11111111111 1111111111111111 11 1111111111111111 11111111111
Cognitive Linguistics Res吾吾rcn
W 川C
问 OUTON
DE GRUYTER
Studies in Linguistic Motivation
国
Cognitive Linguistics Research 28
Editors René Dirven Ronald W. Langacker John R. Taylor
4EL·-i
n
出LM
∞
内υρLV
MBud
GW mUH ri n~k vd4tri >
ZZeng
-
HO 551 -τt古ðies
-
il
-
|
外院图书馆 105
--
ll-
3 3035218 5
in Linguistic Motivation Wω 阳U
HUU Ja
[
-n et nh ρlv
k
JE9uτE
dRE Envd wdAUha
rA
得以只 ek
外门琵恕 J 盼阳川回回常简
绵阳楼窍
Mouton de Gruyter Ber1 in . New York
||| |
Mouton de Gruyter (formerly Mouton, Th e Hague) is a Division of Walter de Gruyter GmbH & Co. KG , Berlin
@ Printed on acid-free paper which falls within the guidelines of the ANSI to ensure pe口nanence and durability.
Library 01 Congress Cataloging-in-Publication Data
Studies in linguistic motivation / edited by Günter Radden , KlausUwe Panther. p. cm. 一 (Cognitive linguistics research ; 28) Inc1udes bibliographical references and index. ISBN 3-11-018245-9 (hardcover : alk. paper) 1. Linguistic analysis (Lin伊istics) 2. Motivation (Psychology) I. Radden , Günter. II. Panther, Klaus-Uwe , 1942- III. Series. P126.S78 2004 410-dc22 2004018907
ISBN 3 11 018245 9 Bibliographi险的iformation
published by Die Deutsche Bibliothek
Die De呗tsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at . 。 Copyright 气E
2004 by Walter de Gruyter GmbH & Co. KG , D-I0785 Berlin
rights reserved. inc1uding those oftranslation into foreign languages. No part ofthis book ?二 X r.:produced or transmitted in any form or by any means , electronic or mechanical , :斗，土 g rhot ()C opy. recording , or any information storage and retrieval system , without permission in writing from the publisher. Printed in Ge口nany
Preface
The present volume grew out of an international workshop "Motivation in . Grammar" organized by the editors at Hamburg University in 1999. Over the years , the scope of the original publication project widened considerably. Apart from a number ofpapers on motivation in grammar we decided to invite contributions focusing on additional aspects of motivation. The s位ucture of this volume is now organized around types of motivation rather than levels of linguistic description. The contributions of the book re f1 ect the manifold facets of motivation as a linguistic concept, both in its explanatory potential and the theoretical problems inherent in accounting for language in functional and cognitive terms. Hence , the rather broad title given to the collection of papers: Studies in Linguistic Motivαtion. We gratefully acknowledge the generous support of the conference by 由e German Research Foundation (DFG) and the Freie und Hansestadt Hamburg. Our gratitude also goes to the editors of the series Cognitive Linguistics Research , in particular John Taylor and René Dirven, who strongly supported the publication and provided us with invaluable comments on earlier drafts of the book. Finally, we would like to th缸lk the authors , whose patience was put to a hard tes t.
Klaus-Uwe Panther Günter Radden Hamburg, October 2004
Table of contents
Preface
V
Introduction: Reflections on motivation Günter Radden and Klaus-Uwe Panther
Section 1: Ecological motivation
47
The ecology of constructions John R. Taylor
49
Expressive binomial NPs in Germanic and Romance languages AdFoolen
75
Section 11: Genetic motivation
101
On genetic motivation in grammar Bernd Heine
103
Emergent aspect constructions in Present-Day-English Chistian Koops
121
Section 111: Experiential motivation
155
Spatial experience, lexical struc阳re and motivation: The case of in Vyryan Evans and Andrea Tyler
157
Motivating the uses ofbasic verbs: Li nguistic and extralinguistic considerations John Newman
193
viii Table 01 contents Section IV: Cognitive motivation
219
The concep阳al motivation of fictive motion Teenie Matlock
221
The goal bias in the encoding of motion events Anatol st，吃fanowitsch and Ada Rohde
249
Motivating the composition of Afrikaans reduplications: A cognitive grammar analysis Gerhard B. van Huyssteen
269
Metonymic motivation in anaphoric reference Francisco J. Ruiz de Mendoza Ibáñez and Olga 1. Díez Velasco
293
Predicative adjectives and grammatical-relational polysemy: the role of metonymic processes in motivating cross-linguistic differences Rita Brdar-Szabó and Mario Brdar
321
Metonymy behind grammar: the motivation of the seemingly "irregular" grammatical behavior of English paragon names Antonio Barcelona
357
Author index Subject index List of contributors
375 381 385
Introduction: Reßections on motivation
Günter Radden and Klaus-Uwe Panther
1. An attempt at explicating ‘ motivation in language'
The present volume addresses a problem that is receiving increasing attention in present-day functional and cognitive linguistics: how much of language is motivated? Motivation is generally seen in con位ast to arbitrariD臼s. De Saussure, who is usually cited in connection with 也e principle of arbitrariness of the linguistic sign, was already aware of the limits of "the 泣rational principle of the arbi位ariness ofthe sign": [The principle of arbitrariness] would lead to the worst sort of complication without restriction. But the mind contrives to introduce a principle of order and regularity into certain parts of the mass of signs and this is the role of relative motivation. If the mechanism of language were entirely rational, it could be studied independently. Since the mechanism of language is but a partial correction of a system that is by nature chaotic, however, we adopt the viewpoint imposed by the very nature of language and study it as it limits arbitrariness. There is no language in which nothing is motivated, and our definition makes it impossible ωconceive of a language in which every也ing is motivated. Between the two extremes - a minimum of organization and a minimum of arbitrariness - we find all possible varieties. (de Saussure 1916/ 证 applied
1959: 133). Th臼e
insightful ideas on motivation make de Saussure appear as a precursor of cognitive linguistics rather than the "founding fa由er of structural linguistics": motivation, in his view, is a cognitive principle that makes Imguage meaningful to its speakers and is necessa叩 as a counterbalance to arbitrariness. De Saussure's notion ofrelative motivation is in the spirit of α惠nitive linguistics. Most scholars would probably subscribe to the view 世皿 motivation is a matter of degree along a continuum ranging between
2
Günter Radden
and ιlaus-Uwe Panther
