier to explain under the empathic-focus-first hypothesis than under the hypothesis that agrammatic aphasics have lost the ability to compute thematic relations. Under the empathic-focus-first hypothesis, empathy or other salience factors would be automatically and appropriately determined in narrative or free conversation about real things that matter. That is, there would always be ‘salience-based guidance’ as to which referent to mention first, so reversal in AnimateAnimate or Inanimate-Inanimate situations would not occur. Complete, fully reversed sentences involving Inanimate acting on Animate, according to this hypothesis, would be vanishingly rare because self-monitoring would be adequate to insure that sentences containing errors of that magnitude would be aborted. Aborted sentences mentioning the undergoer first might indeed occur, but would be inconspicuous among other incomplete noun-phraseonly utterances. 2.3 The Linguistic Expression of Empathy We follow Kuno in our analysis of the linguistic manifestations of the speaker’s empathy or viewpoint. However, as we need the term ‘empathy’ for

324

LISE MENN ET AL.

an affective state, we will use ‘viewpoint’ to describe the cognitive orientation decodable from the form of an utterance. In general, the speaker shares the viewpoint of the grammatical subject of the sentence rather than that of an object or oblique argument, according to Kuno’s surface structure empathy hierarchy (Kuno 1987: 207). Thus, we interpret Kuno as implying, at least for English, that the speaker typically encodes the entity which attracts his/her empathy as the subject of a clause rather than as some other argument. This is how the active/passive contrast can give explicit indications of viewpoint: contrast, again, ‘The ball hit the boy’ (neutral or following the trajectory of the ball) with ‘The boy got hit by the ball’ (viewpoint, and presumably empathy, shared with the boy). The cooperative hearer, in decoding, then has his/her empathy attracted by the sentence subject, other things being equal. However, they are not always equal, as lexical choices also give empathic information. For example, deictic expressions (e.g., demonstratives like ‘here’/’there’, ‘this/that’; a few motion verbs such as ‘come’/’go’, transfer verbs like ‘bring’/’take’, ‘send’/’receive’) give us lexical indications of the speaker’s viewpoint independent of syntax or context. Consider the hypothetical contrast: ‘George took them to Mary’/’George brought them to Mary’. In the first sentence, the speaker shares George’s viewpoint or is neutral; in the second, the speaker clearly shares Mary’s point of view. Explicit indication of the state of mind of a referent is also generally taken as indication that the speaker shares the viewpoint of that referent. However, there is no generally agreed-upon exhaustive listing of indicators of shared viewpoint.

3.

The Experimental Study

3.1 Hypothesis The hypothesis of the present study is that aphasics and normals are alike in preferring to begin sentences by mentioning or referring to the empathic focus first, and further, that the agency and animacy effects found by Saffran, Schwartz, and Marin are consequences of this preference, rather than resulting from a default ‘animacy strategy’ that is used because verb-argument structures can no longer be computed or semantically interpreted. The ‘empathic focus first’ proposal says nothing directly about agency or

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

325

animacy, but the empathic focus of an utterance is most often a person, and very often a person doing something. To test the claim that empathy is really the key factor, rather than animacy or agency, speakers must be induced to talk about events in which a plausible empathic focus is not an agent. Furthermore, either the agent and the undergoer must both be animate, or the agent/cause of the action must be inanimate. 3.2 Experiment 3.2.1 Stimulus Materials Together with colleague Barbara Fox, we commissioned four 3 or 4-frame narrative cartoon strips in addition to the Japanese ‘Hat’ strip, resulting in a total of nineteen frames (Figs. 2a-d) which showed the four possible basic patterns of interactions between animates and inanimates (inanimate subjects acting on animates, two animates, two inanimates), plus several agentive and non-agentive intransitive actions. The stories also varied other factors, such as how many humans were involved, whether there was a single clearlydefined protagonist, how ‘volitional’ the protagonist was, and whether the protagonist (the presumed empathic focus of the story) or another person was the agent in Animate-Animate (A-A) interactions. In four of the stimulus stories, there is a clear protagonist, i.e., a person whom the story is ‘about’. In ‘Hat’ (Fig. 1), ‘Boat’ (Fig. 2c), and ‘Apple’ (Fig. 2d), there is in fact only one person, who is visible in all or almost all of the frames. In ‘Race’ (Fig. 2b), a girl is the only person in the first frame, and she is visually foregrounded in the other three frames. In ‘Ball’ (Fig. 2a), however, the story changes focus; it starts out with the boy, but then follows the ball in through the window and looks at the man’s reactions. At the end, as the man looks out the broken window, one is returned to thinking about the boy, but he is invisible. ‘Protagonist’ is therefore not clearly defined by the cartoon; this is reflected in the stories which our subjects tell, as will become apparent. 3.2.2 Elicitation Method English and Japanese aphasic patients and controls were presented with these pictures, and asked to tell the stories; they were prompted and aided as little as possible. Their narratives were tape recorded, transcribed, and analyzed for error patterns and sentence forms chosen or attempted. Sample English and Japanese narratives are given in Appendix A.

326

Figure 2a. “Ball.”

Figure 2b. “Race.”

Figure 2c. “Boat.”

LISE MENN ET AL.

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

327

Figure 2d. “Apple.”

3.2.3 Subjects The aphasic subjects were unselected for diagnostic type, in order to see whether the type of aphasia had an effect either on the syntax attempted or on the errors made. Data have been analyzed to date from nine English-speaking patients capable of giving narratives (preliminary characterization: two moderate Broca’s, one moderate mixed non-fluent, one anomic, and three mild fluent aphasics) and nine Japanese patients (preliminary characterization: moderate Broca’s, mixed non-fluent and anomic), plus ten English-speaking controls and four Japanese-speaking controls. The ten English-speaking controls were healthy normals from ages 50 to 80; the four Japanese controls were healthy middle-aged adults. Not all the Japanese subjects narrated all the stories; the number of responding subjects for each story varies from 3 to 7 aphasics, and from 2 to 4 controls. 3.3 Analysis 3.3.1 Event Analysis There are, of course, many ways of describing a real or pictured event. The event analysis given in Table 1 reflects our subjects’ typical responses: neither normals nor aphasics, for example, describe Frame 1 of “Apple” (Fig. 2d) by saying “The boy is standing on tiptoe under an apple tree with his arm held up in the air”; instead, almost without exception, they interpret his posture in the context of the whole story. For example, normals said, “The boy is trying to get an apple but he can’t reach it”; a moderate agrammatic aphasic said “The boy like to have apple — but apple is too high.” Similarly, no subject describes Frame 3 of “Ball” (Fig. 2a) by saying “The man’s mouth and eyes are wide open and his eyebrows are raised”; instead, they interpret

328

LISE MENN ET AL.

Table 1. Analysis of events in elicitation cartoon frames Story Panel/ event

agent/cause 1st entity affected*

Hat 1

man

man

Hat 2

wind

hat

Hat 3a

wind (by inertia), hat

hat

Hat 3b

man

hat

Hat 4

man

hat

Ball 1 Ball 2 Ball 3a

boy soccer ball boy (by soccer ball window inertia), ball ball lamp

Ball 3c Ball 3d

ball all above

man

Ball 3e

man

man

Ball 3f Ball 4a Ball 4b

man man man

ball

Ball 4c

boy

Race 1

girl

Race 2

other person number

Race 3

girl

boat man

Boat 2 Boat 3a Boat 3b

boat boat man

modality, location time situation inferred**

background propositions ***

walk

present

wear hat, carry cane

man

remove hat

man

roll

present (instant) present

catch cane

land

shoelace

other racers Race 4a girl ? race Race 4b other person prize cup Boat 1a Boat 1b Boat 1c

2nd entity action affected

boat man fish dock man

pick up

attempt, present present

kick present go through, present break (instant) hit past (instant) present move, present look at move, present look at step on present look out be in present mental state be in present mental state

in wind, windy

along sidewalk into water

from water play soccer of house

window various

shoe

tie shoelace present

girl

place number run, pass

present

win transfer

past present

race

move sleep catch

present

on water in boat in boat

girl

prepare for race girl’s back (pat on back)

present

attempt past collide present fall present be in mental present state

(pass relay turn)

into water

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION Apple 1

boy

apple

Apple 2

boy

broom

Apple 3

boy

broom

get

branch

Apple 4a

apple

Apple 4b apple

boy’s head boy

Apple 4c boy

attempt, present pick up past, instant hit present, instant fall past, instant hit present, instant be in mental present state

329 apple high in tree boy happy loosen apple

surprise boy

these features, imputing an emotional state — here, surprise or annoyance. (Perhaps the actual level of description that is most frequently used, e.g., “The boy is trying to get an apple but he can’t reach it”, should be considered a coherent but context-dependent, intersubjective ‘Basic Event Level’, comparable to the Basic Object Level of Rosch’s prototype theory.) The propositions were divided into ‘foreground’ propositions, which were analyzed fully (columns 1-6), and ‘other’ propositions (column 7). These ‘other’ propositions were those that were only given as additions to one or more ‘foreground’ propositions, and seemed to be ancillary to the main story line. (There is of course room for disagreement with this subjective classification; it could be modified by using a naive reader judgement procedure if a more objective method were needed.) Note here that cartoon narratives divide stories into discrete ‘events’, separated by gaps for which intervening information must be inferred. On the one hand, therefore, the narrator of a cartoon strip bears a greater cognitive burden than the narrator of a comparable filmed story in that he or she must interpret conventional cues for motion and imagine the trajectories of objects that ‘move’. On the other hand, the narrator of a cartoon strip is partially relieved of another cognitive burden: he or she is more clearly directed than the narrator of a film as to which events to focus on. The explicitly pictured events are automatically ‘foreground’ while most of the states (pictured or implicit) and the implicit events are ‘background’. The column headers of Table 1 give the semantic relations of the people and objects mentioned or referred to as indicated by the verbs used and by the word order (for English) or the case particles (for Japanese). The semantic descriptors in the column headings (agent/cause, 1st entity affected, 2nd entity affected) were chosen to be independent of the particular syntactic

330

LISE MENN ET AL.

form that a speaker might have used. The second through fourth columns show which person(s) and/or objects in the frame were mentioned or referred to. This includes entities which were referred to by pronouns or zero-anaphora. In the ‘modality/time’ column, all entries are assumed to depict ongoing action (usually encoded by the progressive aspect), unless ‘instant’ is indicated in the column. ‘Present’ means the exact moment depicted, ‘past’ is slightly before that moment. Modality is marked as ‘attempt’ in two frames, Hat 3 (proposition b, man tries to get hat) and Apple 1 (boy reaches for/tries to get apple); all other events that we have considered to be ‘foreground’ are usually interpreted as actually realized, though in a few cases they occur ‘between frames’ and must be inferred (e.g., Hat 3/4 — the moment where the hat hits the water is not shown). Some arbitrariness is unavoidable in all of these decisions: after all, a schematization must be nearly as complex as the [language + context] being schematized in order to capture all the differences between choices of words. In addition, many of the relations between verb and argument — e.g., between ‘win’ and ‘race’ — are not well described in the usual semantic frameworks. The present analysis is therefore intended only as a convenience for comparison of subjects’ responses. 3.3.2 Event Encoding Analysis The event analysis of the pictures was then used as a framework for analysis of the narrative contents. Every clause or phrase produced in response to each frame was coded for the propositions which were used. Zero-anaphora is of course very common in all the Japanese narratives, as in the following sample (for abbreviations used, see Notes to Appendix A). (2)

Hamaguri-san: Tsue o tsukat-te joozu-ni cane PART:OBJ use:PRED-CONJ skilfully:ADV hippari+age-ta. pull up:PRED-PERF using cane — skilfully — ∆ — pulled [it] up)

Zero anaphora is also found, ungrammatically, in the narratives of more severe English-speaking aphasics (Mr. Zebra, Hat 4: “cane...then pull out water”), and, grammatically, in conjoined sentences of English-speaking

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

331

normals (Mr. Badger, Race 3/4: “She runs the race and wins first prize”). The number of descriptors that the subjects gave for each frame reflects the complexity of the event pictured, as well as its centrality to the narrative: Frame 1 of Ball (Fig. 2a), for example, was always described with the boy as subject of an active verb (either the concrete verb ‘kick’, the more general verb ‘play with’ or the more interpretive ‘play soccer’), but Frame 3 of Ball was described by following the ball’s trajectory, what happens to the lamp, the man’s reactions, or various combination of these. Propositions were further coded for whether they were simple descriptions (‘The man stands in front of the window’) or involved inferences such as mental state (‘The man looks out to see who did it’), and whether they included projections of the probable consequences of the pictured events (‘The lamp breaks’). 3.3.3 Viewpoint Analysis From the subjects, verbs, and verb voices used by our narrators, the various viewpoints they expressed were determined, to the extent possible. Viewpoint analysis is a matter involving a great deal of interpretation, and we regard our attempt in this area as only exploratory. The reasoning used is based on the discussion given earlier: empathy, as an affective mental state of the speaker, may or may not be encoded in a particular utterance; moreover, as Kuno indicates, an utterance may be made from a neutral point of view. Therefore, there may be no overt indication of viewpoint, either because the speaker fails to mark his or her viewpoint, or because he or she takes a neutral point of view. It follows that we cannot always determine the speaker’s attitude from looking at what he or she has said. In analyzing the viewpoints expressed by our four subject groups, we have therefore divided utterances into two classes: those where there is some overt indication of the viewpoint of the speaker (marked empathic focus), and those in which the speaker might be neutral or might be taking the viewpoint of the grammatical subject of the sentence (unmarked empathic focus). We have taken the following pragmatic, semantic and syntactic phenomena as putative markers of empathic focus: explicit attribution of mental state or mental contents (the boy is happy; the man is scolding), imputation of effort or intention (the boy is trying; the man is looking to see who did it), evaluation of good or bad fortune (unluckily; the man had an accident), judgement of

332

LISE MENN ET AL.

quality of performance (carefully), topicalization, passivization, use of deictic verbs (the ball comes in), and direct discourse (Hey! Sorry!). Aphasic speakers produced a few semantically ill-formed active-voice sentences in which the undergoer is the subject (‘He hits on the head — it — the apple’); these we considered as indicating an empathic focus which the speaker was unable to encode grammatically. We have not taken order of mention as a marker of viewpoint unless topicalization or passivization has occurred. This is partly because ‘order of mention’ is difficult to evaluate in Japanese. A protagonist is often mentioned explicitly only once in a story by the normals. And as the examiner often provides the first mention of the protagonist for the aphasic subjects, they may not provide any explicit references to the protagonist at all.2 If there was no marking of empathy such as those just listed, subjectinitial sentences and sentences with zero-anaphor subject were considered to be unmarked for empathic focus. The viewpoint was taken as indeterminate between ‘neutral’ and the viewpoint of the entity encoded as the main clause subject, whether explicit (‘the ball breaks the window) or implicit in the verb (‘. . . and falls’). This procedure was applied uniformly to all speakers. When no subject could be determined because only a fragmentary noun phrase or sound effect (“bump!”) was produced, no viewpoint was attributed.

4.

Results

4.1 Characterization of Aphasic Narratives The individual aphasic subjects in general transmitted less information than the normals. Specific error types found in both languages were omissions/ fragmentary encodings (e.g., omission of verbs), failed attempts to integrate multiple propositions into a clause, and word-choice errors (including perseverations from one story into the next). Examples of these error types are as follows: English 1. Fragmentary encoding/omission He tried. [Mr. Horse, fluent, Hat 3]

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION 2.

333

Failed attempts to integrate multiple propositions syntactically fragmentary: (2)

a. Walking and a cane and a hat [Mr. Moose, mod. non-fluent, Hat 1] b. Then the one — ’round — water. spin the — ’round — water [Mr. Zebra, non-fluent, Hat 3] c. Cane — then pull out water [Hat 4]

syntactically smooth: d. He’s getting his cane to put it on, he’s getting it out that way. [Mr. Wolf, severe anomic, Hat 4] 3.

Word-choice error (3)

a. Wind comes and pulls his hat away [Mr.Wallaby, mod. fluent, Hat 2] b. He pulls his scarf into the water [Hat 3]

Japanese 4. Fragmentary encoding/omission (4)

5.

a. Yuushoo. [Ms. Tampopo, mixed, Race 4] victory b. Soshitara “kora.” [Mr. Kamoshika, mixed non-fluent, then:CONJ “Hey!” Ball 4] c. Otoko no ko ga sakkaa man PART:POSS child PART:NOM soccer o.... [Ball 1] PART:OBJ d. Mado o, mado o kashan. window PART:OBJ window PART:OBJ sound of crash [to waru] [Ball 2]

Failures to integrate multiple propositions into a clause (5)

Gashan to kodomo ga . . . sound of crash PART:QUOT child PART:NOM [Mr. Kamoshika, mixed non-fluent, Ball 4]

334 6.

LISE MENN ET AL. Perseverations and other word-choice errors (6)

a. kodomo ga, booru ga, booru o child PART:NOM ball PART:NOM ball PART:OBJ tor-u, tor-u. take:PRED IMPF [Ms. Tampopo, mixed non-fluent, Perseveration, Ball 1] (The Ball series was presented immediately after the Apple series; the perseveration is the use of the verb toru ‘get, take’, appropriate to the ‘Apple’ story.) b. Otoko no hito ga booru o man PART:POSS person PART:NOM ball PART:OBJ hum-u. step on:PRED-IMPF [Ball 3, Word-choice error]

4.2 Additional Error Types Possible Only in Japanese Japanese patients used ellipses (grammatically permissible omissions, Kuno 1978) which did not seem to be pragmatically appropriate in the discourse context. In 9 out of 25 sessions with the patients, the main characters in the pictures, which are typically expressed as the subjects of the first utterances, were not expressed even when they were not provided by the examiner. In none of the sessions with the controls, on the other hand, did such failure to mention the main characters occur. (Note: because examiner and patient are looking at the pictured stimuli together, the ‘shared information’/’new information’ contrast is blurred, and determination of which ellipses are permissible is correspondingly difficult.) Frequent revisions of particle choices by Japanese patients indicate much difficulty in clause planning. In many instances, the verb was not produced, making it impossible to be sure whether the case particles had been chosen correctly. Occasionally, as in one of our previous studies (Morishima, Menn, Fujita, and Sasanuma 1992), no verb exists that would have been compatible with the case particles chosen.

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION 7.