the poles of arbitrariness on the one hand and predictability on the other hand. De Saussure's view of motivation differs 仕om modem cognitive approaches mainly in the perspective taken: he views motivation as a limiting case of arbitrariness , while cognitive linguists tend to see motivation as the norm and consider arbitrariness as the last resort (Lakoff 1987: 346). Similarly, Heine places the burden of proof on those who cling to the dogma of arbitrariness. He argues that since "[h]uman behavior is not arbitrary but [...] driven by motivations" , language structure, which is a product ofbehavior, "must also be motivated". Still, the term ‘ motivation' is not frequently used, let alone defined , in present-day linguistics and authors often seem to presume that ‘ motivation' is a self-explanatory term. In current functional and cognitive linguistics, the notion of motivation is understood in various ways , which are , however, not necessarily mutually exclusive. For example, Hiraga (1 994: 时， very much in the Saussurean spirit, understands motivation in the sense of a "non-arbitrary relationship between form and meaning". More specifically, Heine (1997: 3) regards linguistic forms as motivated if they "are not invented arbitrarily, but are, rather, already meaningful when they are introduced for some specific function." Within a semiotic tradition, Haiman (1980 , 1985) and others restrict the term ‘ motivation' to one type of diagrammatic iconicity, viz. structural resemblance of language to conceived reality , opposing it to isomorphism, i.e. the principle "one 岛口n 一 one meaning". Geeraerts (2002) uses the terms ‘ motivation' and ‘ isomorphism' in a somewhat different fashion , reserving the former for paradigmatic relations between literal and figurative meanings (e.g. metaphor and metonymy) and the latter for one-to-one mappings from non-figurative to figurative syntagmatic levels. Lakoff (1 987: 448) probably has the naïve native speaker in mind when he characterizes motivation as an independently existing link L between some A and some B that "makes sense". Traditionally, the term ‘ motivation' is applied to the form of linguistic units; more recently scholars have also applied it to the extension of senses (e.g. Lakoff 1987, Beitel , Gibbs and Sanders 1997: 243 , Evans and Tyler, this volume). Although there thus is no definitional consensus regarding the notion of motivation in language
lntroduction: R写flections on motivation
3
In contrast to linguistics, in psychology the notion of motivation is finnly established as a theoretical concept. There seems to be a general consensus that "motivation is an intemal state or condition (sometimes described as a need, desire , or want) that serves to activate or energize behavior and give it direction" (Huitt 2001). Both intemal condition and ex-. temal factors are seen as relevant in psychological theories of motivation. Some theories treat motivation as having an intemal bodily locus and as being an impulse that propels the organism into action; other theories focus on the sources in the surrounding environment that pull or push an organism toward them or away from them. The everyday understanding of motivation seems to be close to the scholarly notion used in psychology. In asking a questions such as What motivates Al Qaeda ω carry out terrorist acts against the Unites States? , the speaker expects to be given an explanation about the intemal state or condition of self-professed fundamentalist Muslims that causes or results in aggressive and self-des位uctive behavior against other people believed to be enemies of Islam. This is the sense given by the 0矿òrd English Dictionary (s. v. motivation): "The conscious or unconscious stimulus for action towards a desired goal provided by psychological or social factors; that which gives purpose or direction to behaviour." The common defining elements ofthe psychological and everyday notions ofmotivation are thus: (1) intemal psychological state or condition that (2) activates , energizes, or stimulates (3) a person's behavior or action (4) in a certain direction or towards an intended goal. We believe that the linguistic notion of motivation is not far removed from its use in psychology and everyday language. (i) As a starting point, motivation in language requires a basis that serves as a potential trigger or enablement for a motivational process to operate on , i.e. the fonn an d/or the content of a linguistic unit. We will refer to a given basis as the source of the motivational process. (ii) As a next step , a motivational process is normally triggered by language-independent factors. By ‘ languageindependent factors' we mean factors that operate in language as well as in other cognitive or semiotic systems. Such factors are e.g. experience , percep阳al gestalt principles , as well the ecological niche in which a linguistic
4
Günter Radden and Klaus-Uwe Panther
stable linguistic structures , i.e. become routinized or entrenched in the linguistic system. It is this final entrenched stage of a linguistic unit that is usually referred to as ‘ motivated\We will refer to this final stage as the target of a motivational process. We need to emphasize , however, that motivation in language not only operates on the diachronic level but also on the synchronic leve l. On the basis of the above-mentioned criteria we propose the following working definition for motivation: A linguistic unit (target) is motivated if some of its properties are shaped by a linguistic source (form an dJor content) and language-independent factors. We thus assume that linguistic motivation involves a causal relation. As suggested by the term ‘ shape' in the above definition , however, the notion of causation is non-deterministic. In what follows , we will discuss criteria that are essential to our understanding of motivation in language. We will first , in Section 2 , look into the relation between a motivational source and a motivational target by way of one example. Section 3 discusses motivation as a causal relation and the role of abductive reasoning. 1 Section 4 contrasts the notion of explanation, as favored in generative grammar, with motivational explanations in functional and cognitive linguistics. Section 5 provides a typology of relations between source and target, which may be exploited for motivational pu甲oses. Section 6 considers languageindependent factors of motivation in more detail and categorizes them into various types. Section 7 briefly addresses the topic of multiple and competing motivations. Pinally , in Section 8, we summarize the contributions to this volume , relating them to the concept of motivation proposed above.
2. Source and target in motivational processes Let us illustrate the notions of motivational source and target with the morphological process of compounding. Already de Saussure noted that compounds such as dix-neuf are relatively motivated by their components , i.e. dix and neuf2 Lakoff (1987: 147) observes that the meanings of compounds 1. To do justice to the complex problem of causality and determination is beyond the scope of this introductory chapter (see Stegmül1 er 1969 for a useful introduction to philosophical aspects of causality). 2. De Saussure (1 950: 133) recognizes two relations involved in relative motivation: a syntagmatic and an associative relation. In dix-ne以 the syntagmatic re-
Introduction: R吃flections on motivation
5
are not compositional, hence not predictable 仕om their p町ts. 3 At a more general level , Langacker (2000: 16, 152, 215) argues 也at complex expressions are not constructed out of their components, but that their component structures only correspond to certain facets of them and, in this respect, complex expressions are motivated to some extent (but not predictable). . Compounds are especially interesting complex expressions in 由at they are conventional names that highlight concep阳al parts of a more complex concep阳alization. A compound evokes a conceptual network, or Idealized c。但itive Model (ICM) (Lakoff 1987: 147).ηle concep阳al parts of the complex ICM that are chosen for naming pu甲oses may vary from language to language; as we demonstrate below, they are motivated by languageindependent factors , namely salience, metonymy, and economy. Consider the names given in eleven European languages to the tool called screwdriver in English, which typically looks as in Figure 1.
Fi.伊 re
1. A typical screwdriver
The concepts the names for screwdriver in different languages denote are listed in (1): (1)
a.
‘ screw-世ive-町'
b. ‘ screw(s)-pull-er'
c. ‘ screws-阳口l-er' d. ‘阳rns-screw'
English screwdriver German Schraubenzieher Danish skruetrækker Hungarian csavarhúzó Dutch schroevendraaier French tournevis
lation is expressed by the juxtaposition of dix and neuJ, while the associative (paradigmatic) relation is supported by tenns such as dix-huit , soixante-dix, etc.
3. As John Taylor pointed out to us one might even que可 whether the meaning of any complex expression is ever fully compositional.
6
Günter Radden and Klaus-Uwe Panther
Spanish destornillador Swedish skruvm矿'se/ Finnish ruuvimeisseli, ruuvita/tta Italian cacciavite Portuguese chave de fenda
e. ‘ de/out-screw-er' E ‘ screw-chisel ' g. ‘ stick -inltake-out-screw' h. ‘ key ofcut'
|
叩町 scre阳W怦 怦叫 -d由ri叫 I
I 伽叫
\
f
.•••.
•••.•••
二二正二....... 气;叫.."，...:::.，.........~气\ 阳也::>'T"-n .-T lI. .ro 旧』
飞气、、
x .......
」二一斗~二 飞
‘ screw由iver'
二二/卜\ ACTION !
MATERIAL
SHAPE
.... 二二~一小气 \ - I OBMs面--PG.RP舰队N，5 METAL long thin rod !1: 二队 I\wlhandle at
sCÍ'ew
卜\
attach
飞..detach turn 飞
oneo喇均加饥叫:t...
heli比 ca 剖1 cut to groove at head
/\ 1 another 飞/\飞
one
and blade
chisel
\
long shaft w/metal blade
川le 盯叫 叫 eω 0创the 阳啤哎r en
..../\ pull-out drive-in loosen tighten
Figure 2. The ‘ screwdriver' ICM and its relation to names for ‘ screwdriver' in various languages
We assume that the different lexemes listed in (1) all evoke the same conceptua/ frame or ICM as their meaning. Figure 2 presents a partial ICM of ‘ screwdriver' relating it to some of the lexical items listed in (1). The dot-
Introduction: R笔flections on motivation
7
ted lines identi命 the elements of the ICM that are highlighted as components of the complex expressions. The first thing to note is that the expressions in (1) name no more than two or three elements ofthe complex ICM ‘ screwdriver' , but these parts are sufficient to evoke the whole ICM. Each of these parts thus metonymically. stands for the whole ‘ screwdriver' ICM. Although we do not c1 aim that this sample is representative, we believe that the selection of certain elements of 也e ICM as components for the naming process is not arbitrary: most of the languages select the object operated on by the ins位ument， i.e. the screw, but none of them selects e.g. the object the screw is driven into such as the board, or part of the screwdriver, such as its blade. The screw is obviously a highly salient element of the screwdriver ICM. Most compounds also highlight one of the actions that is characteristically performed with the tool on the screw, such as ‘阳rn' (French) ,‘pull' , which stands for ‘ pullout' (German , Danish, Hungarian), ‘ drive' , which stands for ‘ drive-in' (English),‘ detach, remove' (Spanish) - a screwdriver is , in fact, used to perform all of these and even some more operations, such as fasten , fix , tighten , loosen , etc. Note , however, that ‘ pull-out' and ‘ drive-in' do not literally describe the activity carried out by means of a screwdriver: these ∞ncepts seem to be motivated by analogy to the driving-in and pulling-out of nails. None of the specific actions performed with a screwdriver stands out as particularly salient so that each of the actions is equally appropriate ωstand metonymically for the whole range of actions. Most of the languages looked at also designate the instrumental character with a nominalizing suffix like in English -er. In comparison to the fairly transp町ent expressions used for ‘ screw世iver' in English, German , Danish, Hungarian , Dutch , French , Spanish , and Italian, the expressions used in Swedish, Finnish, and Portuguese are more opaque , i. e. they are relatively little motivated. In Swedish and its Finnish loan translation a tool outside the screwdriver ICM is profiled, chisel. A possible motivation for selecting the word for chisel is its similarity in shape (a long shaftlblade) and/or function to a screwdriver. The Portuguese term chave de fenda involves a metaphor: a screwdriver is applied ωthe cut in the head of the screw lfenda) like a key (chave) to a lock. In Italian, the prese
8
Günter Radden and Klaus-Uwe Panther
introduced and the need arises in a speech cornrnunity to name this new object, e.g. a screwdriver. There is often a phase when several names compete before one of them wins out and becomes the conventional designation of the object in question. In the 19th cen阳可 there were at least three competing names for ‘ screwdriver' in English: screwturner (attested 1831 in the OED) , turn-screw (attested 1801 , 1837 and 1889) and, of course , screwdriver. Note that screwturner and turnscrew select the same conceptual components 仕om the screwdriver ICM as the French word tournevis. All of these terms for 'screwdriv町， are motivated for designating a screwdriver, but , according to Croft' s (2000: 176) "first law of propagation", there is a natural tendency in human languages to conventionalize one of the competing names at the expense of the others. Which among these competing terms is ultimately chosen is largely a matter of arbitrariness. In conclusion, the following conceptual steps can be identified in the motivational process. First, there is the tool screwdriver that has to be named. This tool is associated with a complex ICM (source) , which provides the basis for naming the thing (target). Second, guided by languageindependent factors such as salience, economy , and metonymy , only certain components of the complex ICM get selected and named by a given speech cornrnunity. The coding of these salient parts is sufficient to evoke the whole ICM by means of a PART FOR WHOLE metonymy.