335

Particle Revisions (7)

hooki o sono ringo no ki broom PART:OBJ that:DEM apple PART:POSS tree ni, ringo no ki o hooki PART:GL apple PART:POSS tree PART:OBJ broom de, u-, hooki de, PART:INSTR uh, broom PART:INSTR hooki o, hooki de butsuke-te broom PART:OBJ broom PART:INSTR knock:PRED-CONJ [Mr. Oshidori, Wernicke, Apple 3]

4.3 Viewpoint Results We found a high degree of similarity of viewpoints across all four subject groups. The same was true of the use (or attempted use) of marked-viewpoint forms. (Recall that our putative markers of empathic focus were: explicit attribution of mental state or mental contents, imputation of effort or intention, evaluation of good or bad fortune, judgement of quality of performance, topicalization, passivization, use of deictic verbs, and use of direct discourse.) Combining the English and Japanese results, Figure 3 shows that aphasics’ choice of marked vs. unmarked forms for a given frame was highly correlated with the markedness choice made by normals. If we consider the use of unmarked forms only (open squares and circles), the correlation is .82; if we consider the use of marked forms only (black squares and circles), the correlation is .80. These results are of course two sides of the same coin: if subjects use more marked forms, they use fewer unmarked ones. (The correlations differ and the percent of marked and of unmarked forms used for each frame does not always sum to 100% because not all subjects gave a response for each frame.) Figures 4a-8a show the same data, but graphed in such a way as to permit comparison of markedness choices with the corresponding cartoon frames. Note especially the shift to marked forms in Boat 3, Race 4, and Ball 4. However, the particular syntactic and lexical devices used to mark the viewpoints differed. For example, quite a number of the more impaired patients marked their viewpoint by the use of direct discourse:

Figure 3. Marked vs. Unmarked Empathic Focus for 5 Stories

336 LISE MENN ET AL.

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

337

8. (8)

Ntoo nto ne “go men ne.” well:INTERJ well:INTERJ PART:SFL “sorry PART:SFL” [Ms. Tampopo, mixed non-fluent, Ball 4]

A few of the deictically marked forms are interesting to examine further. In Apple Frame 4, three of 4 normals used ochite kuru ‘comes falling’, rather than just ochiru ‘falls’. For one of these, atama ni ‘on (his) head’ was the first phrase, but two of them began the statement with ringo ga ‘apple (subj)’. This could indicate competing perspectives; but as this word order seems completely natural to native speakers of Japanese, we suggest instead that ochite kuru indeed indicates taking the boy’s viewpoint. The constituent order reflects the additional factors of cause/effect iconicity, picture scanning order, and the fact that the boy’s desire for the apple, rather than just the boy, may be regarded as the topic of the story. The fourth control used the perspective-neutral ringo ga atama no ue ni ochimashita ‘the apple falls onto (his) head’. As for the patients, 6/7 mentioned ringo ‘apple’ first (usually marked correctly with ga/wa, but once with o). Four of the 6 produced the verb; two of the 4 also used ochite kuru ‘comes falling’, and the other two who expressed the verb used forms of ochiru ‘fall’. 4.4 Encoding Choice Results The choice of which propositions to encode was very similar over members of all four subject groups.3 Overall, the aphasics and the control subjects were also very similar in the way they chose to assign syntactic roles (subject, object) to the participants in the stories (main character, subsidiary character). The choice of which referents to code as topics — either by topicalization or by use as subject — was again similar; we have graphed this set of choices for normals versus aphasics, combining the results from English and Japanese in Figures 4b-8b. Note the switches back and forth among ‘man’, ‘hat’, and ‘wind’ in Fig. 4b; between the girl and the other person in Fig. 5b, Frame 2 (but not Frame 4); among the boy, the ball, and the man in Fig. 6b; and so on: again, the frame-by-frame similarity between the aphasics and the normals in their encoding choices is overwhelming. Much of the linguistic discussion of orientation and empathy has centered around the choice of active vs. passive voice. In the English data, a few patients and normals used get-passives in Race 2, and three patients used

Figure 4b. Referents encoded as Subject or Topic - “Hat”

Figure 4a. Marked vs. Unmarked Empathic Focus - “Hat”

338 LISE MENN ET AL.

Figure 5a. Marked vs. Unmarked Empathic Focus - “Race”

Figure 5b. Referents encoded as Subject or Topic - “Race”

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION 339

340

LISE MENN ET AL.

Table 2. Animacy vs. Markedness, all subjects. .5 assigned to each possible referent of two-way ambiguous referring expressions. Story Panel/event Hat Hat Hat Hat Hat

1 (walk) 2 (blow off) 3a (roll) 3b (chase) 4 (retrieve)

Ball 1 (kick) Ball 2 (break/go thru) Ball 3a (enter/hit lamp Ball 3b (lamp fall) Ball 3c (affect man) Ball 4a (stand/look) Ball 4b (man’s mental state/speech Ball 4 epilog (boy’s state) Race Race Race Race Race

1 (tie) 2 (get number) 3 (race) 4a (win) 4b (give prize)

Boat Boat Boat Boat Boat

1a (move) 1b (sleep) 1c (fish) 2 (hit dock) 3 (fall out)

Apple Apple Apple Apple Apple Apple Total

1 (try to get) 2 (pick up) 3 (hit branch) 4a (fall) 4b (hit head) epilogue

With empathic marking

Without empathic marking

Animate

Animate

Inanimate

Inanimate

man 18 man 2.5

hat .5

wind 15, hat 7 hat 20, wind 2

man 1 man 3

man 8 man 22

man 2

boy 29 ball 1, glass 1 boy 6 ball 2

boy 5 ball .5 man 10 man 3 man 23, boy 1

man 6 man 15

boy 2

boy 1

girl 1 girl 9 girl 2 gril 13

girl girl girl girl girl

man 1 man 7

man man man man

boy 12 boy 1 boy 1 boy 2 boy 1

apple 1

98

6

ball 21, glass 4 ball 13, lamp 1 ball 6, lamp 1

27 7, other 7 20, girls 2 18 1 23 14 1 11

boy 16 boy 18 boy 30 boy 1 boy 4 305

boat 18 ìmpact 6, boat 2 broom 5 apple 23 apple 13 158

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

341

variants of ‘is thrown out’ in Boat 3. However, in the Japanese data, neither patients nor normals made much use of passive voice (one for patients, five for normals) or OSV order (one instance for each group). Both groups often began utterances with locative expressions. We next examined the effect of animacy and other factors on topic choice, as shown in Table 2. Several factors seemed to control topic choice across the four subject groups, as can be determined from comparison of Table 2 with the cartoon pictures. Unsurprisingly, Animates (columns 1 and 3) were chosen as subjects/topics most of the time: 403 animates vs. 164 inanimates. When an Inanimate (columns 2 and 4) was chosen as subject/topic of a particular scene, it was most likely to be a cause (wind, ball, boat, impact of boat, apple) or a freely moving object (ball, lamp, hat, boat, apple). However, a few items coded as subjects/topics are taken from existentials, which introduce them because they are about to play important roles in the story (water, broom). Several others are inanimate objects heavily affected by the action, namely, the broken window and the falling lamp. These factors are summarized in Table 3. One might expect that inanimate subjects/topics would never be marked as empathic foci. However, our definition of ‘marked-viewpoint forms’ allows for this as a logical possibility: inanimates might be topicalized, used with deictic verbs, or used as the subjects of passive-voice verbs. They might also be used as subjects of metaphorical expressions in which the inanimate object is treated as having a mental state. Comparison of the subjects/topics which were marked for empathic focus (according to our definition) with the subjects/topics which are unmarked and might be neutral gives a striking result. While 34% of the subjects/topics of the unmarked clauses are inanimate, only 6% of the subjects/topics of the marked clauses are inanimate: 3.5 references to the ball and 1 to the apple, which are freely moving causes; .5 to the hat, a moving undergoer; and 1 to the broken window glass. (There were two ambiguous referring expressions, which were counted as .5 for each possible referent.) Four of the six were marked by deixis, 1.5 were passivized, and .5 was marked by the judgmental phrase ayamatte “accidentally”; no metaphors were used. Again, aphasics and normals were very much alike in the choices of what to encode as subject/topic; the comparison between the groups is given in Table 4.

342

LISE MENN ET AL.

Table 3. Factors in choice of inanimate subject/topic ? indicates indeterminate Item chosen as subject/topic

moving

cause

future cause

damaged/ endangered

+ + + + – + + +

– – – + – – – + – – + –

+ – + – ++ – + ?

? – +

– + – + – + ? + + + – +

+ + – +

– + – +

– + – –

+ – ++ –

Unmarked hat, Hat 2 wind, Hat 2 hat, Hat 3 ball, Ball 2 glass, Ball 2 ball, Ball 3 lamp, Ball 3 boat, Boat 2 impact, Boat 3 boat, Boat 3 broom, Apple 2 apple, Apple 4

? – ?

Marked hat, Hat 2 ball, Ball 2 glass, Ball 2 ball, Ball 3

Table 4. Aphasic and normal speakers’ choices of animate and inanimate subject/topic With empathic marking

Without empathic marking

Animate

Animate

Inanimate

Inanimate

Aphasic speakers Normal speakers

53 45

3 3

144 161

71 87

totals

98

6

305

158

4.5 Animacy and Clause Structure There were no completed subject-object reversal errors in Japanese or English in these five story contexts, in spite of the fact that an ‘animacy strategy’ would have worked in only half of the frames, namely those in which an animate being acts on (or attempts to act on) an inanimate object: Ball Frame

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

343

1 ‘boy kicks ball’; Race Frame 1 ‘girl ties shoe’; Apple Frame 1 ‘boy reaches for apple’; Frame 2 ‘boy picks up broom’, Frame 3 ‘boy hits branch with broom’; Hat Frame 4 ‘man retrieves hat’. There were no apparent problems of subject choice in the Animate-Animate frames (Race 2, 3, 4), even though two of these were complicated by being ‘dative’ events (i.e., events in which possession of an object is transferred from person to another). With the protagonist recipient, normals and aphasics in both languages successfully used active-voice ‘receive/accept’ verbs (Race 4: “She gets a prize”). In a few instances, both control and aphasic subjects used the secondary character as the sentence subject (e.g., Race 2: “Somebody gave her a number”). The Inanimate-Inanimate frames posed some problems for some of the Japanese aphasics who tried to begin with the object affected (the window in Ball Frame 2, the hat in Hat Frame 2), but English speakers had fewer problems with them. We think that this is because the syntax and morphology of English, in these particular instances, made it possible for English speakers to deal with these frames in the active voice. In the case of the window, most of our patients, like the normals, followed the ball’s trajectory through the window or took the boy as subject (Mr. Hyrax: “an’ he puts it through the plate window”); in the case of the hat, several of them used the intransitive construction “His hat blows off.” Although there were no full reversal errors, some attempts at InanimateAnimate frames indicated problems. A clear example comes from one of the few aphasic sentences which had incompatible subject and verb, e.g., Mr. Wallaby, “He hits on the head” for the final frame of the Apple strip, Fig. 2a. These are of particular interest, as they appear to result from blends of two sentence structures. In this example, the patient starts with the boy as subject, but continues with a predicate that follows the trajectory of the apple. Since this error could be seen more prosaically as substitution of active ‘hits’ for get-passive ‘gets hit’, it is worth pointing out that Mr. Wallaby was a rather mildly impaired fluent aphasic who had few functor problems: for example, he began his narrative by saying “Here he is trying to reach the apple, if it’s an apple.” In some cases, Japanese normals used rather different constructions from Japanese aphasics. For example, in Race Frame 2, intended to show the girl getting a number put on her back, 3/4 normals said zekken o tsukete iru ‘she has a number’. In this phrasing, the action of placing the number is not explicit; the result of the action is presumably more important than the action in the context

Figure 6a. Marked vs. Unmarked Empathic Focus - “Ball”

Figure 6b. Referents encoded as Subject or Topic - “Ball”

344 LISE MENN ET AL.

Figure 7b. Referents encoded as Subject or Topic - “Boat”

Figure 7a. Marked vs. Unmarked Empathic Focus - “Boat”

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION 345

Figure 8a. Marked vs. Unmarked Empathic Focus - “Apple”

Figure 8b. Referents encoded as Subject or Topic - “Apple”

346 LISE MENN ET AL.

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

347

of preparation for the race. The remaining normal used ....zekken o tsukete morau, roughly ‘she gets a number’, which has the function of a passive beneficial to undergoer. Patients either had some difficulty in interpreting the frame (e.g., nambaa o miru ‘looks at a number’; undoo saseru ‘makes her do exercise’) or in finding words to encode the event (e.g., zekken o dasu roughly ‘shows a number’ [lit: makes a number appear]). Only two of 7 patients used either of the constructions used by the normals. 4.6 Summary of Findings Summing up all these observations obtained from our analysis of these data, we may conclude as follows. First, the striking similarity of viewpoints across all four subject groups shown in Figure 3 suggests that aphasics, both in English and in Japanese, take fundamentally the same point of view as do normals in both languages. This means that, in producing an utterance in response to a series of cartoon pictures, aphasic speakers’ capacity to perceive empathically is within normal limits. Second, the interaction of animacy with markedness and subject/topic choice suggests an encoding hierarchy. Humans are strongly preferred as subject/topic (all animates in these stories are human), and are occasionally chosen even when they are not visible in the frame being described (e.g., ‘He hits the dock,’ Boat 2). As indicated in Table 3, when an inanimate is chosen as subject/topic, it is most likely to be an entity with causative power, in motion, not under current human control (wind, the ball in flight, the apple falling). Being in free motion and an object of concern (the hat rolling, the lamp) also encouraged choice of an inanimate as subject/topic; so did being a severely affected object of value, the broken window. Explicit syntactic or lexical indications of empathic focus are overwhelmingly reserved (94%) for animates. The hat, the ball, the apple, and the broken window were the only inanimates that were ever elevated to the status of empathic foci, and this happened very rarely.

5.

Discussion

A pervasive but often unacknowledged problem in functional syntax is the following: functional syntax attempts to develop explanations for syntactic

348

LISE MENN ET AL.

choices and syntactic interpretations in terms of concepts like ‘topic’, ‘focus’, ‘about’, ‘information’ etc. These notions appear to refer to non-linguistic concepts, and the explanations thus appear to make claims about the relation between these concepts and specific linguistic choices. Therefore, logically, testing these explanations requires that we determine the non-linguistic state of affairs independently of the linguistic facts. Only then can we test whether the claimed correlation holds between the non-linguistic state and its proposed linguistic expression. However, when we analyze narratives apart from an analysis of the realword events being encoded, we have no independent information as to what happened. Even when we know what the stories are, as in this and other experimental studies, we have no independent evidence as to the speaker’s state of mind (or what state of mind the speaker wishes to create in the hearer). One way to deal with this is to redefine these notions operationally in linguistic terms, as many functionalists have done (e.g., Givón 1984), but then their explanatory power, being entirely intra-linguistic, is diminished. An alternative is to treat them as psychological primitives, and then attempt to manipulate them experimentally, in narratives of varying degrees of naturalness and control, as in our work, in the studies by Karmiloff-Smith, by Sridhar and by Bates and Wulfeck cited above, and in the ongoing cross-linguistic “Frog Story” studies of children’s narratives by Slobin and associates. Consider ‘empathy’ in its ordinary sense of a feeling of identification with a person, implying the shared viewpoint. What stimulus factors might one try to manipulate as experimental variables in order to increase a narrator’s empathy with a cartoon character? Everyday experience as consumers of comic strips and cartoon films, and now video games, suggests to us that contributory factors include animacy, volitionality, consciousness, power, perceived attractiveness, and perceived similarity to the narrator on any of a large number of social dimensions. The visible manifestations of the first four factors include motion, effect on other entities, and portrayed emotional reactions. It seems likely that ‘perceived similarity to narrator’s self’ is the rubric under which all of the above can be summarized. This is supported by informal observation: different people react quite differently to the various characters in a complex story, and the tendency to share the viewpoint of the character most like oneself is quite evident. However, this ‘perceived similarity to self’ cannot be equated with objectively measured similarity to the

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

349

narrator; our love for powerful heroes suggests that some idealized self is the one from which distance is measured. In addition, biological ‘releasers’ — that is, physical properties that produce automatic reactions in the observer, such as large eyes with large pupils — also play a role in elicting empathy, and they are consciously manipulated by popular artists. (Disney characters have increased eye-size over the years; and Schodt (1983: 91) notes “... by far the most striking visual aspect of [Japanese] girls’ comics is the orblike eyes of the characters...Both males and females in the comics are given this treatment, although the eyes of the males are somewhat smaller, as are those of cold, evil people.” A systematic research program, then, would use narration of sets of animated cartoons or cartoon strips as stimuli, and systematically manipulate variables like similarity to narrator and eye size, looking for corresponding changes in linguistic responses to examine subconscious empathy. Perhaps other methods (questionnaires — ‘Who was your favorite character?’) could be used to examine conscious empathy. A final meta-consideration on the problem of defining (or refining) ‘empathy’ as a technical term: Whenever a term in general usage is adopted as a term of art, its boundaries (even if they remain fuzzy) are sharpened and shifted; when a term defined in one area of science makes connections with notions in a related area, similar shifts are inevitable. This paper attempts to take the study of empathy in language beyond the linguistic basis given to it by Kuno, by providing it with a new foundation in cognitive psychology. As we do so, we face choices. If empathy in the non-technical sense has to do with a sense of likeness-to-self, how far should it extend as a technical term? In particular, should a technical notion of empathy extend to inanimate objects? In everyday behavior, there is evidence that it should: some speakers at moments of stress certainly go beyond ‘Oh, the poor dog!’ to ‘Oh, my poor car!’; they respond with ‘Ouch!’ to the sound of a collision, and cheer the ‘efforts’ of a tow-truck or a construction crane. (Note our factors of damage, motion, and causal efficacy in these imagined but plausible examples.) The approach taken in the present paper, as indicated in section 2.3. above, is to treat the use of certain marked syntactic and lexical devices as linguistic evidence that a particular referent is the empathic focus of the sentence. The devices included passives, get-passives, expressions of judgement and mental state, and deictic expressions. However, as we stated, there is no generally agreed-upon list. More to the present point, if we are trying to

350

LISE MENN ET AL.

put the notion of empathy on a psychological basis, any such list must be tentative. Whether a particular linguistic expression in fact reliably indicates a particular attitude of the speaker (or the attitude that the speaker wishes to be perceived as holding) is an empirical matter. Furthermore, it is a matter that is difficult to test psychologically, although that is what we wish to do by manipulating variables like volitionality, animacy, attractiveness, and value in our drawings. Obtaining experimental support for linguistic intuitions is a very slow and clumsy enterprise, requiring continual revision and refinement of the notions involved. But this is the typical course of scientific development, and it is how linguistics will become a real part of cognitive science. In review: our work with aphasic patients drew attention to the potential utility of an ‘empathy’ concept, and then to problems in employing it as a purely linguistic construct. The fact that the initial subject population is language-disordered has thus become almost incidental. But it is precisely the work with an exceptional population that has made plain the need for deeper analysis of a theoretical notion. We hope that, in the near future, our experimental analysis of abnormal language will make a substantial contribution to the study of functional syntax in general, and to the cognitive and linguistic notions of empathy in particular.

Notes 1.

Viewers of animated films impute animacy and volition on the basis of little more than motion cues: Russell Tomlin at Oregon has been working with computer programs in which two outlined fish move towards one another from opposite sides of a computer screen, one opens its mouth, the other disappears into the mouth, and the first closes its mouth again.

2.

It is even harder to analyze order of mention in natural Japanese conversations than in these narratives, because of the extent of the ellipses: conversations studied by author M.H. have more zero anaphora than the picture narratives of our normal subjects.

3.

A language-specific exception to this is that English speakers have the option of expressing within one clause the trajectory of an entity which is encoded as the object of a sentence [‘the boy kicks the ball through the window’]; this is not natural in Japanese.

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

351

References Bates, Elizabeth, and Beverly Wulfeck 1989 “Crosslinguistic studies of aphasia.” In B. MacWhinney and E. Bates (eds), 328-371. Bock, J. Katherine and Helga Loebell 1991 “Framing sentences.” Cognition 35: 1-39. Chafe, Wallace 1976 “Givenness, contrastiveness, definiteness, subjects, topics, and point of view.” In C. N. Li (ed), 25-55. Dorian, Nancy (ed) 1984 Investigating obsolescence: Studies in language contraction and death. Cambridge: University Press. Fletcher, Paul and Michael Garman (eds) 1986 Language acquisition (2nd ed.). Cambridge: Cambridge University Press. Greenfield, Patricia M. and Joshua H. Smith 1976 The structure of communication in early language. New York: Academic Press. Givón, Talmy 1984 “The pragmatics of referentiality.” In D. Schiffrin (ed), 120-138. Hasegawa, Tsuneo, Keiichi Wada, Ichiro Tsukada, Aiko Takeuchi, and Asako Wada 1977 Standard language test of aphasis. Tokyo: Homeido. Karmiloff-Smith, Annette 1979 A functional approach to child language. Cambridge: Cambridge University Press. Karmiloff-Smith, Annette 1986 “Some fundamental aspects of language development after age five.” In P. Fletcher and M. Garman (eds), 455-474. Kuno, Susumu 1978. Danwa no bumpoo. Tokyo: Taishukan. Kuno, Susumu 1987 Functional syntax: Anaphora, discourse, and empathy. Chicago: The University of Chicago Press. Kuno, Susumu and Etsuko Kaburaki 1977 “Empathy and syntax.” Linguistic Inquiry 8: 627-672. Levelt, Willem 1989 Speaking. Cambridge, MA: MIT Press. Li, Charles N. (ed) 1976 Subject and topic. New York: Academic Press. Linebarger, Marcia. C., Myrna F. Schwartz, and Eleanor Saffran 1983 “Sensitivity to grammatical structure in so-called agrammatic aphasics.” Cognition 13: 361-392. Menn, Lise and Loraine K. Obler (eds) 1990 Agrammatic aphasia, Vol. II. Amsterdam and Philadelphia: John Benjamins.

352

LISE MENN ET AL.

Menn, Lise and Loraine K. Obler 1982 “Exceptional language data as linguistic evidence: An introduction.” In L.K. Obler and L. Menn (eds), 3-14. Menn, Lise and Loraine K. Obler 1990 “Conclusion: Cross-language data and theories of agrammatism.” In L. Menn and L.K. Obler. (eds), 1369-1389. MacWhinney, Brian and Elizabeth Bates (eds) 1989 The cross-linguistic study of sentence processing. Cambridge: Cambridge University Press. Menn, Lise 1989 “Some people who don’t talk right: Universal and particular in child language, aphasia, and language obsolescence.” In N. Dorian (ed), 335345. Morishima, Yasunori, Lise Menn, Ikuyo Fujita, Akio Kamio, and Sumiko Sasanuma 1992 “Bunsanshutsu no ‘on-line processing’: nihongo oyobi eigo ni okeru shitsugoshoo kara no shoorei (Online processes of sentence production: Evidence from aphasic speakers of Japanese and English)”. Proceedings of 9th Annual Meeting of the Japan Society for Cognitive Science, 106107. Rumelhart, David E., James L. McClelland, and the PDP group 1988 Parallel distributed processing. Cambridge, MA: MIT Press. Obler, Loraine K. and Lise Menn (eds) 1982 Exceptional language and linguistics. New York: Academic Press. Ono, Tsuyoshi, and Suzuki, Ryoko 1992 “Word order variability in Japanese conversation: Motivations and grammaticization.” Text 12: 429-445. Saffran, Eleanor, Myrna F. Schwartz, and Oscar S.M. Marin 1980 “The word order problem in agrammatism: II. Production.” Brain and Language 10: 263-280. Schodt, Frederick 1983 Manga manga: The world of Japanese comics. Tokyo and New York: Kodansha International. Schiffrin, Deborah (ed) 1984 GURT ’84: Meaning, form, and use in context linguistic applications. Washington, D.C.: Georgetown University. Sridhar, S.N. 1989 “Cognitive structures in language production”. In B. MacWhinney and E. Bates (eds), 209-224. Wulfeck, Beverly 1988 “Grammaticality judgements and sentence comprehension in agrammatic aphasia”. Journal of Speech and Hearing Research 31: 72-81.

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

353

Appendix A: Sample narratives (The “Hat” story) Normal English speaker (Ms. Bluebell) BLU: A man’s walking down the street happily and the wind comes along and blows off his hat... he starts to chase it and it’s blowing over and over and oh... it almost goes in the water but he hooks it with his cane. Aphasic English speaker (Mr. Wallaby; mod. fluent) WAL: Wind comes and pulls his hat away. Pulls — he pulls his scarf into the water and he picks it up with his — with his cane with his with his with a cane. Normal Japanese speaker (Ms. Murasaki) MUR: otoko no hito ga man PART:POSS person PART:NOM

michi o street PART:OBJ

arui-te i-mashi-ta. walk:PRED-CONJ be:PRAUX-PRAUX:POL-PERF “(A) man was walking down (the) street.” MUR: totsuzen kaze suddenly:ADV wind

ga hui-te ki-te PART:NOM blow:PRED-CONJ come:PRAUX-CONJ

booshi o tobas-are-mashi-ta. hat PART:OBJ blow:PRED-PRAUX:PASS-PRAUX:POL-PERF “Suddenly the wind blew and (he) had (his) hat blown off.” MUR: booshi wa kawa no hat PART:TOP river PART:POSS ochi-soo fall:PRED-seem:AN

naka e inside PART:DIR

ni nari-mashi-ta PART:GL become:PRED-PRAUX:POL-PERF

ga otoko no but:PART:CONJ man PART:POSS

hito wa person PART:TOP

mot-te carry:PRED-CONJ

sutekki cane

i-ta be:PRAUX-PERF

de PART:INSTR

hiroi-age-mashi-ta. pick+raise:PRED-PRAUX:POL-PERF “Though (the) hat almost fell into (the) river, (the) man picked (it) up with (the) cane that he carried with him.” Aphasic Japanese speaker (Mr. Hamaguri; mixed) EXA: koko ni otoko no hito ga i-ru desh-oo. koko kara chotto tsuzuki it-te mi-te kudasai. “Here is (a) man, ok? From here, please tell me (the story).” HAM: tsukk- tsue o sui-te i-ru. cane PART:OBJ poke:PRED-PART:CONJ be:PRAUX-IMPF “(He) is walking with a cane.”

354

LISE MENN ET AL.

HAM: kara booshi ga and:CONJ hat PART:NOM

kabu-tte wear:PRED-PART:CONJ

i-ru. be:PRAUX-IMPF “And (a) hat is wearing.” HAM: ndee byuu tte fu- fu- fui-te and:CONJ sound.of.wind PART:QUOT blow:PRED-PART:CONJ ku-ru. fui-ta come:PRAUX-IMPF blow:PRED-PERF

kaze - kaze ga wind wind PART:NOM

fui-ta. blow:PRED-PERF “And blows whoosh. Wind that blew - The wind blew.” HAM: nde kondo and:CONJ this.time

wa PART:TOP

tobas-are-chat-ta. blow off:PRED-PRAUX:PASS-PRAUX- PERF “And then, has been blown off.” HAM: nde kondo wa okkot-ta and:CONJ this.time PART:TOP fall:PRED-PERF

okkottfall:PRED-

okkochi-ru - okkot-ta ja fall:PRED-IMPF fall:PRED-PERF be:PRAUX naku-te ne ... not:PRAUX-PART:CONJ PART:SFL “And then, fell, fell, fell, it’s not ‘fell’ ...” EXA: doo shi-ta? okkot-ta-no? “What happened? Did (it) fall?” HAM: okkot-ta. fall:PRED-PERF “(It) fell.” HAM: tsue o tsukat-te cane PART:OBJ use:PRED-PART:CONJ hippari-age-ta. pull:PRED-lift:PRAUX-PERF “With (the) cane, pulled (it) up skilfully.”

joozu-ni skilfully:AN-be:PRED

THE ROLE OF EMPATHY IN SENTENCE PRODUCTION

355

Notes on conventions of interlinear morphemic translations These abbreviations are derived from those used in Sasanuma, S., Kamio, A., and Kubota, M. 1990. “Agrammatism in Japanese: Two Case Studies,” in Menn, L. and Obler, L. K., Agrammatic Aphasia: A Cross-language Narrative Sourcebook (Amsterdam: John Benjamins) and those used in MacWhinney, B. 1991. The CHILDES Project: Tools for Analyzing Talk (Hillsdale, NJ: Lawrence Erlbaum Associates). ADV AN CONJ DEM DIR IMPF INSTR NOM GL OBJ

= = = = = = = = = =

adverb adjectival noun conjunctive demonstrative direction imperfective intrument nominative goal object

PART PASS PERF POL POSS PRAUX PRED QUOT SFL TOP

= = = = = = = = = =

particle passive perfective polite possessive predicate auxiliary predicate quotative sentence final particle topic

PERSONAL PRONOUN SHIFT IN JAPANESE

357

Personal Pronoun Shift in Japanese A Case Study in Lexical Change and Point of View John Whitman Cornell University

1.

Introduction

There have been two influential uses of the notion of EMPATHY in linguistic research partly or mainly inspired by Japanese language data over the past three decades. One is in the work of Susumu Kuno studying the effect on linguistic form of speaker identification with various speech act participant and syntactic roles. The other is the notion of EMPATHETIC IDENTIFICATION found in the writings of Takao Suzuki. Suzuki discusses instances in Japanese where choice of address form appears to involve identification of the speaker with a potential speech act participant other than herself. Both uses are based on the idea that a speaker naturally ‘empathizes’ with certain (actual or potential) speech act participants, less naturally with others,1 and that the direction of empathy, the orientation of the camera angle in Kuno’s terms, has concrete linguistic consequences. Both can thus be described as theories of how linguistic form is shaped by communicative context, and in this sense at least both are functionalist theories. The purpose of this paper is to explore the very fundamental difference between these two modes of functionalist explanation as possible accounts for a phenomenon brought to our attention by the historical syntax of Japanese: the shift of a pronoun from one personal category to another.

358 2.

JOHN WHITMAN Intrapersonal Pronoun Shift

A striking fact about the history of Japanese is the frequency with which pronouns2 shift over time to designate different speech act participants. Ware (wa- ‘ego’ + -re pronominal suffix), for example, occurs in the earliest texts of the 8th century as a speaker-designator: (1)

Kojiki (song 49) Susuko1ri ga kam-i1-si mi-ki1 ni Susukori NOM brew-RY -PAST.RT HON-sake on ware weφ-i1-n-i-ke1r-i I drunk-RY-PRF-RY- PAST-SS ‘On the fine sake that Susukori has brewed, I (ware) got drunk’ ko2to2-na-gusi we-gusi ni ware weφ-i1-n-i-ke1r-i matter-none-sake laugh-sake on I drunk-RY-PRF-RY-PAST-SS ‘On that blameless sake, that laughing sake, I (ware) got drunk’

In this song the emperor praises the Korean winemaker Susukori. Ware designates the drinker and the speaker (the emperor), the I-singer of the verse. This is the primary function of ware identified in dictionaries and handbooks of the classical language. The function remains standard into the Kamakura period, but in the stories of Konjaku monogatari (1106) there are a few instances of ware functioning as a hearer-, not a speaker-designator: (2)

ware ni mo tug-e maus-i ni φito wo (KM, 27.32) you DAT also tell-RY say-RY DAT person ACC tukaφas-i-tar-i-si-ka-ba send-RY-PERF-RY-PAST-IZ-COND ‘Since I had sent someone to inform you (ware) as well...’

In (2), the referent of ware is an equal’s wife; the deferential auxiliary mausu (deferential ‘say’) indicates that its function is not derogatory or ‘lowering’; but it clearly designates the hearer. In the Uji-shûi monogatari (1218) ware appears as a hearer-designator with a stronger coded nuance: (3)

ware φa miyako no φito ka. iduko φe you TOP captial GEN person Q where to oφasu-ru zo. go(HON)-RT EMPH ‘Are you from the capital? Where are you going?’

(UM, 10)

PERSONAL PRONOUN SHIFT IN JAPANESE

359

Here the speakers are pirates addressing a captured priest, whose shipmates they have thrown overboard. Elsewhere in the same narrative, ware appears in its still standard function as speaker-designator: thus the usage in (3) sharply marks (for the narrator and audience of this text) a special class of speakers and a special context. The ‘rough’ second person use of ware may strike the modern Western reader as discordant with the use of honorific oφasu ‘go’ to designate the same hearer in the same utterance, but ware here is a slang/argot tag situating the piratical speakers for a contemporary reader/ hearership, while the honorific signals the social fact that the priest is, after all, a personage worthy of deference from the standpoint of pirate, author, or reader. The point is that by the time of (3), ware can be used to mean ‘you’. The shift in the social function of hearer-designating ware in (2) and (3) presages the regular use of ware as a familiar or derogatory hearer-designator in Muromachi (1378-) and subsequent texts: (4)

itu ware ga ore ni sake wo when you NOM me to sake ACC kureta zo. give-PERF EMPH ‘When did you give me sake?’

(Kyôgen: Morai muko)

In modern Japanese dialects, ware is best known as a very rough and derogatory hearer-designator in Kawachi-type varieties of the Osaka region. This usage descends directly from (3-4). A cluster of Western dialects retain this secondary hearer-designating function of ware we saw develop in (2-4), while a smaller number of other dialects retain the earlier speaker-designating function. Reduplicated wareware is, of course, still a first person plural pronoun in the standard language. The shift of ware from speaker-designator to hearer-designator is by no means unique: (5-8) is a summary list of personal pronouns which have undergone similar shifts in Japanese. (5)

Speaker-Designator > Hearer-Designator a. ware ‘I’( 8th c.) Late Heian period on b. ono2re ‘I’ (8th c.) Heian period on c. konata ‘here’ (Muromachi) Muromachi on. In this case the direction of the shift is not completely clear; konata originates as a proximal locative which might have been used in either function.

360

JOHN WHITMAN (6)

Hearer Designator ore ‘you’ (8th c.)

>

(7)

Reflexive > a. ono2 ‘self’ (8th c.) b. ono2re ‘self’ (8th c.)

(8)

Hearer Designator na ‘you’ (8th c.)

>/<

Speaker Designator Kamakura on Hearer Designator unu/una/ona (non-central dialects) Heian period on Reflexive na (non-central dialects)

See the Appendix for representative examples of each usage. It cannot always be shown that the usage on the left is the source for the usage on the right; thus, for example, a certain amount of mystery necessarily attends the relationship between ‘self’ in central dialects of the 8th century and use of this pronoun as a hearer-indicator in Hachijôjima and the Ryûkyûs (7). Similarly, na appears to show some variability in person designation in the oldest texts as well as in the modern non-central dialects where it is attested.3 In all of (58), however, the items on the left and those on the right are historically related, and where it is possible to tell, the usage on the left is older.

3.

Personalization

The developments of (5-8) are instances of INTRAPERSONAL PRONOUN SHIFT (leaving aside for the moment the reflexives in (7)). They are shifts within the personal categories in the sense of Benveniste (1946, 1956). In Benveniste’s famous formulation, 1st (speaker-designating) and 2nd (hearer-designating) person are the true personal categories: 3rd designates the discourse-invariant ‘non-person’ (1966: 228). I am unaware of shifts of the intrapersonal type in Indo-European languages. What we commonly encounter in the histories of many languages is PERSONALIZATION: shift of a non-person indicator (a 3rd person pronoun as conventionally labeled, or an epithet, title, or common noun) into a speaker- or hearer-designating role. Spanish usted and German Sie are instances of personalization. Personalization is more common yet in Japanese, attested from the earliest texts of the 8th century on. The absence of non-deictic third person pronouns precludes an exact equivalent of a Sie-type shift, but demonstratives shift into speaker- and hearer-designating roles throughout the history of Japanese. Proximal konata ‘here, close to here’ (5c) exemplifies this type, as do mesial sonata ‘there, close to hearer’ > ‘you’, and

PERSONAL PRONOUN SHIFT IN JAPANESE

361

anata ‘yonder, in the vicinity distant from speaker and hearer’ = third person pronoun > 2nd person ‘you’. While the first two of these demonstratives have an original deictic orientation toward speaker and and hearer, respectively, and retain this orientation in the modern language, anata originally represented orientation toward the non-person or non-particpant in the discourse (away from speaker and hearer); its shift to hearer-designator is thus a true case of pronominal personalization, comparable to Sie. In the history of Japanese, epithets and titles follow the path of usted to become hearer-designators: examples present in the language of the 8th century include ki1mi2 ‘lord’, which is a familiar hearer-designator in the modern standard language. More striking from a European standpoint are parallel shifts of titles and epithets into a speaker-designating role. Thus maro2 (noble title) beomes a speaker-designator for male members of the imperial family in the Heian period; boku (modern Tôkyô male speakerdesignator) originates as a deferential (originally epistolary) term, the loan pronunciation of a Chinese character meaning ‘slave’. Its function as a speaker-designator is quite recent. The majority of items used as speaker- or hearer-designators in the contemporary standard (Tôkyô) language result from personalization, as has been widely observed. This is the source of the very common view that modern standard Japanese lacks personal pronouns, in the strict sense. However person categories are as central in Japanese as in any language; this point has recently been reinforced by Nitta (1991). Benveniste (1946/1966: 226-7) took pains to argue (against a certain interpretation of Ramstedt 1950) that the absence of person agreement in Korean is no obstacle to the expression of the categories of person in the verb or by independent pronouns; Benveniste’s arguments go through for Japanese as well. The non-personal source of many personal pronouns in Japanese is not the most distinctive feature of the pronominal system; personalization is a widespread phenomenon across languages. What seems to be special about Japanese (together with the languages discussed in the next section) is the high frequency of personalization, and also the occurrence of intrapersonal pronoun shift. Let us first consider what typological features might correlate with personalization.