3. 脚:lotivation
as a causal relation
Let us now tum to the notion of causation that is involved in motivation. We will consider an example of grarnrnaticalization: the development of a bound grarnrnatical mo甲heme out of a free lexical item. Heine , Claudi and Hünnemeyer (1991: 188ff.) observe that in the African language Ewe all case markers are derived from verbs. These case markers synchronically "form a continuum ranging from a fully verbal behavior at the one end to a prepositional behavior at the other" (p. 188). For example , the Ewe verb ná ‘ give' has grarnrnaticalized into the case functions BENEFACTIVE, PURPOSE , and DATIVE. 4 The concept ‘ give' with its participant roles ‘ giver' ,‘ reclpl-
4. The usual functional explanation of the grammaticalization of ná ‘ give' into a case marker with the meaning of, for example,‘ BENEFACTIVE' is based on the fact that, with the exception of ná , verbs in Ewe as well as in some other Kwa languages allow no more than two participants. 1n order to introduce an addi-
Introduction: R电(1ections on motivation
9
ent' and ‘ transferred object' constitutes a linguistic source that may be exfurther conceptual elaboration. 白le sense ‘ benefactive' is most likely derived 仕'Om the sense ‘ give' by way of two metonymically-based implicatures: first , NEUTRAL FOR POSITIVE，也rough which the recipient is positively evaluated as BENEFACTIVE, and, second, ACTION (of giving) FOR. SALIENT PARTICIPANT (recipient). The concept ‘ give' is a linguistic source 也at， in conjunction with the above-mentioned implica阳res， is causal in bringing about the meaning ‘ benefactive\But what kind of causal relation is involved in this case? It is not a sufficient cause because the concept ‘ give' not only 位iggers the meaning ‘ benefactive' but, as demonstrated by Newman (1996; this volume) , also gives rise to other meaning extensions such 臼‘ interpersonal communication' (give a ωlk)， ‘ emergence' (give a yell), ‘ causation' (give someone a headache) , etc. It is not a necessary cause either because the concept ‘ benefactive' may also derive 丘'Om other sources than give. For example, in English the semantic role BENEFACTIVE is expressed by the preposition for , which is derived 企'Om the spatial and/or temporal meaning ‘ in front' and ‘ be岛re' (仕om Germanic 节ra 'before in place or time'). This shift in meaning was probably also motivated by a language-independent principle of metonymic reasoning: objects 由at are in 齿。nt of a person are perceptible and accessible and hence potentially beneficial to this person. The notion of causation 由at is involved in. all these examples is that of a contributing cause and comprises both a linguistic source and language-independent 岛.ctors. To summarize, the concept ‘ give' and language-independent factors (such 臼 metonymic and other inferential principles) are neither sufficient nor necessary causes 岛r the development of 由e grammatical function BENEFACTIVE, but they 缸e contributing causes (among others). The same applies to the notion ‘仕ontness' ， which together with inferential principles ploited 岛r
ti'Onal participant, Ewe probably used, at an earlier 阳ge in its hist'Ory , a serial verb c'Onstructi 'On with ‘ give' , which finally grammaticalized int'O a marker 'Of a semantic r'O le. One 'Of 也e examples given by Claudi and Heine (1986: 319), cited in Hiraga (1994: 16), is the f'Oll 'Owing: é -fi ga ná m he-steal m'Oney give me ‘ He stole m'Oney fr'Om me.'
The verb ná in this functi 'On is "desemanticized" t'O也at 'Ofapr，叩'Ositi'On.
.1 0
Günter Radden and Klaus-Uwe Panther
is a contributing cause to the development of English for in the sense of ‘ benefactive' . In con c1 usion , we assume that there are two types of contributing cause 出at are relevant in linguistic motivation: (i) a linguistic source and (ii) language-independent factors. A linguistic source is a linguistic unit as a whole , its content or its form. The other contributing causes can be called language-independent since they involve general cognitive faculties , such as inferential abilities , or metaphoric and metonymic mappings , that are not restricted to language. It is such contributing causes , or factors , that are at work when a linguistic unit, its content or its form is called ‘ motivated\ In using the term contributing cause , or motivational factor , we take a "realistic" stance, acknowledging that it is impossible to exhaustively list the set of causes that are necessary and jointly sufficient to trigger a linguistic prope此y. In fact , the attribution of motivational sources to linguistic phenomena by the linguistic analyst is usually based on post hoc abductive reasoning , i.e. inferencing from some observed fact plus assumed general principles of reasoning (which may be merely probabilistic) to a conc1 usion that "best explains" the observed fact (see also Goldberg 1995: 69ff. and Hopper and Traugott 1993: 64). In the case of the example discussed above , an abductive reasoning process might run as follows: (2)
Observed fact: Target: Ewe ná has the grammatical meaning BENEFACTIVE. Abductive reasoning: (i) Source: ná originally meant ‘ give' , i.e. ‘ Agent causes Recipient to have Objec t' [contributing cause: linguistic source] (ii) The metonymic principle NEUTRAL FOR POSITIVE invites the inference from RECIPIENT to BENEFACTIVE [contributing cause: language-independent factor)] (iii) The metonymic principle EVENT FOR SALIENT PARTICIPANT OF EVENT invites the inference from event of ‘ giving' to ‘ benefiting recipient' of this event [contributing cause: language-independent factor] (iv) The grammatical meaning BENEFACTIVE of ná is caused (= motivated) by the contributing causes (i) 一(i ii).
Introduction: R写flections on motivation
11
The step-by-step process outlined above is meant as an i1lustration of abductive reasoning by linguists , not as a claim about conscious reasoning processes in native speakers. Whether the abductive inferences happened exactly as outlined in (2) is a question that, in hindsight, cannot be answered. We believe, however, that general cognitive principles (including. metonymies) guide native speakers and hearers subconsciously in constructing and comprehending meanings.