362 4.

JOHN WHITMAN Agreement and Personalization

The grammatical expression of agreement appears to be a major constraining factor on the scope of personalization. The best known instances of personalization in European languages show some type of third person agreement: this is true of usted in standard Castilian and American varieties, Sie in German (although merger of the plural agreement patterns makes Sie interpretable as colligating with second person plural), and the polite second person pronoun pan in Polish. Instances where an original non-personal (third person) form trigger personal (first or second person) agreement might be called COMPLETE PERSONALIZATION: colligation of ustedes with second person agreement in West Andalusian and Canary Island varieties is such an example, as is the second person agreement triggered by the Rumanian polite (dumneata) and deferential (dumneanoastra) pronouns.4 Complete personalization requires two changes in a language with morphological agreement: a change in indexical function of the new person-indicator (from discourse non-participant to discourse participant), and a change in agreement pattern. The former change involves merely grammaticalization of an option that is presumably found in all languages: use of a non-personal form to designate a discourse participant (your honor = usted; your father = speaker; son = hearer). The second change signals obliteration of the non-personal origins of the the new personal forms; complete personalization in this sense appears to be relatively rare in languages with morphological agreement. It is thus surely not an accident that Southeast Asia is the best known locus for extensive complete personalization (see for example Cooke 1968). Using Nichols’ (1992) survey of the areal distribution of head marking as a rough guide to the distribution of agreement, we find that South and Southeast Asia has the highest proportion of dependent to head marking among the areas sampled by Nichols (1992: 218). Even more to the point, South and Southeast Asia has the highest proportion of any area in Nichols’ sample of languages with no head or ‘detached’ marking in the clause (i.e. no agreement marked on a verb or auxiliary): 4/10 languages, including Mandarin and Thai. Korean and Japanese are the only languages in Nichols’ Northeast Asian area sample with no head or detached clausal marking. In areal terms, these languages are transitional between Southeast and Northeast Asia; absence of verbal agreement is one of the typological features that set them off from the rest of Northeast Asia. All show extensive personalization.

PERSONAL PRONOUN SHIFT IN JAPANESE

363

A hallmark of personalization in Southeast and East Asia is obligatory use of epithets/titles (including kin terms) as hearer-designators for social superiors.5 Where there are one or more hearer-designators for equals or inferiors (sometimes identified in grammars as the unmarked second person pronouns, although more often than not they themselves result from personalization), this phenomenon might appear characterizable as second person pronoun avoidance. Thus in Korean, Vietnamese, and Japanese, when addressing a kin elder, the kin term is obligatory, the ‘second person pronoun’ inappropriate: (9)

a. Korean emma / #ne eti-lo ka-si-ni? mother you where-to go-HON-Q ‘Where are you going?’ (to mother) b. Vietnamese má / #mày di dau day? mother you go where ‘Where are you going?’ (to mother) c. Japanese okaasan / #anata doko(-e) ik-u no? mother you where-to go-hon Q ‘Mother, where are you going?’

The same restriction holds for addressing other status superiors. It is possible to show that the kin terms used as hearer-designators in the languages listed in (9) are instances of complete personalization, even in the absence of morphologically expressed agreement. This is different from the situation in English, for example, where ‘mother’ may be used as a (vocative) term of address in combination with hearer-designating pronoun (Mother, where are you going), but not as an independent hearer-designator (#Where is Mother going? addressed to one’s mother). Three pieces of evidence make this point. First, kin terms in English, for example, used as non-grammaticalized speaker- or hearer-designators, antecede only third person pronouns (as they trigger only third person agreement); see (10). In contrast, third person pronouns are impossible with personalized antecedents in Korean, including kin terms used as speaker-designators as in (11). (I leave it to the reader to duplicate this result and the following two in Japanese):

364

JOHN WHITMAN (10) Daddy thinks that hisi/*myi idea is best. (said by father) (11) Appa nun *kui-uy / nay sayngkak i kacang NOM most father TOP he-GEN my idea cohta-ko sayngkakhanta. good-COMP thinks ‘Fatheri thinks that *hisi/myi idea is best.’ (said by father)

While Daddy in (10) can antecede his and cannot antecede my, appa ‘Dad’ used as speaker-designator in (11) cannot antecede ku ‘he’ but can antecede nay ‘my’. Second, personalized pronominal subjects in Korean, as well as kin terms used as speaker- or hearer-designators, must take verb endings appropriate for the discourse participant they designate, not for a non-person (third person) subject: (12) Emma to ka-ko siphe /*siphe hay. mother too go-COMP wants want does ‘Mother wants to go too.’ (said by mother) Desiderative siph- in Korean is generally restricted to first person subjects in declaratives and second person in questions; a similar restriction applies to the Japanese desiderative suffix-ta-i. With a third person subject, the Korean desiderative requires a ‘light verb’ form consisting of the infinitive siph-e + ha- ‘do’. This form is impossible with a first person subject, and we see that it is impossible in (12) with personalized ‘mother’ designating the speaker. Siph-e + ha- would be acceptable in (12) if the speaker were someone else, and ‘mother’ a third (non-) person designator. Finally, kin terms used as third person designators co-occur with honorific verb forms. This is somewhat easier to show in Korean (see 9a), where children use honorifics in addressing parents and grandparents, than in contemporary Japanese. Crucially, however, when kin terms are used as speakerdesignators, honorifics are impossible, as they are with speaker-designating subjects in general: (13) *Emma to ka-sey-o mother too go-HON-POL ‘Mother is going too.’ (said by mother)

PERSONAL PRONOUN SHIFT IN JAPANESE

365

This similarity in the behavior of personalized pronouns and kin terms used as discourse participant indicators masks some differences in their behavior. Failure to antecede third person pronouns and co-occurrence with endings limited to discourse participants suffice to show, however, that these are not examples of incomplete personalization. I have suggested that absence of grammatical agreement — obligatorily expressed, overt morphological agreement — correlates with relative ease of complete personalization, that is, shift of a non-personal item into a personal role. This is a contributing factor to the extraordinary variety of personal forms in Southeast Asian languages as well as Japanese. Lack of agreement may be a facilitating, or limiting factor for intrapersonal shift as well. This would account for the absence of clear cases of intrapersonal shift in Indo-European, or other languages with grammatical agreement. However even within the languages of Southeast and East Asia, intrapersonal shift remains a rarer phenomenon than personalization, although as we see below, it occurs outside of Japanese. The question is, what other factors might prompt a language to switch the functions of first and second person pronouns? Two types of explanation come to mind. The first is essentially sociocultural: it identifies some common factor(s) in the sociocultural context to account for the possibilty of this type of flexibility in function. It is here that Takao Suzuki’s notion of ‘empathetic identification’ suggests itself as the basis for such a sociocultural explanation. I explore this possibility in the following section. The second type of explanation is structural/typological: it identifies the common linguistic properties of languages exhibiting intrapersonal shift (we have already seen one, absence of agreement). As we explore this type of approach below, we see that it too requires application of a certain notion of empathy in language, but one substantially more independent from sociocultural variation than Suzuki’s concept.

5.

Empathetic Identification

I suggested above that one very plausible account for the phenomenon of intrapersonal pronoun shift might come from Suzuki’s (1967, 1973/1995) notion of empathetic identification. In the course of a discussion of the

366

JOHN WHITMAN

‘pronominal’ use of kin terms in Japanese, Suzuki notes a striking fact: first person ‘pronouns’ in modern Tôkyô, for example, may be used as second person (hearer-designating) referents. The context described by Suzuki (1995: 172) involves a young mother speaking to her single or youngest son. The mother may use the male speaker-designator boku to address her young son: (14) Boku hayaku irassyai I (boy) quickly come ‘Hurry up’ In fact, this usage is available to young women (at least) addressing small children quite unrelated to them: coming upon a small boy (15) or girl (16) crying, with no parent in sight, a young woman might say: (15) Boku doo si-ta no? I (boy) how do-PERF Q ‘What’s wrong (little boy)?’ (16) Atasi doo si-ta no? I (girl) how do-PERF Q ‘What’s wrong (little girl)?’ The adult speaker in (14-16) designates the hearer (the child) using the first person (speaker-) designator that the child would normally use to refer to her or himself. The background of Suzuki’s discussion is his notion of ‘empathetic identification,’ a stance whereby a speaker identifies with — adopts the viewpoint of — another potential discourse participant. Suzuki points out that speakers select the kin/address term appropriate for the youngest child in the kin group, so that wife designates or addresses husband as ‘father,’ grandparent designates or addresses a granddaughter with a younger sibling as ‘big sister’ etc. According to Suzuki, (14) represents the limiting case of empathetic identification, when a particular addressee (the child) is her/himself the target of empathetic identification: that is, when the child is the youngest member of the kin group. In this case, the adult speaker applies empathetic identification to take the viewpoint of the child hearer, and thus uses the speaker-designator from the child’s viewpoint (for a boy, boku ‘I’) to address the hearer. Let us consider how empathetic identification might be used as an explanation for intrapersonal pronoun shift of the sort we saw in (1-4). The scenario would proceed as follows: a specific first-person form (speaker-

PERSONAL PRONOUN SHIFT IN JAPANESE

367

designator) comes to be used, in some appropriate social situation, to designate the hearer in empathetic identification contexts, just like boku and atasi in (14-16); such usage becomes widespread; eventually the form is grammaticalized as a hearer-designator. I should emphasize that this explanation for intrapersonal pronoun shift has not actually been proposed by Suzuki or any other researcher, as far as I am aware. I raise it here simply as a very plausible consequence of the notion of empathetic identification. I would now like to argue that this proposed explanation is incorrect. My argument is based on the fact that the historical shifts we saw in (1-4) and the apparent example of synchronic intrapersonal shift in (14-16) are very different phenomena, despite their superficial similarity. In the remainder of this section, let us take a closer look at what is really going on in (14-16). First, note that boku and atasi are the only speaker-designators that may be used in the empathetic identification role of (14-16). For example, the ‘rough’ male speaker-designator ore is the unmarked form in in-group discourse even among young boys, once they have contact with male peers. However no mother in any circumstances could ever say to her young son: (17) #Ore hayaku irassyai I (boy) quickly come ‘Hurry up’ If the striking use of a speaker-designator to address a hearer in (14-16) were really a matter of taking the hearer’s point of view, (17) might be closer to the actual viewpoint of small boys than (14), but ore is completely uninterpretable in a hearer-designating sense. Thus the actual scope of empathetic identification with hearer-designators is extremely limited, basically to the two items boku and atasi in (14-16). Of course the conception of empathetic identification sketched by Suzuki only allows a speaker to take the viewpoint of the youngest child in the family: speaker-designators not used by a child in this position will not be accessible to empathetic identification. This restriction on the phenomenon itself casts some doubt on its appropriateness as an historical explanation for intrapersonal shift. Presumably, to account for the range of intrapersonal shifts listed in (5-8), a more extended conception of empathetic identification is required, one which permits speakers to adopt the viewpoint of their interlocutors in a wider range of circumstances. There is no evidence for such

368

JOHN WHITMAN

a conception in Suzuki’s work or elsewhere. But to return to our main point, even the concept of empathetic identification as restricted by Suzuki does not immediately explain the contrast between boku and ore in (14) and (17). Suzuki’s own examples point up a difference between boku and ore which I believe is crucial. Suzuki observes that hearer-designating boku may be suffixed with hypocoristic-forming-tyan (the diminutive/affectionate form of the honorific/polite suffix -san): (18) Boku-tyan kore hosi-i n desyoo (1995: 172) I (boy)-AFFEC this want-IMP COMP PROP ‘You want this, don’t you?’ This same possibility holds for all of the kin terms with ‘pronominal’ uses (i.e. as terms of address = hearer-designator, etc.) in Japanese. All are suffixed with -san or affectionate -tyan in this usage: (19) obaasan/tyan oziisan/tyan okaasan/tyan otoosan/tyan

‘grandma’ ‘grandpa’ ‘mom’ ‘dad’

oneesan/tyan oniisan/tyan obasan/tyan ozisan/tyan

‘older sister’ ‘older brother’ ‘aunt’ ‘uncle’

On the other hand, ore (and other speaker-designators) are unsuffixable with (-tyan/san).6 The fact that boku patterns with the kin terms in (19)7 may be related to its historical origin as a personalized epithet; in any event, the phenomenon that Suzuki describes as empathetic identification seems closely related to the distribution of these suffixes. Two further facts support the importance of this relation. First, as Suzuki points out, the kin terms eligible for use as hearerdesignators cannot designate relations specifically lower in status (younger) than the speaker. Thus imooto(-tyan/san) ‘younger sister’, otooto(-tyan/san) do not occur in this function. This is consistent, as Suzuki states, with his view that the reference point is the youngest child in a family, but it also follows directly from the distribution of -tyan/san. As has been widely pointed out, -san (and -tyan, despite its affectionate status) are non-ingroup markers, applying a dynamic definition of ingroup, where the minimal ingroup is speaker. These suffixes in the minimal case mark non-speaker, then superiors in a group (such as elders in a kin group), then non-group members. Since imooto ‘younger sister’ otooto ‘younger brother’ cannot be superiors in a kin group, imooto-san, otooto-san can designate only non-relatives (‘your

PERSONAL PRONOUN SHIFT IN JAPANESE

369

younger sister’, ‘your younger brother’). This fact explains the exclusion of kin terms designating specifically younger relatives from (19) quite independently of the notion of empathetic identification. Second, the use of -san (contracted from -sama ‘appearance, likeness’) as a suffix with names and status terms (including kin terms) dates only from the Muromachi period (14th century). Prior to this time, it is difficult to find examples of kin terms used in exactly the pattern described by Suzuki as empathetic identification. Thus while terms such as ani ‘older brother’ (nonhonorific), ane ‘older sister’ (non-honorific) are used prior to this date to designate both kin and non-kin young men or women, the same is true of oto (> oto+φito ‘younger brother + person’ > otooto) ‘younger brother’, imo (> imo + φito > imooto) ‘younger sister’. This fact suggests that the specific pattern of empathetic identification in Suzuki’s restricted sense is in fact a byproduct of the broader pattern of status term + honorific marker (-tyan, -san, -sama, earlier Japanese -tono). More importantly for our present concerns, it suggests that empathetic identification in this restricted sense cannot have been the source for intrapersonal pronoun shift in earlier Japanese.

6.

Intrapersonal Pronoun Shift and Reflexives

In the preceding section I failed to mention a basic fact which poses perhaps a more serious problem than any other for a scenario relating intrapersonal pronoun shift to empathetic identification. The empathetic identification scenario works only for cases of speaker-designators shifting into a hearerdesignator role. While this is the direction of shift in (5), (6-8) show that speaker-designators are not the only source for intrapersonal shift. Of these latter, (7-8) involve a shift to or from a reflexive function. Both (7) and (8) are difficult to assess because the modern patterns occur in non-standard varieties whose history is unknown. However the flavor of this type of shift can be understood from the historical and current usage of the pronoun onore ‘self’ < ono ‘self’ + -re substantivizing (pronominal) suffix (5b), (7b). In the modern standard language, onore has a ‘literary’ function as a reflexive pronoun (used in Bible translations, proverbs, or stock sayings such as (20a), for example), and a usage as an expresion of anger or outrage (20b), descended from its function as a derogatory hearer-designator (see Appendix 2c).

370

JOHN WHITMAN (20) a. Mazu onore wo kaerimi-yo (Hayashi, Nomoto and first self ACC reflect-IMPERATIVE Minami 1984) ‘First, reflect upon your self(‘s actions).’ b. Onore! Ima ni mite i-ro (Hayashi, Nomoto and son-of-bitch so on seeing be-IMPERATIVE Minami 1984) ‘Son of a bitch! Just you wait.’

The reflexive function of onore (Appendix 2a) is basic and original, but from early along it has hearer- and speaker-designating (Appendix 2b) functions as well, both with a lowering effect. Onore has never undergone complete personalization, in the sense that its reflexive function remains available; but it has allowed throughout its history the option of a shift to a personal function. What might explain the possibility of such a shift? Kuno (1972) provides a framework for understanding the link between reflexives in their long-distance (sometimes ‘logophoric’8) function and the function of person-designator. According to Kuno’s analysis, long-distance reflexives (such as modern Japanese zibun ‘self’) are represented as first or second person pronouns in the ‘direct discourse representation’ of logophoric complements (clausal complements of verbs of saying, hearing, etc.). It is not difficult to imagine how a long-distance reflexive with a speaker- or hearerdesignating function in a direct discourse representation could be reanalyzed as speaker/hearer-designator outside of that context as well. This scenario might involve a type of analogical extension of contexts where Kuno would posit a direct discourse representation. In fact I think a number of factors favor the analogical ‘closeness’ of long-distance reflexives and person designators, even outside of direct discourse representations narrowly construed. First, long-distance reflexives are empathy foci. Kuno and Kaburaki (1977) demonstrate that when zibun is used with a long-distance (non-clausemate) antecedent, the speaker must empathize with the referent (antecedent) of zibun rather than other referents in the same clause. This is a crucial point of similarity with first and second person pronouns, since these rank highest on the Speech Act Empathy Hierarchy (see footnote 1). Kuno & Kaburaki’s example is the following: (21) Tarooi wa Taroo TOP okane o money ACC

[Hanako ga zibuni ni kasite kureta/*yatta] Hanako NOM self to lend gave gave tukatte simatta. use ended-up

PERSONAL PRONOUN SHIFT IN JAPANESE

371

‘Taroo has spent all the money that Hanako had lent to him.’ (1977: 636) In (21) kureta ‘gave’ (to empathy target) is appropriate, but yatta ‘gave’ (to empathy non-target) results in a conflict of empathy foci. Use of zibun marks its referent (Taroo) as the target of empathy, but yatta marks Hanako, not Taroo as the target. Examples like (21) show that long-distance zibun marks the target of empathy even in an embedded clause which does not receive a direct discourse representation, as there is no logophoric predicate in (21). A second related fact emerges about contexts such as (21) in English. It is well known that in most languages reflexives (long-distance or local) may appear in positions where they have no syntactically eligible antecedent (see in particular the study of picture noun reflexives in Kuno 1987). In such cases the reflexive is often said to have a discourse antecedent, as in the picture noun example in (22): (22) Mary opened the album. There on the first page was a picture of herself. The discourse antecedent Mary in (22) is perhaps better described as the narrative ‘I’ of the text, again certainly the target of the narrator’s empathy, in Kuno’s terms. As Kuno (1987) points out, such an antecedent must be sentient, although she need not be an active agent: (23) a. The wind opened the album. Mary looked up from the floor. There on the first page was a picture of herself. b. #The wind opened the album. Mary lay dead on the floor. There on the first page was a picture of herself. There is a (rather crude) literary device that allows us extend the overt counterpart of a direct discourse representation to the empathy-sensitive reflexive examples in (21-23). Consider the following: (24) a. John spent all the money that “Mary lent to me”. b. Mary opened the album. “There on the first page is a picture of me”.9 c. The wind opened the album. Mary looked up from the floor. “There on the first page is a picture of me”. d. #The wind opened the album. Mary lay dead on the floor. “There on the first page is a picture of me”.

372

JOHN WHITMAN

We might call the device in (24) INJECTED QUOTATION; some readers may feel that in this usage he said, she thought, etc. is somehow ellipted. The more important point is that in (24a-c), injected quotations are felicitous, and the designee of me is clearly John (24a) and Mary (b-c). In (d) the designee of me is unclear, and the sentence is infelicitous without tacit supply of another narrative ‘I’ to antecede it. The injected quotation pattern is related to both the acceptability and the interpretation of the reflexives in (20-22). I would speculate that a first or second person pronoun in an injected quotation is possible wherever long-distance Japanese zibun occurs, although full substantiation of this conjecture exceeds the scope of this paper. If this is correct, there may be an argument for extending Kuno’s direct discourse representation beyond the domain originally proposed for it. In any event, both the empathy-sensitivity of long distance (including discourse-dependent) reflexives and their interchageability with personal pronouns in injected quotation paraphrases further supports the close relation between reflexives and personal (speaker- and hearer- designating) pronouns. We have seen above an explanation of how reflexives (self-designators) might shift into a person-designating role, that is, undergo personalization. Such a shift would be based on the personal status of reflexives in direct discourse representations, and their high empathy status. We have seen that the reflexive onore permits such shifting throughout its history. Let us return to the case of ware, the pronoun whose intrapersonal shift we saw in (1-4). From the earliest textual evidence, ware has a so-called ‘reflexive’ (selfdesignating) usage (see Appendix 1b) parallel to its more basic speakerdesignating function. Compare ware in (26) below to onore in (25): (25) kirikake-datu mono ni, ito aφoyaka naru kadura no board fence-like thing on very greenish be vine GEN kokoroti yoge-ni φafi-kakareru ni, siroki fana zo, onore feeling good-ly creep-attach on white flower EMPH itself fitori wemi no mayu φiraketaru one smiling GEN eyebrows opened ‘On a very green vine which was creeping up something like a board fence, white flowers, all by themselves, raised their smiling eyebrows.’(GM: Yûgao)

PERSONAL PRONOUN SHIFT IN JAPANESE

373

(26) Ware φitori sakasiki fito nite, omofosiyaru kata I one strong person being think(HON) way zo naki ya EMPH not.be EXCLAM ‘He (Genji) himself was the only reliable person, and there was just no way to think it out.’ (GM: Yûgao) Both onore φitori in (25) and ware φitori in (26) mean ‘oneself alone’. The primary difference between (25) and (26) is that the latter is arguably direct discourse (quotation). However modern translators (Japanese and English) assign (26) to the narrator’s voice rather than Genji’s (that is, they do not analyze it as direct quotation), partially because the punctuated quotation in modern editions is reserved for speech uttered out loud (26, if direct quotation, is uttered by Genji to himself), and partially because the honorific verb omosiyaru is consistent with Murasaki Shikibu’s third person verbal reference to Genji.10 In fact, (26) is an example of what Kuno (1988) calls ‘blended quasi-direct discourse’: the utterance has elements of direct discourse (the emotive sentence-final particle ya) and indirect discourse (the honorific verb). Blending of this type, which is probably the norm across languages, is a further factor contributing to the blurring of the distinction between self- and person-designators. In other contexts ware is unambiguously self-designating: (27) Kimi mo e taφe-tamaφa-de, ware φitori sakasi-gari Lord too can bear-HON-not I one strong-ishly idaki-mo-tamaφeru ni, kono φito ni iki wo embrace-hold-HON although this person to breath acc nobe-tamaφite zo, kanasiki koto obosarekeru, tobakari, let.out-HON EMPH sad thing think(HON) awhile ito itaku e mo todome-zu naki-tamaφu. very exceedingly can even stop-not cry-HON ‘The lord (Genji) too could not bear it; although he alone had showed strengh and embraced (her), to this person he let out his breath, the sad things came to mind, and for a while he cried very hard, without being remotely able to stop.’ (GM: Yûgao) In (27) ware, in the exact collocation of (25-6), patterns as a long-distance reflexive bound by kimi (‘lord’ = Genji). The status difference seems most crucial in distinguishing self-designating onore and ware by the time of Genji

374

JOHN WHITMAN

monogatari: onore in (25) refers to flowers, and in Appendix (2) all examples refer to inferiors of the speaker or protagonist. Onore is thus presumably unavailable to refer to Genji in (26) and (27). Otherwise, it is difficult to distinguish the two. The hypothesis that I would like to propose in this section is that intrapersonal pronoun shift is always mediated by a reflexive function. That is, reflexives (unspecificed for person) shift into a speech act participant-designating function, and speech act-participant designators (usually speakerdesignators, it appears, but see (8)) may shift into a reflexive function. The derogatory second person use of onore represents the former shift, while the reflexive function of ware represents the latter. Direct intrapersonal shift — shift of a speaker-designator immediately into the function of a hearer designator — is rare or non-occurring, I suggest.

7.

Reflexive Personalization in Other East Asian Languages

In the preceding section I proposed that intrapersonal shift is always mediated by a reflexive function: that is, reflexives shift to a speech act participantdesignating function, and personal pronouns may shift to a reflexive function. The first type of shift is well attested in East Asian languages outside of Japanese. For example, the modern Korean humble speaker designator ce ‘I’ is derived from reflexive ce (Lee 1979/1991: 46).11 In Middle Korean ce functions only as a reflexive; the modern speaker-designating function is a later development. Middle Korean ce itself is most likely a loan from Chinese (Late Middle Sino-Korean ccó , Mandarin zì ‘self’) (Martin 1992: 439). The shift of a reflexive to a humble (‘lowering’) speaker-designating function is precisely parallel to the speaker-designating use of onore (Appendix 2b). Modern Chinese provides a similar example of reflexive > speakerdesignating personalization with the form zá ‘self’, ‘I/we ourselves’, ‘we’ (authorial), attested from the Song dynasty onward.12 Lü Shu-hsiang (1985) shows that zá is derived through the contraction of zì (Middle Chinese dzi, the source of the Korean pronoun discussed above) ‘self’ and jia (MC ka) ‘house, family’. In Mandarin, zá survives in the first person plural inclusive (speaker+hearer-designating) pronoun zámen (zá + men plural suffix). In Chinese, it is unclear whether speaker-designating za marked a

PERSONAL PRONOUN SHIFT IN JAPANESE

375

status difference with the standard speaker-designator wo ‘I’; in most dialects singular zá seems to have been lost at the expense of wo, as in Mandarin (Lü 1985: 99). We have seen that Korean ce is humble; Japanese onore is ‘lowering’ in both its speaker- and hearer- designating roles; Japanese hearerdesignating ware is lowering, as are the reflexes of ono ‘self’ as hearerdesignator in dialects (Appendix D). The same pattern can be seen in varieties of Japanese that use reflexive zibun ‘self’ in a person-designating role, such as the prewar military usage of zibun as a speaker-designator for private soliders addressing officers,13 or Kansai varieties where zibun occurs as a hearer designator (typically for men addressing women14). In general, then, person-designators derived from a reflexive source involve some type of ‘lowering’, resulting in a humble (or in some cases perhaps ‘rough’) signification with speaker-designators, and a familiar or derogatory signification with hearer-designators. I would argue that this regularity is a direct reflection of the status of (long-distance) reflexives as empathy targets. Humble forms reference entities highest on the speech act participant hierarchy (speaker and speaker’s group), familiar forms the next highest (in-group hearer and hearer’s group). Honorifics, and the distal (formal) verbal forms found in Japanese and Korean, on the other hand, never reference speaker, frequently reference entities lowest on the speech act participant hierarchy (third person or non-participants), and reference only non-ingroup hearers. There is thus an inverse relationship between empathy status and honorific or deferential status.15 We might tag this the ‘familiarity breeds contempt’ relationship, although its actual dynamic is much subtler: the gambit of co-opting another’s point of view works against the naturally distancing force of honorification. Seen from this standpoint, the inclusive function of Mandarin first person plural zámen is consistent with its reflexive source. Inclusive pronouns reference speaker and hearer, the top two categories in the speech act participant hierarchy. The consistent feature of reflexive-derived personal pronouns is that they retain their high empathy status.

8.

Summary: Point of View and Pronoun Shift

In this paper I distinguished two types of lexical shift involving personal pronouns: personalization, the shift of a non-personal item into a speech act

376

JOHN WHITMAN

participant-designating role, and intrapersonal shift, a ‘switch’ in the persondesignating function of a pronoun (from speaker to hearer or the opposite). Intrapersonal shift is a remarkable type of lexical change from the European/ Indoeuropean standpoint, and indeed the examples I have discussed all come from East Asian languages. It is thus extremely tempting to regard intrapersonal shift as not merely an areal linguistic phenomenon, but an historical phenomenon closely related to sociocultural area. In exploring the applicability of Takao Suzuki’s notion of empathetic identification to the explanation of intrapersonal shift, my intention was to consider the utility of a socioculturally-based analysis, for Suzuki’s fundamental argument is that empathetic identification is a disposition, somehow uniquely available in a Japanese cultural context (I expect that Suzuki would be willing to broaden this context to include East Asian beyond Japan). This type of explanation is always potentially available for a certain line of functionalist analysis, and it was because the explanation was so obviously tempting in this case that I explored it. Functionalist explanations attempt to relate linguistic phenomena to the nonlinguistic needs and desires of speakers and hearers. Where linguistic variation is observed, there is always the possibility, for at least the line of functionalist explanation under discussion, of relating variation to cultural differences in needs and desires. Empathy in language (in the broadest sense) is a notion which might seem particularly susceptible to cultural variation; this is of course is exactly Suzuki’s claim. A familiar stereotype lurks just around the corner: languages belonging to the groups under discussion (East and Southeast Asia) are marked by a tendency to be ‘empathetic’ because the cultures of their speakers place a premium on empathetic interaction. The concept of empathy developed in Kuno’s work is not unrelated to Suzuki’s: it also involve the notion of point of view, and the possibility of adopting the viewpoint of an actor other than the speaker. However a salient feature of Kuno’s concept (one that can be said to apply to Kuno’s functionalism in general) is that it is universalistic: Kuno holds that empathy is a salient feature of human language in general. This is not to say that there might not be some relationship between the distribution of sociocultural traits and empathy-sensitive linguistic phenomena (such as long-distance reflexives, verbs of giving and receiving, honorific marking); but the relationship should be the generally indirect relationship that holds between linguistic and ‘cultural’ domains, the latter even more notoriously difficult to define than the former. The general predic-

PERSONAL PRONOUN SHIFT IN JAPANESE

377

tion made by the approach represented by Kuno proceeds from presence or absence of the linguistic trait: if an empathy-sensitive trait is present, empathy-sensitive phenomena will be observed, regardless of the cultural context. I have argued that the two types of pronominal shift, personalization and intrapersonal shift, are sensitive to the presence of two respective typological features: absence of agreement, and presence of long-distance reflexives. The striking phenomenon of intrapersonal shift is always mediated by a reflexive function, shift of a personal pronoun out of or into a reflexive role.

Notes 1.

This notion is expressed explicitly in the Speech-Act Empathy Hierarchy of Kuno and Kaburaki (1977: 652): Speech-Act Empathy Hierarchy (revised) It is easiest for the speaker to empathize with himself; it is next easiest for him to empathize with the hearer; it is most difficult for him to express more empathy with third persons than with himself or with the hearer: Speaker > Hearer > Third Person

2.

I use this term pretheoretically here, well aware of the view that Japanese lacks personal pronouns in the normal sense, on which more below.

3.

Hirayama et al (1992) identify na as a familiar hearer-designator in a number of Tôhoku dialects.

4.

I am indebted to Wayles Browne for directing me to the Rumanian examples.

5.

Suzuki (1973/1995) represents the obligatory use of kin/status terms and the disallowance of pronouns in addressing superiors as a distinctive characteristic of Japanese; in fact, as (9) shows, it is a feature of languages with no grammatical agreement and extensive personalization. The question of a possible ‘cultural basis’ for this cluster of typological properties is of course open.

6.

Ore-sama ‘Mr. Me’ is a jocular possibility, but it retains the speaker-designating function of ore.

7.

Only -tyan, not -san is possible with boku, but this is because in (14-16), (18) boku(-tyan) designates the lowest member of the kin hierarchy. Hypocoristics may be used to designate lower status members; honorifics may not.

8.

The issue of whether the distinction between reflexive and logophor is significant in a language such as Japanese is beyond the scope of this paper. Much of the generative literature on this topic has argued for such a distinction, using the term ‘long distance reflexive’ for anaphors with a non-clausemate c-commanding antecedent, and reserving the term logophor for contexts where a reflexive has an antecedent which has speaker or hearer status in the discourse, but which does not necessarily c-command. These two

378

JOHN WHITMAN notions overlap in unmarked cases. From the standpoint of this paper, I have chosen to designate the items in question as (long-distance) reflexives. This is because logophors in the West African languages where they first attracted intensive study are grammatically third person entities, while the reflexive items involved in intrapersonal shift are crucially not specified for person.

9.

Note that present tense seems to be obligatory in the injected quotation here.

10.

Seidensticker (1978: 72) translates (26): ‘He (Genji = ware) was the only rational one present, and he could think of nothing to do.’

11.

Lee argues (1991: 46) against Choe Hyon-bae’s (1937) earlier suggestion that ce is derived from the distal demonstrative ce < Middle Korean tye. Lee’s view that reflexive personalization is a more semantico-pragmatically plausible shift prefigures the argument in this paper (Lee also adduces morphological evidence). Indeed, I know of no instance of a distal demonstrative shifting into a speaker-designating role.

12.

I am indebted to Tsu-lin Mei for bringing this example to my attention and explaining it in detail to me, as well as directing me to Lü Shu-shiang’s discussion.

13.

American military cadet and Marine use of third person for self-reference in this context follows the same pattern.

14.

Akio Kamio (p.c.) points out to me that women may also use zibun as a hearer designator in the Kansai dialects in question, but that this is considered “inelegant”. Use of zibun as self-designator is of course widespread.

15.

This relationship makes examples such as (26), where self-designating ware is used in combination with an honorific in what I suggested was a quasi-direct discourse pattern, particularly interesting.

16.

A kunchû (reading gloss) in the Nihon shoki spells the deionym in man’yôgana as ohoana-muti. It is a matter of dispute as to whether this is more than just a later literal reading of the characters in question. Ana is not attested elsewhere as a pronoun in eighth century sources.

17.

Izuyama (1994: 14) notes the existence of una as a second person form in Hachijôjima and reconstructs a reflexive form *una, which she also relates to Ryûkyûan reflexive and eighth century second person/reflexive na-. The pronouns are certainly cognate, but it would be hasty to reconstruct *una as their protoform. This would not be regularly relatable to supposed ana, and the prothetic vowel is at least as likely the result of analogy.

18.

Ore is already lexicalized in a (low) second person function in the eighth century. This form clearly involves the suffix -re, but there is no independent morpheme o- in a relevant interpretation. This indicates loss of medial /n/ in earlier *ono2-re, perhaps through assimilation to /r/ and medial *r loss as suggested by Whitman (1991) for forms like fari ‘needle’ : LMK pànól id. Alternatively, *ono2re > ore simply involves the reduction of medial syllables with nasal onset attested in rendaku. Once this process occurred the internal structure of the pronoun becomes opaque, so that it coexists with eighth century ono2-re in the transparent meaning of ‘self’. Subsequently onore itself shifts into a second person function, as ore shifts into first person.

PERSONAL PRONOUN SHIFT IN JAPANESE 19.

379

Some analyses identify the final syllable in ane ‘older sister’ and irone ‘older sibling of the same mother (iro -’same mother’ + ne) with suffixal -ne. The argument for this identification is completely unclear: since iro- is a bound morpheme, ne in this case must be substantive, and is clearly identifiable with a morpheme designating older siblings (cf. ani ‘older brother’). Suffixal -ne appears in examples like the following: Wo.kakitu-no wo-wo φik-i φos-i [imo na-ne-ga] little.yard-GEN flax-ACC pull-ing dry-ing wife you-AFF-GEN mak-ing tukur-i kis-e-ke-m-u siro.taφe no fimo-wo-mo tok-a-zu. (M1800) make-RY dress-RY-PAST.PRESUMP-SS white.cloth GEN cord-ACC-also untie-MZ-not “His dear wife must have made it to dress him with, plucking the hemp in the garden and drying it, he doesn’t untie the cord of that white garment”

References Benveniste, Emile 1946 “Structure des relations de personne dans le verbe”. Bulletin de la Société Linguistique 43: 1-12. Reprinted in (1966), 225-236. English translation in (1971), 195-204. Benveniste, Emile 1956 “La nature des pronoms.” In M. Halle et al. (eds), 34-37. Reprinted in (1966), 251-257. English translation in (1971), 195-204. Benveniste, Emile 1966 Problèmes de linguistique générale. Paris: Gallimard. Benveniste, Emile 1971 Problems of general linguistics. Miami Linguistics Series No. 8. Coral Gables: University of Miami press (translation of 1966). Choe, Hyon-bae 1939 Wuli mal pon. Seoul: Cengum-sa. Cooke, Joseph 1968 Pronominal reference in Thai, Burmese, and Vietnamese. Berkeley and Los Angeles: University of California Publications in Linguistics, No. 52. Halle, M. et al. (eds) 1956 For Roman Jakobson. The Hague: Mouton. Hayashi Shirô, Kikuo Nomoto and Fujio Minami 1984 Reikai shin kokugo jiten. Tokyo: Sanseidô. Hirayama, Teruo (ed) 1992 Gendai Nihon hôgen daijiten. Tokyo: Meiji shoin. Izuyama, Atsuko 1992 “Ryûkyû hôgen no ichi ninshô daimeishi (First person pronouns in Ryûkyûan dialects)”. Kokugogaku 171. Izuyama, Atsuko 1994 “Ryûkyû · Hachijôjima hôgen ni okeru nishu no ichi ninshô daimeishi’ (Two types of first person pronoun in Japanese Dialects - Hachijôjima and Okinawa”. Dokkyo University Bulletin of Liberal Arts 28.2: 1-18.