4. The notions of motivation and explanation
The notion of motivation is closely related to that of explanation. The description of a motivational relation in language constitutes one type of explanation, but not every explanation is based on motivation. Berg (1 998: 11), in accordance with Popper (1963: 241) , regards an explanation as an act of "establishing a connection between hitherto unconnected things." In principle, both lay people and experts constantly establish connections between hitherto unconnected things and thus provide "explanations". The regularities underlying most linguistic phenomena, however, are below the level of awareness of lay people. For example , the Bavarian and Austrian noun Watschen ‘ slap in the face' is a singular (feminine) noun whose ending in -en suggests that it is a plural form; hence the backformation Watsche as a singular form (witness a recent newspaper headline Keine Watsche für Viatcheslav ( ‘ No slap in the face for [handball player] Viatcheslav'). Lay persons can, in alllikelihood , not account for their usage of Watsche as the singular and Watschen as the plural. The linguistic expert, in contrast, can explain the linguistic processes subconsciously guiding the lay person by resorting e.g. to the notions of feminine singular and plural "gestalts飞 cf. Bratsche ‘ viola' vs. Bratschen ‘ violas' or Sage ‘ legend' vs. Sagen ‘ legends\ As is well-known , linguistic experts often do not agree on what counts 部 a "connection" between two things and hence as an explanation. Explanations are always embedded within a theoretical framework and reflect the researcher' s theoretical commitments. Let us consider how a puzzling phenomenon of English grammar is "explained" within the formalist 仕ame work of generative grammar and how it might be accounted for in terms of motivation. The data are taken from an article by Hoekstra and Kooij (1 988: 38) on the innateness hypothesis , i.e. the conjecture that infants are born with a richly structured universal grammar. These authors observe that
12
Günter Radden and Klaus-Uwe Panther
sentence (3a) has two readings , as opposed to the single reading of sentence (3b): (3)
a. Wh ere did John say that we had ω get off the bus? ‘ John said where that we had to get offthe bus?' (wide scope ofwhere) ‘ John said that we had to get offthe bus where?' (narrow scope ofwhere) b. Wh ere did John ask whether we had to get off the bus? ‘ John asked where whether we had to get offthe bus?' (wide scope ofwhere) 呻 John asked whether we had to get offthe bus where?' (naηow scope ofwhere)
Question (3a) is either about the location at which John said that we had to get off the bus (wide scope of where) or about where the event of getting offthe bus should take place (narrow scope ofwhere). In contrast, question (3b) has only the reading with wide scope, i.e. , it is a question about where John asked whether we had to get offthe bus. Movement (as a transformational process) of where is possible from the embedded that- c1 ause in (3 时， but blocked in cases like (3时， where whether is c1 aimed to form a "barrier" to movement. Thus that- c1 auses allow extraction of elements while whether forms a barrier to this movement operation. This account of extraction in terms of barriers provides an "explanation" in the sense of establishing "a connection between hitherto unconnected things". However, this explanation reeks of circularity: whether forms a "barrier" to movement because where is not moved and extraction of where is blocked because whether forms a barrier. One way out for the formalist approach is to stipulate. that children are born with an innate universal grammar that, on a subconscious level , contains theoretical concepts such as ‘ movement' and ‘ barrier to movement'. In the case of the different structural behavior of that- and whether-c1 auses , a connection is drawn to language acquisition and innate constraints. It is assumed that the child cannot acquire the scope differences between (3a) and (3b) on the basis of empirical evidence alone. Therefore, children must be guided by innate universal constraints on extraction, i.e. barriers are assumed to be p盯t of the innate universal grammar that humans are endowed with. An explanation in terms of universal grammar is rather speculative, and there is no empirical evidence that children are guided by such
Introduction:
R吃flections
on motivation
13
supposedly universal principles in acquiring this particular grammatical contrast. A motivational explanation ofthe contrast between (3a) and (3b) would take other syntax-extemal factors into account: say in (3a) is one of the most neutral or schematic verbs of communication and consequently highly . non-salient. The interpretation that is likely to come to mind first for (3a) is the narrow scope reading of where: the information that is most relevant (岛>regrounded) in this kind of question is the location where we have to get offthe bus rather than where John performed the act of saying. The latter is l臼s relevant (backgrounded) information. If say is replaced by more specific speech act verbs such as insist, proclaim , announce , or indicate, attention is drawn to the communicative act itself performed by the subject participant, i.e. an interpretation with a wide scope ofwhere is far more likely. The same observation applies to the verb ask in (3b) , which is semantically richer than say and therefore automatically more foregrounded. The con位出t between (3a) and (3b) thus does not reside in the presumed syntactic properties of that as opposed to those of whether but is motivated by communicative factors. In (3a) the speaker's communicative goal is to inquire about where we , according to John , have to get off the bus; this communicative goal can be achieved by backgrounding the act of saying and instead focusing on the propositional content of John's utterance. In the default case, main clauses are foregrounded and subordinate clauses convey background information; however, in (3a) this principle is reversed due the low degree of informational content of say. As a result, the syntactically subordinate clause becomes the focus of attention. In contrast, in (3 b) , because of the more specific speech act of asking , the pragmatic focus is immediately on the act of asking and the propositional content of the question is backgrounded. 5 We claim that the cognitive and communicative factors mentioned above , such as communicative goal , background information, focus of attention, and the concep阳al content of the message determine or, at least, influence linguistic structure - in the above case, the constraints on "whextraction". We do not wish to give the impression that every linguistic 5. Similar observations are made by Verhagen (forthcoming) for English and Dutch. He observes that, apart from the verb itself, the possibility of whextraction depends on variables such as the tense of the matrix verb , the complexity of the subject of the ma位ix clause (pronoun vs. lexical NP; person of pronoun, etc.) , and the complexity ofthe matrix clause as a whole.
14
Günter Radden and Klaus-Uwe Panther
phenomenon is totally explainable in motivational terms. However, we regard the search for motivation as the best heuristics to gain insights into the nature of language. Given that motivational explanations do not, in general , reach the level of prediction, many linguists working in a formalist paradigm are inclined to discard motivational accounts of language form and s位ucture as unSClentific. 6 However, the impressive amount of data that has been amassed in the last forty or fifty years of linguistic research strongly invites an account in terms of motivation. In the following sections we provide only a small sample of the available evidence that speaks for ‘ motivation' as an important explanat。可 device in linguistics.
5. Semiotic relations underlying motivational processes Since motivational processes involve language, they necessarily operate on linguistic signs , i.e. they pertain to semiotic relations within the linguistic unit an d/or across linguistic units. In Section 5.1 we list possible basic semiotic relations between source and target, and in Section 5.2 we illustrate the role of basic semiotic relations for motivation.
5.1.