380

JOHN WHITMAN

Kuno, Susumu 1972 “Pronominalization, reflexivization, and direct discourse”. Linguistic Inquiry 3: 161-195. Kuno, Susumu 1987 Functional syntax. Chicago: The University of Chicago Press. Kuno, Susumu 1988 “Blended quasi-direct d iscourse in Japanese”. In W. J. Poser (ed), 75-102. Kuno, Susumu and Etsuko Kaburaki 1977 “Empathy and syntax”. Linguistic Inquiry 8: 627-672. Lee, Ki-moon 1979 “Kwuke uy inching taymyengsa (The personal pronouns of Korean)”. Kwanak emwun 3. Collected in Lee (1991) Lee, Ki-moon 1991 Kwuk.e ei-sa yenkwu (Studies on the history of Korean vocabulary). Seoul: Tong-A Chwulphan-sa. Lü, Shu-hsiang 1985 Chin tai Han yû chih tai tz’u. Shanghai: Hsüeh lin ch’u pan she. Martin, Samuel E. 1992 A reference grammar of Korean. Rutland, VT: Charles E. Tuttle. Miyara, Tôsô 1930 Yaeyama goi (Yaeyama vocabulary). Tokyo: Tôyô bunko. Murayama, Shichirô 1950 “Kodai Nihongo ni okeru daimeishi (Pronouns in Old Japanese)”. Kokugogaku 13: 40-47. Nichols, Johanna 1992 Linguistic diversity in space and time. Chicago: The University of Chicago Press. Nitta, Yoshio 1991 Nihongo no modariti to ninshô (Person and modality in Japanese). Tokyo: Hitsuji Shobô. Omodaka Hisakata et al. (eds) 1967 Jidaibetsu kokugo daijiten: jôdai hen. Tokyo: Sanseidô. Ôno, Tôru 1978 Nihongo no sogenteki kenkyû. Tokyo: Takayama honten. Poser, William J. (ed) 1987 Japanese syntax. Stanford: CSLI. Ramstedt, G.J. 1950 A Korean grammar. Helsinki: Mémoires do la Société Finno-Ougrienne. Seidensticker, Edward 1978 The tale of Genji (English translation). Rutland, VT: Charles E. Tuttle. Suzuki, Takao 1967 “On the notion of teknonymy”. Studies in Descriptive and Applied Linguistics, Bulletin of the Summer Institute of Linguistics, vol. IV. Tokyo: Interntional Christian University.

PERSONAL PRONOUN SHIFT IN JAPANESE

381

Suzuki, Takao 1973/95 Kotoba to bunka, Tokyo: Iwanami shoten. Suzuki, Takao 1984 Words in context. Tokyo: Kodansha International. (English translation of Suzuki 1973/95.)

Appendix (A) Speaker Designator (1st Person) > Hearer Designator (2nd Person) (1) ware 8th century uses of this pronoun are as a 1st person indicator (a) or a reflexive (selfindicator) (b). Use as a 2nd person indicator is attested from the end of the Heian period (c), where it designates an equal or inferior of the speaker. From the Kamakura period on ware is a derogatory or familiar second person designator (d). a. ware φito wo okos-a-mu I person ACC waken-MZ-PRESUMP ‘I will waken somebody.’

(GM, Yûgao)

b. womuna wo, sasite sono φito to tazun-e id-e- (GM, Yûgao) put out-RY woman ACC especially that person COMP ask-RY tamaφ-a-n-e-ba ware mo na.nori wo si-tamaφ-a-de HON-MZ-not-IZ-while self too name.stating ACC do-HON-MZ-not-ing ‘While not particularly asking as to her identity, nor getting into giving names re himself...’ c. ware ni mo tug-e maus-i ni φito wo (KM, 27.32) you DAT also tell-RY say-RY DAT person ACC tukaφas-i-tar-i-si-ka-ba send-RY-PERF-RY-PAST-IZ-COND ‘Since I had sent someone to inform you as well...’ (to an equal’s wife) d. itu ware ga ore ni sake o kure-ta zo (Kyôgen: Morai muko) when you NOM me to wine ACC give-PERFEMPH ‘When have you given me wine?.’ (2) onore Onore seems to originate as a self-designator, even an emphatic reflexive (a), and actually retains this function in the modern language, in such expressions as onore o sire ‘Know thyself’. As a first person indicator it was used in humble or lowering contexts (b). It appears in addressee-designating (second person) contexts from early on, always directed toward inferiors or in a derogatory sense (c).

382

JOHN WHITMAN

a. nak-i-tamaφ-u sama, ito wokasige ni rauta-ku, (GM, Yûgao) very affecting DAT cute-GER weep-RY-HON-RT sight mi-tatematur-u φito mo ito kanasi-ku-te see-DEFER-RT person too very sad-GER-ing onore mo yoyo to nak-i-n-u self too boohoo COM cry-RY-PERF-SS ‘The sight of (Genji) crying was affecting and charming, and the people seeing it were so saddened that they themselves wept.’ b. onore kakar-u winaka.udo nar-i to-te (GM, Suma) I such country.person be-SS COMP-GER ACC especially that person COMP ask-RY put out-RY obos-i sut-e-zi think-RY discard-MZ-NEG.PRESUMP ‘When he sees what a country person I am, surely he won’t be able to put me out of his mind’ (woman surmising about Genji) c. Kaguya Hime φa, tumi wo tukur-i tamaφ-er-i-kere-ba (TM) Kaguya Hime TOP sin ACC make-RY HON-PERF-RY-PAST-IZ-COND kaku iyasi-ki onore ga moto ni sibasi oφas-i-t-uru nar-i GEN abode in briefly be-RY-PERF-RT be-SS thus low-RT you ‘As for Kaguya Hime, because she committed a transgression, she has been briefly residing in this lowly abode of yours’ (3) konata This word originates as a proximal (close-to-speaker) demonstrative of location ‘here, close to here’ < ko ‘this, here’ (proximal) + no GENITIVE + kata ‘side, direction’. It is one of a number of demonstratives of location that serve in a ‘personal’ pronominal function: sonata (medial (close to hearer) ‘there, close to there’ has been a speakerdesignator throughout the history of Japanese; anata (distal) ‘yonder’ originates as a third person designator but shifts to a second person desigator in modern Japanese. The ‘personal’ use of konata first emerges in the Muromachi period with speaker- (a) and hearer- (b) designation attested almost simultaneously. It is thus difficult to say with certainty that konata attests a 1st person > second person shift; I have included it in this category because the basic deictic function of the word is proximal (close-to-speaker). As a hearer-designator konata is honorific; as a speaker-designator it is formal. These statusmarking features put konata outside the paradigm for intrapersonal shift explicated in this paper, as does the likely fact that konata never passed through a specifically ‘reflexive’ function. It seems preferable to regard speaker- and hearer-designating konata as simply two different options for personalization of a demonstrative. sono kotoba mo, konata φa mimi ni tomar-u mono wo (SG) those word too I TOP ear in stay-RT thing-ACC (EXCL) ‘You know, those words also stay in my ears.’ b. waraφa mo konata no on-sosyô no koto wo anji-te wi-mas-i-ta ga (KN) I too you GEN HON suit GEN matter ACC worry-ing be-POL-RY-PERF but ‘I too was worrying about the matter of your lawsuit.’ a. nau , EXCL

PERSONAL PRONOUN SHIFT IN JAPANESE (B) Hearer Designator (2nd person)

>

383

Speaker Designator (1st person)

ore It is difficult to track the relationship between speaker-designating ore and the homophonous hearer-designating form found in the oldest texts (a). The speakerdesignating usage (b) appears first in the 12th century, and occurs in Kamakura texts in conversations between young males. Hearer-designating ore, on the other hand, is rare in Heian literature. (It was plausibly avoided by refined women authors, but appears fairly frequently in prose literature during the Kamakura period). Its demise in the Muromachi period is probably related to spread of the speaker-designating usage. In central dialects hearer-designating ore undergoes a sound change to ure (c) from the 12th century on. One view derives speaker-designating ore from contraction of onore (see above). It seems likely that onore retained its ‘reflexive’ (i.e. either speaker- or hearer- designating potential) because of its transparent relation to reflexive ono- ‘self’; contracted ore makes this relation opaque, and ure further so. If this is correct, 8th century ore may simply reflect an original occurence of the onore > ore contraction, lexicalized in a hearer-designating function; the 12th century contraction was able to lexicalize in a speaker-designating function because of the shift of ore > ure in the hearerdesignating role. a. φototogisu, ore kayatu yo (MS, Kamo e mairu michi) cuckoo you (derogatory) EMPH ore nak-i-te koso ware wa ta u(w)-ure you sing-RY-GER FOCUS I top field plant-IZ ‘Hototogisu, you jerk/It’s when you sing that I must plant the fields.’ b. oya no umi-i tuk-e-te oka-syat-ta (OD 4.1) parent GEN bear-RY attach-RY-GER put-HON-PERF hana nar-e-ba ore ga mama ni wa nar-a-nu nose be-IZ-COND I NOM as is DAT TOP become-MZ-not ‘Because it’s the nose my parents gave birth to me with, it isn’t as I’d like it.’ (HM, Noto dono no saiki) c. iza ure, ore-ra si-de no yama no sonafe s-eyo hey you you-PL death-go GEN mountain GEN offering do-MR ‘Hey you, make your offering to the mountain of death!’ (C)

Reflexive >

Hearer Designator (2nd person)

ono As noted above in (B) ono ‘self’ retains its reflexive meaning throughout the history of Japanese, at least as far as central dialects are concerned. The same is true of its derivative reduplicated onoono, ‘each’ (distributive), and as we saw in (A2) onore < ono ‘self’ + re substantivizing suffix. Onore in a speaker- or hearer- designating function is the prime historical example of a reflexive serving in the role of a personal pronoun. There is some evidence that the same has occurred in non-central dialects with ono. Both some eastern dialects (particularly Hachijôjima) and some Ryûkyûan dialects attest

384

JOHN WHITMAN

forms from this source in a second person usage, as in the Hachijôjima example (a) from Izuyama (1994). Hirayama (1992) cites this as a derogatory or familiar hearer-designator (= Tôkyô omae) form for Hachijôjima [una], Niigata [una], and Akiyama (Nagano) [ona]. a. ung-ga ik-ou da:-ba ai-mo ik-o-wa (Izuyama 1994: 10) you-GEN go-IMP be-if I-too go-IMP-EMPH ‘If you’re going, I’m going too. Ono also occurs in a clear speaker-designating role for certain classes of speakers in Heian texts. In Genji monogatari, for example, it is a speaker-designator for males, monks and old men. It is for this reason that we are able to tell that the apparition of a beautiful woman that appears at Genji’s bedside in Yûgao to admonish him for his infidelity is not a live woman: b. Ono ga, ito medeta-si to mi-tatemat-uru woba, tazune-omofos-a-de, self NOM very splendid-SS COMP see-HON-RT ACC visit-think-MZ-not kaku koto nar-u koto na-ki φito wo w-i-te ofas-i-te, thus thing be-RT thing not.be-RT person ACC bring-ry-ing come(HON)-RY-ing tokimekas-i-tamaf-u koso, ito mezamasi-ku turak-ere EMPH very unexpected-RY unbearable-IZ (GM, Yûgao) favor-RY-HON-RT ‘That you you should think to not visit me (myself), who considers you most splendid, and bring with you this woman who is of no significance, and favor her with your love, that is most unexpected and difficult to bear’ (D) Hearer Designator (2nd person) >/< Reflexive na A tradition dating back at least to Murayama (1950: 42-43) claims that the hearerdesignator na is derived from an original first person usage. There are two pieces of evidence for the hypothesis. The first comes from examples where ‘self’ is glossed as na in contexts with a first person subject/speaker. These include the glosses ofo-na-mo2ti and ofo-na-muti or ofo-ana-muti for the deionym ‘great-self’s-esteemed one muti’ in the Nihon shoki (see (a) below), and M 9: 755, M 13: 3239). The second piece of evidence comes from combinations of na with a kin or other personal relation term: a. namuti (Honorific 2nd person < na + muti ‘honored person’) b. nabito2 (Familiar 2nd person < na + -ga GENITIVE + pito2 ‘person’). c. nanimo ‘my wife’ (< na + -ga GENITIVE + imo ‘sister, wife’) d. nase ‘my husband’ (< na + se ‘brother, husband’) Ôno Tôru (1978: 305) disputes this hypothesis, arguing that ofo-ana-muti is the correct reading for the deionym,16 and that the examples in (a-d) are appositive constructions. The first counterargument is based on the putative existence of a reflexive ana. The evidence for this form is slight, but Ôno is right that (a)na in this example, as well as in the Man’yôshû examples cited above, functions as a reflexive, not as a first person. The occurrence of rendaku in (8b) and retention of /n/ from genitive -no2 or (more likely) -ga (/nga/) shows that these are genitive compounds, not appositives. But the genitive compound analysis is also perfectly consistent with a reflexive interpretation of na.

PERSONAL PRONOUN SHIFT IN JAPANESE

385

Murayama’s (1950) version of this hypothesis is based on the idea that loss of initial /n/ provides a source for the first person pronoun a from earlier speaker-designating *na. But na co-occurs with a in Ryûkyûan dialects as it does in eighth century sources (Izuyama 1992, 94). Murayama’s hypothesis provides no explanation of how /n/ might be lost in forms with a specifically speaker-designating function, but retained otherwise. Most importantly, na occurs pervasively in Ryûkyûan (Yaeyama, Miyako) dialects as a reflexive form, not as a hearer-designator (Miyara 1930, Izuyama 1992: 113-112). Izuyama provides the following contrast involving na- and a- from Irabe (Miyako) in a context where a teacher asks a pupil if (s)he had his/her mother buy an object. Two answers are possible: (e) i.

aran, a?a-du kat-ta no I-EMPH buy-PERF ‘No, I bought it.’ ii. aran, nara?a-du kat-ta. buy-PERF no self-EMPH ‘No, (I my) self bought it.’

(Izuyama 1992: 112)

Both the Ryûkyûan data and the evidence from eighth century Japanese for speakeroriented uses of na indicate that this form was originally a reflexive, or more precisely, logophoric pronoun. This view is hardly revolutionary, even among researchers who focus on eighth century central dialect materials: most Japanese dictionaries gloss 8th century and subsequent na as second person and/or reflexive (Omodaka et al 1967). Many of these researchers have noted the fact that na appears to be a ‘vowel harmonic’ alternant of reflexive ono2. It may be that lexicalization of this alternant in a second person function involves analogy with first person a. As I noted above, evidence for a productive form ana is poor,17 but in the one case where this form is attested it has a reflexive function; the prothetic vowel may in turn involve analogy with ono. In any event both dialect comparative evidence and evidence from 8th century materials indicate that na originates as a reflexive pronoun. Its lexicalization in a second person function exactly parallels lexicalization of ore (probably < ono2 ‘self’ + -re Pronominal suffix) and subsequently ono2re itself in the same function.18 Among the reflexive functions noted by Izuyama (1994, 1992) in Ryûkyûan dialects is a function as the second member of a reflexive compound, such as Miyako (Ôkami) tu:na ‘each’ < tu: ‘self’ ( ‘body’) + na. Reflexive compounds of this sort, often formed through reduplication such as in 8th century ono2-ono2 ‘self-self, each’ derive a distributive ‘plural’ interpretation. It is possible that this usage is also the source of ‘affectionate’ uses of na as the second member of eighth-century sena (se ‘husband/brother + na) , senana (reduplicated), and Azuma (Eastern) dialect ko1na (ko1 ‘child’ + na ) in the Man’yôshû. Since these examples lack a plural meaning, it is also possible that they are derived from second person na, but other personal pronouns do not show a similar pattern. A second ‘affectionate’ suffix -ne may well have reflected exactly the reduplicated pattern, since -ne in the Man’yôshû appears only after na.19

386

JOHN WHITMAN

Abbreviations (premodern Japanese) RY = ren’yôkei = CONT(INUATIVE) IZ = izenkei = COND(ITIONAL) RT = rentaikei = ADNOM(INAL) SS = shûshikei = CONC(LUSIVE) MZ= mizenkei = IR(REALIS) Premodern Japanese texts cited TM = Taketori mongatari (859) IM= Ise monogatari (900) YM = Yamato monogatari (c. 950) MS = Makura no sôshi (c. 1000) GM = Genji monogatari (1002) SM= Sumidagawa (Yôkyoku: 1500?) KN= Kuronuri (Kyôgen: 1600?) OD= Oridome (Saikaku, c. 1640)

SUBJECT INDEX

387

Subject Index A a-, 307, 309, 311 accompaniment, 50 adjective, 75 subjective ____, 80 adjunct adverbial ____, 215 atode ____, 232 maeni ____, 232 toki ____, 235, 237 adverbial ____ complement, 25-26, 50 after, 218, 229, 233, 245-247 agency, 324-325 agent, 325, 329 agreement, 362, 365 already-learned stage, 295 ambiguity, 61-62, 252, 255-256, 262 American, 196, 202 ____ English, 207, 210 anaphor null ____, 307, 309, 312 zero ____, 332 anaphora associative ____, 127-129, 137 zero ____, 330 anchor, 238 anchored, 226, 228-229, 231, 241, 246, 248 anchoring, 227, 230, 232, 239-240, 242, 244, 247, 249 animacy, 324-325, 340-342, 347-348, 350 ____ strategy, 342 animate, 323, 325, 341-343, 347 antecedent, 85-88, 90-93, 96-97, 999-100, 102-105, 108-115, 117, 200-201, 203, 289-290, 294, 296-298, 306, 307, 312

nomimal ____, 302 aphasia, 321 aphasic, 317, 324, 332, 335, 337, 341-343, 347 ____ patient, 317, 320, 325, 350 ____ speaker, 323 agrammatic ____, 321, 323 anomic ____, 327 fluent ____, 327, 343 non-fluent ____, 327 appearance, 41-44 argument, 330 ____ reversing construction, 14 aspect, 161-162, 185, 221, 223 atode, 215-216, 218, 223-225, 228-230, 232-233, 245, 247 B before, 224, 227, 233, 245-248 bridge verb, 30, 51 Broca, 327 C cardinal ____ NP, 259, 261, 264, case ____ particle, 329, 334 causality, 118 causation, 86-87, 94 cause, 86-87, 115, 329 cause-and-effect, 86 ce, 304-307, 311, 314 certainty, 198-199 Chinese, 374 clause matrix ____, 215-216, 224-228, 233, 238-239, 241-242, 244-248