Basic semiotic relations
In this section we propose five basic semiotic relations that may obtain between the form and the content of a linguistic unit or across linguistic units. Figure 3a represents the case of an arbitrary semiotic relation be-
6. Among the few generative linguists who have shown an interest in questions of motivation is Frederick Newmeyer, most recent1y in Newmeyer (2000). Newmeyer is generally skeptical as to motivational explanations of grammatical structure or, as he calls them, "extemal explanations". Nevertheless, he concedes that there are two "prime candidates" (p. 127): The first is the performance theory of constituent order developed by John Hawkins (1 994) , i.e. the thesis that the human parser prefers sentences whose immediate constituents can be identified as quickly as possible. The second is "structure-concept iconicity" (Newmeyer 2000: 129) , i. e the tendency for languages to align grammatical structure and conceptual structure.
Introduction:
R写flections
on motivation
15
tween content and fonn , indicated by a simple line. Motivated semiotic a.
b.
c.
SOURCE
TARGET
CONTENT
CONTENT
CONTENT
FORM
FORM
FORM
TARGET
arbitrary semiotic relation content motivating form
SOURCE
form motivating content
d. SOURCE
e. TARGET
SOURCE
TARGET
\ / \
contentlmotivating content2
formlmotivating form2
Figure 3. Basic semiotic relations
links between content and fonn are represented by means of arrows. Figure 3b illustrates a relation in which a content (source) motivates a fonn (target), Figure 3c depicts a situation in which a fonn (source) motivates a ∞ntent (细rget). Figures 3d and 3e differ from the preceding relations in 也at they do not relate fonn and content within one unit but contents or 岛nns across units. These relations , unless elaborated by other relations , only exist outside the semiotic system: the content-content relation applies ωpurely mental associations and the fonn-fonn relation to uninterpreted physical phenomena. In language, when these two relations are exploited for motivational pu甲oses ， they are always part of a complex relational ∞nfiguration (see Sections 5 .2 .3 and 5.2 .4).
16
5.2.
Günter Radden and Klaus-Uwe Panther
Basic semiotic relations in motivational processes
As shown in Figure 3 there are , in principle, four basic semiotic relations that may be exploited in motivation: (i) a content may motivate a form , as in iconicity, (ii) a form may motivate a content, as in isomorphism or folk etymology, (iii) a content may motivate another content (in conjunction with other relations) , as in polysemy, (iv) a form may motivate another form (in conjunction with other relations) , as in phonological change. (v) Finally, a form-content unit may motivate another form-content unit, as in grammaticalization. In this case, content-content relations and form-form relations apply simultaneously. 5.2.1. Content-form relations
Forms that are motivated by their content are generally regarded as the most typical type of motivation. Particularly convincing cases of contentform motivation are iconicity and metonymy. They represent the situation of a motivated linguistic unit par excellence. In the case of iconicity, the linguistic unit is assumed to reflect the content it expresses. The iconic relation between content and form might ultimately be metonymically motivated: due to its conceived similarity to the concept, the linguistic form may naturally stand as an image (or representation) for the concept. Haiman (1 980) , following Peirce's (1932) taxonomy of signs , distinguishes two basic types of iconicity: imagic iconicity and diagrammatic iconicity. Imagic iconicity applies to a simple sign that resembles its conceived referent. The most obvious applications of imagic iconicity in language are pictograms used in many writing systems. For example, in ancient Chinese a circle with a dot in the middle and rays radiating from it was used to represent the concept ‘ sun'. Later on the iconic pictogram underwent changes both in its form and content: the circle was squared off, the dot was stretched to a line , the rays were dropped, and the content was extended to senses such as 'day' and ‘ warmth\The content-from relationship thus became less transparent or iconic. Since language is usually vocal , the only kinds of image that speech can imitate are sounds or noises. Imagic iconicity in language is therefore described by Taylor (2002: 46) as imitative iconicity. This simplest case is graphically represented in Figure 3b above. It is 可pically restricted to the fairly small set of onomatopoeic words of a language - the majority of simple words being, at least from a synchronic
Introduction: Reflections on motivation
17
point ofview , usually seen as arbitrary. A well-known example ofimitative iconicity is the simple form cuckoo. The phonetic form of cuckoo is felt to resemble the bird's call and, by metonymy , refers to the bird producing the call. It should be mentioned though that the bird' s c可 is not acoustically identical to the phonological shape of the word. To this extent even ono- . matopoeic words have an element of arbitrariness. The motivational force underlying imitative iconicity may be very powerful: thus , the expected pronunciation of the first syllable of cuckoo would have had the STRUT vowel as in cuckold, but the pronunciation with the FOOT vowel , or even with a short version of the GOOSE vowel , "has prevailed as the supposed echo of the bird's cry" (Onions 1966, s.v. cuCkOO).7 It is also noteworthy that cuckoo superseded the Old English word gëac , which lost its original sound-imitative quality through phonetic developments. Diagrammatic iconicity applies to the arrangement of signs that re f1 ects the relationships of their conceived referents. In language , diagrammatic iconicity pertains to linguistic structure and is therefore also described by Taylor (2002: 46) as structural iconicity. This term is preferable to diagrammatic iconicity because the latter subsumes isomorphism (Haiman 1980), which applies to a different type of iconic situation (see 5.2 .2 below). Well-known 可pes of structural iconicity inc1ude linear iconicity (Harη1 came in and sat down as opposed to *Harry sat down and came in) , proximity iconicity (Susan is not happy but content as opposed to *Susan is unhappy but content) , quantity iconicity (full reduplication: Mandarin Chinese xiao ‘ small' , xiaoxiao ‘ very small'; partial reduplication: Finnish yksin ‘ alone' , ypo-yksin ‘ completely alone') , etc. The conceived similarity between source and target in structural iconicity is more abstract than in imitative iconicity. Highly schematic correspondences between a linguistic structure and its referent may sti1l be seen as iconically motivated. Thus , 也e plural morpheme of a noun, which usually carries more sound than the singular 岛rm， may be seen as motivated by the iconic principle of quantity: MORE SOUND IS MORE MEANING. 8 It should be noted though that in struc7. Cf. Ul1mann (1972: 95) for a similar development of coucou and other examples in French. The Latin word cucu/us developed to cocu in Old French, which was felt to be inexpressive and was ousted by the purely imitative coucou. "[I]t has survived, however, in French cocu and English cucko/d as a crudely jocular metaphor based on the notorious habits ofthe bird". 8. Exceptions to the general tendency of the plural form being more weighty phonological1 y are certain declensional classes of Latin as illustrated by oppi-
18
Günter Radden and Klaus-Uwe Panther
tural iconicity the component units , at least in natural language , are always arbitrary. The type of metonymically motivated content-form relation was already illustrated by the ‘ screwdriver' example discussed above. The conceptualization of ‘ screwdriver' was shown to provide the potential for a wide range of complex expressions in various languages. Since the concept of the composite expression is invariably richer than the combined "literal" meanings of its constituents , complex units tend to be motivated by a PART FOR WHOLE metonymy.