388

SUBJECT INDEX

subordinate ____, 215-217, 223-229, 233, 235, 238-239, 241-242, 244248 complementizer, 91-92, 118 Complement Principle, 49-50 concessive, 203 conditional, 195-203, 211 ____ meaning, 84-86, 94, 96-98 BA ____, 89, 105, 109, 118 concessive ____, 106 NARA ____, 89-90 TARA ____, 89, 98, 104, 109, 112115, 118 TO ____, 89, 98, 112, 118 configuration, 323 connective temporal ____, 216-217, 225, 232233, 246-247, 249 consequent, 85-88, 90, 92-93, 96-97, 108, 111-112, 114-117, 199, 201 Conservation Principle, 125 contemporaneity, 236-238, 242 contemporaneous, 218-219, 222-223, 226-228, 233-235, 237, 239-241, 243-244, 246, 248 c(ontext)-construable, 42-43 contingency meta-propositional ____, 88-90, 92, 108, 116-117 phenomenal ____, 87, 89-90, 108109, 117 control, 325, 327, 334-335, 343 counterfactual, 197, 203, 205-206, 208, 210-211 ____ conditional, 195, 204 counterfactuality, 196, 207 D dative construction, 35-36, 40 definite, 256, 259, 283 ____ NP, 255, 259, 263 definiteness, 11, 166 deictic, 228, 290, 294, 319 ____ center, 225-226, 228-229, 245, 247

____ expression, 324, 349 ____ verb, 332, 335, 341 deictics, 110 deixis, 341 demonstrative, 360 desirability, 195-196, 201-202, 207, 210211 ____ hypothesis, 196, 201 ____ table, 201 ____ table hypothesis, 202 desirable, 196, 201-204 determination, 161-163 determiner, 290, 314 discourse, 199, 205, 207, 211, 248, 318, 334 ____ deletion, 27 ____ link, 260 ____ -new, 4 ____ -old, 4 ____ processing, 295 direct ____, 280, 332, 335, 370, 372373 experimental ____, 320 indirect ____, 279-280, 282 natural ____, 320 dislocation left ____, 19 right ____, 18-19 d-linked, 12, 259 ____ NP, 274 d-linking, 274 domain context ____, 117 epistemic ____, 117 speech-act ____, 117 donkey ____ anaphora, 265, 284 ____ antecedent, 271 ____ pronoun, 271 ____ sentence, 263-264 double object construction, 35-36, 39-40 E effect, 86-87, 115 ellipsis, 334, 350

SUBJECT INDEX emargination, 19 empathetic identification, 357, 365-367, 376 empathic, 318 ____ focus, 318, 321, 323-325, 331, 335, 341, 347, 349 ____ -focus-first hypothesis, 323324 empathy, 57, 61, 63-64, 68-70, 72-75, 140, 142-144, 154-156, 317, 319, 323-325, 331-332, 337, 347-350, 357, 370-371, 375-376 Speech Act ____ hierarchy, 57-59, 6869, 72-76, 79, 370, 377 English, 202, 218, 245-247, 249, 317, 319, 324-325, 327, 329-330, 332, 335, 337, 342-343, 347, 350, 363 ____ existential there-sentence, 6 ____ presentational there-sentence, 7 ____ there-sentence, 5 entity ____’s motion, 319 attention-getting ____, 319 epistemic, 87-89, 92, 97, 116 ____ operator, 271-273, 277-278, 281-283 ____ scale, 204 epithet, 360-361, 363 error reversal ____, 323 event ____ analysis, 327 exclamation, 162, 178-185, 188 existential ____ construction, 261 ____ sentence, 160, 165-166, 259 experimental, ____ study, 324 extractability, 262 extraction, 255, 261, 284 ____ from NP, 254, 256 WH ____, 262 extraposition ____ link, 47-48, 53

389

F factuality, 83 family-of-question interpretation, 275-278 Flow-of-Information Principle, 27, 36, 39, 45-46 focus, 26-28, 40, 50-51 French, 304, 314 functionalism, 376 functional sentence perspective, 27 functional syntax, 347, 350 experimental ____, 318 future, 219-220, 223-224, 226-229, 231, 235, 239-240, 243, 245-246 ____ eventuality, 85 simple ____, 227-228, 238-239, 241242, 248 G ga, 308 generic, 95, 167 German, 360, 362 grammar ____ of closeness, 68 ____ of restraint, 68 grammaticality ____ judgment, 321 grammaticalization, 362 H head ____ marking, 362 hearer ____ -new, 4 ____ -old, 4 ____ -oriented, 318, 320 Heavy NP Shift, 39-40, 52 higher order event, 87 higher predicate, 87-88, 110, 116 honorific, 359, 364, 373, 375, 377 hypocoristic, 368, 377 hypothetical ____ conditional, 83 ____ meaning, 83-84, 93-94, 98-101, 103-111, 113-116

390

SUBJECT INDEX

I il, 304-307, 311 inanimate, 323, 325, 341-343, 347, 349 indefinite, 253, 256-259, 261-262, 264271, 273-274, 280, 283, 285 ____ NP, 253-254, 259-260, 262263, 275, 279, 284 indefiniteness, 177 indexical ____ function, 362 inferrable, 128-129, 135 information ____ already assimilated, 295 ____ processing, 299, 301 absolute ____ status, 14 already-integrated ____, 299, 303 already-known ____, 298, 312 already-learned ____, 299, 304 already-processed ____, 298, 312 familiar ____, 312 given (old) ____, 3, 25, 28, 50, 111112, 115, 117 important ____, 25-27, 29, 31-32, 3440, 45-46, 48, 50-51 More/Less Important ____ Condition, 27-34, 37-40, 45-49, 51 new ___, 3, 25-26, 28, 50, 111-113, 115, 117, 300, 320 newly-acquired ____, 303, 307 newly-learned ____, 204, 299 novel ____, 292, 294-295, 300, 309, 312 presupposed ____, 302 prior ____, 312 recently-acquired ____, 295 recently-learned ____, 294 relative ____ status, 14 interaction, 323 interpretive nesting requirement (INR), 47-48, 52 inversion locative ____, 18 subject ____, 19 inverted WH-cleft, 297, 313 island

____ -escaping behavior, 261, 264, 268-269, 274, 277-278 ____ -escaping interpretation, 270 -escaping NP, 272, 283 scopal ____, 254, 256, 261, 264, 267268, 271, 283-284 it, 289-290, 296-304, 307, 309, 312-314 Italian ____ inversion, 14 ____ presentational ci-sentence, 9 ____ subject postposing, 12 J Japanese, 196-199, 202-203, 207-208, 210, 215, 217, 219-220, 223, 228, 230, 234, 245, 247-248, 307, 317, 319-320, 325, 327, 329-330, 333-335, 337, 341-343, 347, 349-350, 357, 361, 365 K kin ____ term, 363, 366, 368-369 knowledge, 293-294, 298, 300, 303, 312 prior ____, 293-296, 298-304, 307, 309, 311 ko-, 307, 309, 311 Korean, 196, 198-199, 202-203, 207, 210, 361, 363-364, 374 Middle ____, 374, 378 L locative, 24-25, 31, 42 logophor, 377 logophoric, 271, 273, 282, 370-371 logophoricity, 258 M maeni, 215-216, 218, 223, 225, 227-228, 232-233, 245, 247-248 manner, 25, 50 marked, 335 deictically ____, 337 markedness, 340, 347 Markedness Principle for Discourse-Rule Violations, 36, 40

SUBJECT INDEX means, 24-25, 50 mental space, 195, 204-205 modality, 161-162, 170, 185 model, 319 ____ of sentence production, 323 ____ of the hearer’s mental state, 318 parallel-processing constraint-satisfaction ____, 320 morpheme non-past ____, 220-221, 223, 225228, 232-233, 235, 238, 240, 242, 245, 247-248 non-past tense ____, 247 past ____, 221-224, 226, 229, 232233, 235-236, 242, 244-245 past tense ____, 223, 247 N narrative agrammatic ____, 323 aphasic ____, 332 newly-learned, 295 nonanaphoric, 12 nonspecific, 257, 259-260 ____ definite, 261 ____ interpretation, 251 normal, 324, 327, 335, 337, 341- 343, 347 noun phrase full ____, 318 novel, 293 novelty, 319-320 O objective, 75 ____ predication, 170 ____ statement, 161-162, 165, 171 order ____ of mention, 319, 332, 350 P passive, 347, 349 get-_____, 337, 343, 349 passivization, 332, 335 passivize, 323, 341 past, 216, 218-219, 221, 241, 243, 249

391

simple ____, 218, 222, 226, 229, 241242, 244-246 patient, 341, 347 perfect ____ of result, 220-221, 247 perfective, 222-223 perseveration, 334 person second ____, 363 personalization, 360-361, 363, 365 complete ____, 362 picture ___ -description experiment, 317 point ____ of reference (R), 218-223, 226232, 238-244, 246, 248 ____ of speech (S), 218-223, 226232, 238-248 ____ of the event (E), 218-223, 225235, 237-248 point of view, 161-163, 169-170, 177, 185-186, 347, 367, 375-376 extrinsic ____, 66 intrinsic ____, 66 Polish, 362 pragmatics, 195 predication ____ link, 47-48, 53 ____ relation, 24, 49 preposition stranding, 27, 51 present, 219-220 ____ progressive, 220-221 simple ____, 220-221, 226-228, 238239, 241, 245-248 presupposition, 28 existential ____, 256-257, 281 prior to, 218-219, 223-229, 232-238, 241244, 246, 248 pronoun, 289-290, 307, 312, 318-319, 330, 358, 363 first person ____, 366, 370 inclusive ____, 375 intrapersonal ____ shift, 358, 360, 366, 369, 374, 376 personal ____, 359

392

SUBJECT INDEX

possessive ____, 58-59, 61, 63, 65, 68-69 second person ____, 370 third person ____, 363 proposition, 318, 329-333, 335 protagonist, 325, 332, 343 prototype, 98, 160-161 psychological ____ primitive, 348 psych-verb, 50 Q qualification, 160, 177-178, 182, 185 quantification, 160, 164, 169, 176, 181, 185 quantificational ____ interpretation, 254 quotation injected ____, 372 R realis, 204 reference arbitrary ____, 310 narrow ____, 295, 306, 309 wide ____, 295, 305, 307-308, 310311 reference point, 159, 161, 180-181 referential, 267 ____ interpretation, 254, 270 ____ knowledge, 60 ____ NP, 267 referentiality, 265, 270, 278-279 reflexive, 369-370, 374, 377 long-distance ____, 370, 372-373, 375 picture noun ____, 371 rheme, 18 rhetorical question, 183 ru, 216-221, 223-228, 232-233, 235, 238, 240, 242, 245, 247-248 Rumanian, 362 S salience, 319-320

scopal value, 257 scope, 251-254, 256-257, 259, 263, 265, 268-270, 272-273, 278 maximal ____ reading, 270 narrow-____ reading, 266 wide ____ reading, 270 widest-____ reading, 266 semantics, 195, 197, 211 sentence ____ production, 317-318 active-voice ____, 321, 332 agent-first ____, 321 simultaneous, 223, 226, 230, 234 so-, 307, 309, 311 social accommodation, 300 sore, 309 Spanish, 360 speaker ____’s extrinsic view, 79 ____’s impulse, 319-320 normal ____, 318, 323 primary ____, 279-282 secondary ____, 279-282, 285 specific, 257-260, 266-267, 270, 279 ____ indefinite, 284 ____ interpretation, 251 ____ NP, 251-252, 255-256, 258, 265, 268, 271-274, 280-282 ____ reference, 310 specificity, 251-253, 255-258, 261-263, 265, 269-270, 273, 277-279, 282-283 speech act, 88, 92, 116, 118 ____ conditional, 87 stress, 290 stressed, 290 strong, 264 subject, 331-332, 334, 337, 340-343, 347 ____ of the clause, 324 aphasic ____, 318, 327 control ____, 337 grammatical ____, 324 normal ____, 350 normal control ____, 318 subjective, 75 ____ appreciation, 166-167, 171, 176177

SUBJECT INDEX ____ assertion, 162, 169, 171, 174, 178, 185 ____ assessment, 162, 182 ____ evaluation, 161, 163, 165, 169, 171, 177 ____ expression, 138, 140-142, 151152 ____ predication, 169, 173, 177-178, 187 switch-reference, 114 T ta, 216-219, 221-224, 226, 229, 232-233, 235-236, 242, 244-245, 247 tense, 218-219, 221-223, 225, 245, 247248 ____ morpheme, 215, 218, 228 absolute ____, 215, 225, 228 fixed ____, 218 future ____, 218-219 matrix ____, 215-216, 240 non-past ____, 215, 224 non-past ____ morpheme, 217 past ____, 215, 223 perfect ____, 219 present ____, 218 relative ____, 215, 225, 228 simple past ____, 221 territory, 111 that, 289-290, 296-304, 307, 309, 312314 thematic relation, 323 theme, 18 this, 312-314 toki, 215, 217-218, 232-234, 236, 238245, 247 topic, 25-26, 42, 113, 318, 319, 337, 341342, 347-348 discourse ____, 131-132, 134-135 topicalization, 332, 335, 337 topicalize, 323, 341 transitivity, 180, 185, 188 truth, 83, 85-86, 110, 117, 195 ____ condition, 196 ____ function table, 85

393

____ table, 85-86, 199-201 ____ value, 85-86, 99, 102-103, 105, 108, 195, 200 U uncertainty, 198 undergoer, 323, 325, 332, 341, 347 undesirable, 196, 201-204, 210-211 unexpectedness, 319 unmarked, 335 unstressed, 290 V vantage point, 97 verb, 330-331, 334, 343 ____ voice, 331 achievement ____, 104 active ____, 331 light ____, 364 non-stative ____, 219-220, 227 passive-voice ____, 341 stative ____, 220, 223 Vietnamese, 363 viewpoint, 324, 331-332, 335, 337, 347348 ____ analysis, 331 marked-____, 335, 341 voice active ____, 337, 343 passive ____, 337, 341 volition, 350 volitionality, 348, 350 W weak ____ NP, 264 when, 232, 236, 238-245 WH-extraction, 276 WH-movement, 274, 276 word-order, 329 ____ error, 321 canonical ____, 317 Z zibun, 370-372, 375, 378

NAME INDEX

395

Name Index A Abusch, Dorit, 254, 261, 265-266, 268270, 273, 278, 283 Aissen, Judith, 5, 8 Akatsuka, Noriko, 111, 117, 196, 201204, 212, 295, 312 Arita, Setsuko, 117-118 B Bach, Emmon, 254 Baker, Carl L., 77 Baker, Carl R., 179 Baltin, Mark R., 52 Barwise, Jon, 254, 264, 284 Bates, Elizabeth, 320-321, 348 Bazenov, Nikolaj M., 78 Belletti, Adriana, 12 Benveniste, Emile, 198, 360-361 Berruto, Gaetano, 9 Birner, Betty, 3, 5-7, 12, 14, 18, 284 Bock, Katherine J., 320 Bolinger, Dwight, 170, 187, 274 Bresnan, Joan, 18, 179 Brown, Gillian, 128 Browne, Wayles, 377 Burzio, Luigi, 12 C Calabrese, Andrea, 12-13, 19 Cantrall, William R., 77 Chafe, Wallace, 4, 28, 320 Chatras, Pascale M., 188 Choe, Hyon-bae, 378 Chomsky, Noam, 51, 58, 179, 309 Chvany, Catherine V., 61 Cinque, Guglielmo, 274-276, 278 Clancy, Patricia M., 196, 201-203, 212

Comorovski, Ileana, 274 Comrie, Bernard, 83, 86-87, 221, 225, 247 Cooke, Joseph, 362 Cooper, Robin, 254, 264, 284 Coppieters, René, 304, 306-307 Corblin, Francis, 128 Culicover, Peter, 5, 24, 35, 41-45, 47, 4950, 52 Culioli, Antoine, 159, 187-188 D Declerck, Renaat, 297, 306 Diesing, Molly, 254, 256, 259, 265, 267, 284 Dik, Simon C., 131-133 Donnellan, Keith, 252-253, 255-256, 261 Dowty, David, 117, 119 E Enç, Mürvet, 252, 256-259, 261263, 274, 284 Erdmann, Peter, 8 Erteschik-Shir, Nomi, 51, 254 F Fauconnier, Giles, 197, 204-207, 210 Ferguson, Charles, 195 Fiengo, Robert, 255-256, 258, 261-262 Finkel, Aleksandr M., 78 Firbas, Jan, 4, 18 Fodor, Janet, 47 Fodor, Jerry A., 252-256, 258, 261, 263270, 273, 275, 278 Fox, Barbara, 325 Frajzyngier, Zygmunt, 313 Frege, Gottlob, 280 Fujita, Ikuyo, 334

396

NAME INDEX

G Givón, Talmy, 348 Greenfield, Patricia M., 320 Grice, Paul H., 199 Guéron, Jacqueline, 24, 29, 34, 41-45, 4952, 257 Guillemin-Flescher, Jacqueline, 187-188

H Haiman, John, 113, 119, 195 Hankamer, Jorge, 294 Hasunuma, Akiko, 96, 118 Harris, Martin B., 117 Hawkins, John A., 125-127, 135 Hayashi, Shirô, 370 Heim, Irene, 251, 253, 263, 265, 269, 271, 283-284 Higginbotham, James, 255-256, 258, 261262, 265, 279, 282 Hintikka, Jaakko, 265, 271-273, 277, 281283 Hirayama, Teruo, 377, 383 Hooper, Joan, 187 Hopper, Paul, 188 Horn, George, 254 Hornstein, Norbert, 215, 221, 245-249, 271 I Iida, Masayo, 156 Inoue, Kazuko, 156 Isard, Stephen, 302 Iwasaki, Shoichi, 207, 212 Izuyama, Atsuko, 378, 383, 384-385 J Jackson, Frank, 195 Jacobsen, Wesley M., 118-119, 249 Jakobson, Roman, 77 Jasperson, Robert, 313 Johnson, Kyle, 47, 52 K Kaburaki, Etsuko, 57, 123, 125, 138, 156, 317, 370, 377