5.2.2. Form-content relations
The form-content relation is represented in Figure 3c. It is instantiated by the principle "one form - one meaning" , known as the principle of isomorphism. 9 A corollary of this principle is sameness in form signals sameness in meaning and distinctness in form signals difference in meaning. The principle "sameness of form - sameness of meaning" is exemplified in phonesthemes: certain non-morphemic sound sequences tend to be associated with invariant meanings. 10 For example, many, typically monosyllabic , words starting in /sp/ have unpleasant connotations such as 甲it， 再pew， 再pi/l， 早 ic (derog. for 'P因此o Rica的 ， spot , speck, spy , spα时， etc. The form spam fits this pattem perfectly and may therefore have been chosen to express the
dum ‘ town' vs. oppida ‘ towns' or murus ‘ wall' vs. muri ‘ wa l1 s'. Another
counterexample was pointed out to us by John Taylor: the Maori demonstratives teenei/teenaa ‘ this/tha t' vs. eenai/eenaa ‘ these/those\ 9. Despite Croft's (1 990: 164ff.) criticism of the "unfortunate selection" of the terms ‘ isomorphism' and ‘ motivation' as used by Haiman , we will adopt the latter author's established use of ‘ isomorphism\Croft notes that the term ‘ isomorphism' is used in a different sense in mathematics , i.e. in reference to both coηespondence of elements and relations; it thus includes ‘ motivation' as used by Haiman. Unlike Haiman , who understands ‘ motivation' as ‘ iconic motivation' only , we wi l1 use the term in the widest sense. 10. An exce l1 ent co l1 ection of English phonesthemes is found on Benjamin K. Shisler's homepage Dictionary of English Phonesthem臼.
Introduction: R笔flections on motivation
19
new meaning ‘junk e-mail'; possibly its earlier sense ‘ tinned meat' also con位ibuted to this meaning extension. 11 An interesting example in which phonological form (stress pattern) impacts on highly schematic (grammatical) content is provided by John Taylor (this volume). He observes that the form hamburger has primary stress . on ham and secondary stress on burger, i.e. it exhibits the same stress pattern as countless compounds like dog-lover. Because of this stress pattern the morphological structure hamburg + er has been reinterpreted as ham + burger. Thus burger came to be understood as the head of a compound and ham as its modifier. The reanalysis of the form hamburger as an endocentric compound also motivated a lexical change of meaning: both ham and burger came to be understood as meaning-bearing units so that a hamburger is understood to be a kind ofburger just like a beefsteak is a kind of steak. 12 The reanalysis of hamburger is also an instance of folk etymology, i.e. the belief that a phonological "word" carries meaning 一 in this case, the fact that ham is a lexeme has reinforced the folk analysis. Even on the level of grammatical constructions constant form-meaning correspondences can be obse凹ed. For example, the constructions NP/ 01 NP2 as in a bear 01 a man investigated by Foolen (this volume) and bang goes NP studied by Taylor (this volume) are associated with specific expressive meanings. Even highly schematic constructions display isomorphism in the sense that they impose aspects of their meaning on the lexical content. For example, the transitive construction prototypically expresses an action schema with an Agent participant as the subject and a Patient participant as the direct object. Non-prototypical participant roles in the 11. John Taylor kindly drew our atiention to the fact that spam did not originally have a negative connotation. Its derogatory meaning may have originated in a Monty Python sketch, in which a restaurant offers all dishes with spam: egg and spam; egg bacon and spam; egg bacon sausage and sp缸n; spam bacon sausage and spam, etc. Possibly, some computer nerd adopted the word and its subsequent success was probably fostered by the negative connotation of the phonestheme sp-. 12. However, one could add that the reinte甲retation of hamburger as a compound was probably also reinforced by certain developments in the world of fast food. Once fast food chains put other burgers on their menu such as cheeseburgers , chickenburgers, etc. , "real-world" facts might have established a new conceptual system that in tum extemal1 y motivated (i怠 strengthened) the analysis of burger as a class term.
20
Günter Radden and Klaus-Uwe Panther
subject and object position of transitive sentences are interpreted in accordance with the action schema. For example, the Instrument in the NRA slogan Guns don 't kill people is contrasted with the Agent in People kill people, i.e. , some people see guns as self-acting energy sources and it is exactly this implicature that is refuted in the NRA slogan. Similarly, in The driver honked the horn , the Instrument the horn as a direct object is seen as more strongly affected than if it were expressed as a prepositional phrase. Thus , honking the horn refers to the hom's function of waming while honking with the horn may also refer to situations such as using the hom for fun or as an attention-getter. The principle of isomorphism also implies that difference in form signals difference in meaning. According to this principle, there is no complete synonymy or paraphrase in language. People even tend to associate different meanings with different pronunciations of the same word. 13 Lexical doublets tend to develop different meanings .I ike shirt (from Old English) and ski问(企om Old Norse). A minimal morphological contrast as between economic and economical results in a clear-cut semantic distinction: ‘ pertaining to the economy' vs. ‘ using money/resources carefully\Minimal syntactic contrasts display different syntactic behavior, which indicates that each construal has its own meaning. Hiraga (1 994: 14) cites Bolinger's (1 968) examples , which contradict the claim that the gerund and the infinitive , and the active and passive voice are synonymous. Thus , both the gerund (4a) and the infinitive (4c) are compatible with hypothetical si阳ations but only the gerund (4b) , not the infinitive (4d) , can be used to describe a fac阳al situation. (4)
a. b. c. d.
Waiting would have been a mistake. Waiting has been a mistake. To wait would have been a mistake. *To wait has been a mistake.
5.2.3. Content-content relations A paradigm case for a content-content relation is polysemy , which is regarded by some cognitive linguists as a prime case of motivation (see Sec13. Labov noted that some people think of /va: zJ as a big vase and /vei zJ or /veis/ as a smaIl vase.