Kameyama, Megumi, 156 Kamio, Akio, 51, 111, 186-187, 378 Kamp, Hans, 253, 263, 268 Kaplan, David, 254 Karmiloff-Smith, Annette, 320, 348 Karttunen, Lauri, 251-253, 261, 269 Kawanishi, Yumiko, 207, 210 Kennedy, Becky, 51, 265 King, J., 265 Klenin, Emily, 58, 61, 77-78 Koide, Keiichi, 112, 119 Komatsu, Noriko, 112, 119 Koster, Jan, 51 Kripke, Saul, 251, 253-254 Krivinkova, Katerina, 78-79 Kuno, Susumu, 25, 27-28, 36, 40, 46, 4952, 57, 59, 63- 65, 69, 74-77, 79, 84, 90, 96, 123, 125, 138, 142, 156-157, 186187, 254, 258, 280, 284-285, 308-309, 317, 323-324, 331, 334, 349, 357, 370371, 376-377 Kupferman, Lucien, 186 Kuroda, Shige-Yuki, 156 L Lambrecht, Knud, 4-5, 12 Larson, Richard, 5, 52 Lee, Ki-moon, 374, 378 Lee, Kiri, 247 Levelt, Willem, 318 Levin, Lori, 18 Linde, Charlotte, 303-304, 314 Linebarger, Marcia C., 321 Loebell, Helga, 320 Lü, Shu-hsiang, 374-375 Ludlow, Peter, 251-254, 258, 262, 265, 267, 269-270 M Machida, Ken, 223 Marin, Oscar S. M., 321, 323-324 Martin, Samuel E., 374 Masuoka, Takashi, 117 Matsushita, Daisaburoo, 118 May, Robert, 271 Mayes, Patricia, 203

NAME INDEX McCawley, James D., 77 McCawley, Noriko A., 140, 156 McClelland, James L., 320 McNally, Louise, 5 McNulty, Ellen, 52 Menn, Lise, 317, 323, 334 Mihara, Kenichi, 247 Milner, Jean C., 179, 183, 188 Milsark, Gary, 256 Minami, Fujio, 370 Miyara, Tôsô, 385 Morishima, Yasunori, 334 Murayama, Shichirô, 384 N Nakajima, Heizo, 52 Nakau, Minoru, 223 Napoli, Donna Jo, 52 Neale, Stephen, 251-254, 258, 262, 265, 267, 269-270 Needham, Rodney, 187 Nichols, Johanna, 362 Nitta, Yoshio, 361 Nomoto, Kikuo, 370 O Obler, Loraine K., 317, 323 Ohso, Mieko, 156 Omodaka, Hisakata, 385 Ono, Tôru, 384 Ono, Tsuyosi, 210, 319 Oyakawa, Takatsugu, 156 P Penhallurick, John, 18 Pesetsky, David, 47, 259, 274 Peskovskij, Aleksandr M., 65 Pinchon, Jacqueline, 186 Pinto, Manuela, 12 Pottier, Bernard, 186 Prince, Ellen, 4-6, 50, 125-127, 129-130, 135 Q Quine, Willard van Orman, 280

397

Quinn, Charles, 118 Quirk, Randolph, 27, 46 R Ramstedt, G. J., 361 Reichenbach, Hans, 218-219, 222, 248 Reilly, Judy S., 116, 195 Reinhart, Tanya, 77, 263-264, 271, 284 Reuland, Eric, 77 Riegel, Martin, 188 Roberts, Ian, 52 Rochemont, Michael, 5, 24, 27, 41-45, 47, 49-50 Rothstein, Susan, 52 Rugaleva, Anelya, 78-79 Rumelhart, David E., 320 Ruys, Eduard, 265 S Saccon, Gabriella, 4, 12-13, 19 Saffran, Eleanor, 321, 323-324 Sag, Ivan, 252-256, 258, 261, 263-270, 273, 275, 278, 294 Saita, Izumi, 112, 119 Sakakibara, Sonoko, 156 Sasanuma, Sumiko, 334 Schiffman, Rebecca J., 302, 304 Schodt, Frederick, 349 Schwarts, Myrna F., 321, 323-324 Seidensticker, Edward, 378 Siewierska, Anna, 132-133 Smith, Joshua H., 320 Soga, Matsuo, 223 Sohn, Sung-Ock S., 203 Sridhar, S. N., 319, 348 Strauss, Susan, 196, 201-203, 207, 210, 212 Sunagawa, Yuriko, 247 Suzuki, Ryoko, 319 Suzuki, Takao, 357, 365-369, 376-377 Sweetser, Eve, 117 T Takami, Ken-ichi, 27, 49-52, 76-77, 151, 156, 314

398 Teramura, Hideo, 247 ter Meulen, Alice, 195 Thompson, Sandra A., 187-188 Timberlake, Alan, 61, 63, 71, 77-78 Togo, Yuji, 305-306, 314 Tomlin, Russell, 350 Tomoda, Etsuko K., 156 Traugott, Elizabeth, 195 U Upfield, Arther W., 4, 8 Uspenskij, Boris A., 77 V Vendler, Zeno, 103 Veyrenc, Jacques, 78 W Wagner, Rene-Louis, 186

NAME INDEX Ward, Gregory, 3, 5-7, 12, 18, 129-130, 284 Whitman, John, 378 Wilkins, Wendy, 52 Williams, Edwin, 52, 257, 285 Wulfeck, Beverly, 320-321, 348 Y Yokoyama, Olga T., 58, 60, 65, 68, 70, 7780 Yoshida, Eri, 210 Yule, George, 128 Z Zemskaja, Elena A., 77 Zimmerman, Thomas Ede, 257 Zribi-Hertz, Anne, 77, 79

In the PRAGMATICS AND BEYOND NEW SERIES the following titles have been published thus far or are scheduled for publication: 1. WALTER, Bettyruth: The Jury Summation as Speech Genre: An Ethnographic Study of What it Means to Those who Use it. Amsterdam/Philadelphia, 1988. 2. BARTON, Ellen: Nonsentential Constituents: A Theory of Grammatical Structure and Pragmatic Interpretation. Amsterdam/Philadelphia, 1990. 3. OLEKSY, Wieslaw (ed.): Contrastive Pragmatics. Amsterdam/Philadelphia, 1989. 4. RAFFLER-ENGEL, Walburga von (ed.): Doctor-Patient Interaction. Amsterdam/Philadelphia, 1989. 5. THELIN, Nils B. (ed.): Verbal Aspect in Discourse. Amsterdam/Philadelphia, 1990. 6. VERSCHUEREN, Jef (ed.): Selected Papers from the 1987 International Pragmatics Conference. Vol. I: Pragmatics at Issue. Vol. II: Levels of Linguistic Adaptation. Vol. III: The Pragmatics of Intercultural and International Communication (ed. with Jan Blommaert). Amsterdam/Philadelphia, 1991. 7. LINDENFELD, Jacqueline: Speech and Sociability at French Urban Market Places. Amsterdam/Philadelphia, 1990. 8. YOUNG, Lynne: Language as Behaviour, Language as Code: A Study of Academic English. Amsterdam/Philadelphia, 1990. 9. LUKE, Kang-Kwong: Utterance Particles in Cantonese Conversation. Amsterdam/Philadelphia, 1990. 10. MURRAY, Denise E.: Conversation for Action. The computer terminal as medium of communication. Amsterdam/Philadelphia, 1991. 11. LUONG, Hy V.: Discursive Practices and Linguistic Meanings. The Vietnamese system of person reference. Amsterdam/Philadelphia, 1990. 12. ABRAHAM, Werner (ed.): Discourse Particles. Descriptive and theoretical investigations on the logical, syntactic and pragmatic properties of discourse particles in German. Amsterdam/Philadelphia, 1991. 13. NUYTS, Jan, A. Machtelt BOLKESTEIN and Co VET (eds): Layers and Levels of Representation in Language Theory: a functional view. Amsterdam/Philadelphia, 1990. 14. SCHWARTZ, Ursula: Young Children’s Dyadic Pretend Play. Amsterdam/Philadelphia, 1991. 15. KOMTER, Martha: Conflict and Cooperation in Job Interviews. Amsterdam/Philadelphia, 1991. 16. MANN, William C. and Sandra A. THOMPSON (eds): Discourse Description: Diverse Linguistic Analyses of a Fund-Raising Text. Amsterdam/Philadelphia, 1992. 17. PIÉRAUT-LE BONNIEC, Gilberte and Marlene DOLITSKY (eds): Language Bases ... Discourse Bases. Amsterdam/Philadelphia, 1991. 18. JOHNSTONE, Barbara: Repetition in Arabic Discourse. Paradigms, syntagms and the ecology of language. Amsterdam/Philadelphia, 1991. 19. BAKER, Carolyn D. and Allan LUKE (eds): Towards a Critical Sociology of Reading Pedagogy. Papers of the XII World Congress on Reading. Amsterdam/Philadelphia, 1991. 20. NUYTS, Jan: Aspects of a Cognitive-Pragmatic Theory of Language. On cognition, functionalism, and grammar. Amsterdam/Philadelphia, 1992. 21. SEARLE, John R. et al.: (On) Searle on Conversation. Compiled and introduced by Herman Parret and Jef Verschueren. Amsterdam/Philadelphia, 1992.

22. AUER, Peter and Aldo Di LUZIO (eds): The Contextualization of Language. Amsterdam/Philadelphia, 1992. 23. FORTESCUE, Michael, Peter HARDER and Lars KRISTOFFERSEN (eds): Layered Structure and Reference in a Functional Perspective. Papers from the Functional Grammar Conference, Copenhagen, 1990. Amsterdam/Philadelphia, 1992. 24. MAYNARD, Senko K.: Discourse Modality: Subjectivity, Emotion and Voice in the Japanese Language. Amsterdam/Philadelphia, 1993. 25. COUPER-KUHLEN, Elizabeth: English Speech Rhythm. Form and function in everyday verbal interaction. Amsterdam/Philadelphia, 1993. 26. STYGALL, Gail: Trial Language. A study in differential discourse processing. Amsterdam/ Philadelphia, 1994. 27. SUTER, Hans Jürg: The Wedding Report: A Prototypical Approach to the Study of Traditional Text Types. Amsterdam/Philadelphia, 1993. 28. VAN DE WALLE, Lieve: Pragmatics and Classical Sanskrit. Amsterdam/Philadelphia, 1993. 29. BARSKY, Robert F.: Constructing a Productive Other: Discourse theory and the convention refugee hearing. Amsterdam/Philadelphia, 1994. 30. WORTHAM, Stanton E.F.: Acting Out Participant Examples in the Classroom. Amsterdam/Philadelphia, 1994. 31. WILDGEN, Wolfgang: Process, Image and Meaning. A realistic model of the meanings of sentences and narrative texts. Amsterdam/Philadelphia, 1994. 32. SHIBATANI, Masayoshi and Sandra A. THOMPSON (eds): Essays in Semantics and Pragmatics. Amsterdam/Philadelphia, 1995. 33. GOOSSENS, Louis, Paul PAUWELS, Brygida RUDZKA-OSTYN, Anne-Marie SIMONVANDENBERGEN and Johan VANPARYS: By Word of Mouth. Metaphor, metonymy and linguistic action in a cognitive perspective. Amsterdam/Philadelphia, 1995. 34. BARBE, Katharina: Irony in Context. Amsterdam/Philadelphia, 1995. 35. JUCKER, Andreas H. (ed.): Historical Pragmatics. Pragmatic developments in the history of English. Amsterdam/Philadelphia, 1995. 36. CHILTON, Paul, Mikhail V. ILYIN and Jacob MEY: Political Discourse in Transition in Eastern and Western Europe (1989-1991). Amsterdam/Philadelphia, 1998. 37. CARSTON, Robyn and Seiji UCHIDA (eds): Relevance Theory. Applications and implications. Amsterdam/Philadelphia, 1998. 38. FRETHEIM, Thorstein and Jeanette K. GUNDEL (eds): Reference and Referent Accessibility. Amsterdam/Philadelphia, 1996. 39. HERRING, Susan (ed.): Computer-Mediated Communication. Linguistic, social, and cross-cultural perspectives. Amsterdam/Philadelphia, 1996. 40. DIAMOND, Julie: Status and Power in Verbal Interaction. A study of discourse in a closeknit social network. Amsterdam/Philadelphia, 1996. 41. VENTOLA, Eija and Anna MAURANEN, (eds): Academic Writing. Intercultural and textual issues. Amsterdam/Philadelphia, 1996. 42. WODAK, Ruth and Helga KOTTHOFF (eds): Communicating Gender in Context. Amsterdam/Philadelphia, 1997. 43. JANSSEN, Theo A.J.M. and Wim van der WURFF (eds): Reported Speech. Forms and functions of the verb. Amsterdam/Philadelphia, 1996. 44. BARGIELA-CHIAPPINI, Francesca and Sandra J. HARRIS: Managing Language. The

discourse of corporate meetings. Amsterdam/Philadelphia, 1997. 45. PALTRIDGE, Brian: Genre, Frames and Writing in Research Settings. Amsterdam/Philadelphia, 1997. 46. GEORGAKOPOULOU, Alexandra: Narrative Performances. A study of Modern Greek storytelling. Amsterdam/Philadelphia, 1997. 47. CHESTERMAN, Andrew: Contrastive Functional Analysis. Amsterdam/Philadelphia, 1998. 48. KAMIO, Akio: Territory of Information. Amsterdam/Philadelphia, 1997. 49. KURZON, Dennis: Discourse of Silence. Amsterdam/Philadelphia, 1998. 50. GRENOBLE, Lenore: Deixis and Information Packaging in Russian Discourse. Amsterdam/Philadelphia, 1998. 51. BOULIMA, Jamila: Negotiated Interaction in Target Language Classroom Discourse. Amsterdam/Philadelphia, 1999. 52. GILLIS, Steven and Annick DE HOUWER (eds): The Acquisition of Dutch. Amsterdam/ Philadelphia, 1998. 53. MOSEGAARD HANSEN, Maj-Britt: The Function of Discourse Particles. A study with special reference to spoken standard French. Amsterdam/Philadelphia, 1998. 54. HYLAND, Ken: Hedging in Scientific Research Articles. Amsterdam/Philadelphia, 1998. 55. ALLWOOD, Jens and Peter Gärdenfors (eds): Cognitive Semantics. Meaning and cognition. Amsterdam/Philadelphia, 1999. 56. TANAKA, Hiroko: Language, Culture and Social Interaction. Turn-taking in Japanese and Anglo-American English. Amsterdam/Philadelphia, 1999. 57 JUCKER, Andreas H. and Yael ZIV (eds): Discourse Markers. Descriptions and theory. Amsterdam/Philadelphia, 1998. 58. ROUCHOTA, Villy and Andreas H. JUCKER (eds): Current Issues in Relevance Theory. Amsterdam/Philadelphia, 1998. 59. KAMIO, Akio and Ken-ichi TAKAMI (eds): Function and Structure. In honor of Susumu Kuno. 1999. 60. JACOBS, Geert: Preformulating the News. An analysis of the metapragmatics of press releases. 1999. 61. MILLS, Margaret H. (ed.): Slavic Gender Linguistics. 1999. 62. TZANNE, Angeliki: Talking at Cross-Purposes. The dynamics of miscommunication. 2000. 63. BUBLITZ, Wolfram, Uta LENK and Eija VENTOLA (eds.): Coherence in Spoken and Written Discourse. How to create it and how to describe it.Selected papers from the International Workshop on Coherence, Augsburg, 24-27 April 1997. 1999. 64. SVENNEVIG, Jan: Getting Acquainted in Conversation. A study of initial interactions. 1999. 65. COOREN, François: The Organizing Dimension of Communication. 2000. 66. JUCKER, Andreas H., Gerd FRITZ and Franz LEBSANFT (eds.): Historical Dialogue Analysis. 1999. 67. TAAVITSAINEN, Irma, Gunnel MELCHERS and Päivi PAHTA (eds.): Dimensions of Writing in Nonstandard English. 1999. 68. ARNOVICK, Leslie: Diachronic Pragmatics. Seven case studies in English illocutionary development. 1999.

69. NOH, Eun-Ju: The Semantics and Pragmatics of Metarepresentation in English. A relevance-theoretic account. 2000. 70. SORJONEN, Marja-Leena: Recipient Activities Particles nii(n) and joo as Responses in Finnish Conversation. n.y.p. 71. GÓMEZ-GONZÁLEZ, María Ángeles: The Theme-Topic Interface. Evidence from English. 2001. 72. MARMARIDOU, Sophia S.A.: Pragmatic Meaning and Cognition. 2000. 73. HESTER, Stephen and David FRANCIS (eds.): Local Educational Order. Ethnomethodological studies of knowledge in action. 2000. 74. TROSBORG, Anna (ed.): Analysing Professional Genres. 2000. 75. PILKINGTON, Adrian: Poetic Effects. A relevance theory perspective. 2000. 76. MATSUI, Tomoko: Bridging and Relevance. 2000. 77. VANDERVEKEN, Daniel and Susumu KUBO (eds.): Essays in Speech Act Theory. n.y.p. 78. SELL, Roger D. : Literature as Communication. The foundations of mediating criticism. 2000. 79. ANDERSEN, Gisle and Thorstein FRETHEIM (eds.): Pragmatic Markers and Propositional Attitude. 2000. 80. UNGERER, Friedrich (ed.): English Media Texts – Past and Present. Language and textual structure. 2000. 81. DI LUZIO, Aldo, Susanne GÜNTHNER and Franca ORLETTI (eds.): Culture in Communication. Analyses of intercultural situations. n.y.p. 82. KHALIL, Esam N.: Grounding in English and Arabic News Discourse. 2000. 83. MÁRQUEZ REITER, Rosina: Linguistic Politeness in Britain and Uruguay. A contrastive study of requests and apologies. 2000. 84. ANDERSEN, Gisle: Pragmatic Markers and Sociolinguistic Variation. A relevance-theoretic approach to the language of adolescents. 2001. 85. COLLINS, Daniel E.: Reanimated Voices. Speech reporting in a historical-pragmatic perspective. 2001. 86. IFANTIDOU, Elly: Evidentials and Relevance. n.y.p. 87. MUSHIN, Ilana: Evidentiality and Epistemological Stance. Narrative Retelling. n.y.p. 88. BAYRAKTAROGLU, Arin and Maria SIFIANOU (eds.): Linguistic Politeness Across Boundaries. Linguistic Politeness Across Boundaries. n.y.p. 89. ITAKURA, Hiroko: Conversational Dominance and Gender. A study of Japanese speakers in first and second language contexts. n.y.p. 90. KENESEI, István and Robert M. HARNISH (eds.): Perspectives on Semantics, Pragmatics, and Discourse. A Festschrift for Ferenc Kiefer. 2001. 91. GROSS, Joan: Speaking in Other Voices. An ethnography of Walloon puppet theaters. n.y.p. 92. GARDNER, Rod: When Listeners Talk. n.y.p. 93. BARON, Bettina and Helga KOTTHOFF (eds.): Gender in Interaction. Perspectives on feminity and masculinity in ethnography and discourse. n.y.p.