Introduction:
R笔flections
on motivation
21
ti'On 1). The study 'Of the "causes", i. e. m'Otivati 'Ons, 'Of semantic changes has a l'Ong phil'Ol'Ogical traditi'O n. Sweetser (1990: 9) 'Observes 曲创 "[s]ynchr'Onic p 'Olysemy and historical change 'O f meaning really supply the same data in many ways." In recent times , c'Ognitive semantics has led t'O substantial new insights in the motivational pathways 'Ofpolysemy. As sh'Own in Figure 4 , p'O lysemy inv'Olves a c 'Ombinati'On 'Of three basic relati 'Ons: an arbitrary semi'Otic relati 'On between a c'Ontent and a f'Orm, a m'Otivati'Onallink between tw'O ('Or m'Ore) c'Ontents , and a m'Otivated semi'Otic link 仕om 由e target c'Ontent t'O the source form. This link reflects the native speaker's intuiti'On that the target c'Ontent is conceptually sufficient1y close t'O the source content s'O由at it can be subsumed under the same (s'Ource) 岛rm. The m 'Otivati'Onal links themselves can be regarded as being, at least partially, rein岛rced by language-independent factors. SOURCE CONTENT 1
FORM 1
1: arbitrary semiotic relation 2:
TARGET 2
3
motivated concepωal relation 3: motivated semiotic relation (e.g. in my way)
Example: 切‘ containment' ， (e.g. 切 the house)..→~‘ blockage'
F盼lre 4. 咀le
motivational structure of polysemy
In this v 'Olume, Evans and Tyler's contribution is dev'Oted t'O the issue of m'Otivated, or what they call principled, polysemy. They demonstrate that the senses of 由e English prepositi'On in f'Orm a radial netw'Ork of motivated extensi'Ons (see Section 8). 5.2.4. Form -form relations
Strict1y speaking, foll 'Owing Langacker, linguistic form-岛rm relations can 'Only apply to the phon'O logical level. The relation between one phon'Ol'Ogical 岛rm and another phon'O l'O gical form may inv'O lve individual phones 'O r phonemes, for example, in regular s'Ound shifts 'Or c'O mbinati 'Ons
22
Günter Radden and Klaus-Uwe Panther
of phones or phonemes, such as in phonological processes like assimilation , vowel harmony , metathesis , consonant cluster simplification, etc. 5.2.5. Form /c ontent-form/c ontent relations
This configuration applies to motivated changes of whole linguistic units as form-meaning pairings. An example illustrating this phenomenon is the taboo-avoiding use of Gosh! or Golly! for ‘ God!' or (what the) heck! for ‘ (what the) hell!'. As shown by the bold arrow in Figure 5, the form of a unit undergoes a slight phonological change , which results in a new form. The meaning of this form is motivated by a variant of the principle of isomorphism: similarity of form corresponds to similarity of meaning: like God! , Gosh! is used as an exclamation of surprise , bewilderment, etc. Since the target form is also sufficiently dissimilar to the source form , however, it is no longer felt to offend people's religious feelings and is not an appropriate expression to be used in invocations of God. SOURCE
TARGET /
CONTENT ,
~ CONTENT2
4
+ 3
FORM ,
2
FORM ,
1: arbitrary semiotic relation; 2: motivated form-form relation; 3: motivated semiotic relation; 4: motivated content-content relation. Example: God! Gosh!
•
Figure 5. The motivational structure ofrelated forms
The reverse case of a motivated fo rmJ content-formJcontent relation is found in grammaticalization. Typically, a lexical or less grammatical content is shifted towards a more grammatical content, a process that is often accompanied by phonological attrition. For example, the demonstrative that has developed the additional grammatical sense of a complementizer (for its motivation see Hopper and Traugott 1993: 185ff.). The demonstrative is only pronounced /δæt/， whereas the complementizer also has the attrited
Introduction: 岛nn /ðe tJ.
R写flections
on motivation
23
Thus , the new target content has motivated a new phonological
岛口n.
SOURCE
TARGET
lexical CONTENT 1
grammatical 2
CONTENT2
3
FORM 1
.....1.................... 4 ....................I....~
+
FORM 2
1: arbitrary semiotic relation; 2: motivated content-content relation; 3: motivated semiotic relation; 4: motivated form-form relation (optional); Example: demonstrative that complementizer that
•
Figure 6. The motivational structure of grammaticalization
In other cases of grammaticalization, the new target fonn is noticeably distinct from the source fonn , as in be going 10 versus be gonna or one versus a(n).
6. Language-independent factors of motivation Section 5 dea1t with combinatorial possibilities of motivational relations. This section will take a closer look at some of the language-independent factors that may have an impact on the linguistic unit. As shown in Figure 7, language-independent factors may operate either on a linguistic unit as a whole or on its content or 岛口n. Motivational factors usually do not function in isolation but tend to apply jointly. All motivational factors , as they show up in language , are mediated through conceptualization. In other words , we sunnise that linguistic motivation is u 1timately guided by cognition. For analytical reasons and ease of exposition, however, we will discuss the motivational factors one by one in the following sections.
24
Günter Radden and Klaus-Uwe Panther LINGUISTIC UNIT
Experiential motivation [3] (e.g. embodirnent, image schema)
Perceptual motivation [4] (e.g. viewpoint, similarity, /sa叫
CONTENT Genetic motivation [2] (e.g. grammaticalization)
Cognitive motivation [5] (e.g. inferences , mappings , blending)
FORM
八JJ
刊酌
hu-m-m
n
白'班
·"吃肌
ρLVBJ'
EL-h 、H
l-
ole nuob Edcdbl ,
[l ]
飞\Communicative motivation [6] (e.g. maxirns , economy, expressivity) Other motivations
Figure 7. Some language-independent motivational factors
6.1.
Ecological motivation
Ecological motivation is meant to refer to the motivation of a linguistic unit due to its place, or "ecological niche" , within a system. The notions ‘ ecology' and ‘ ecological niche' have been introduced by Lakoff (1 987: 487). Taylor (this volume) elaborates these concepts within a framework of linguistic motivation. Ecological motivation is certainly not restricted to language but is much rather a general human principle. The term ‘ ecology' suggests that Lakoff and Taylor see a very strong parallel between ecological systems and linguistic systems. In this sense, the systematic aspect of language is a language-independent phenomenon. The ecology of a linguistic unit is to be understood in the sense that it has "pointers" to other units and, to the extent that the unit is related to other units in the language, it is motivated. Since each linguistic unit is related to other units within a system, all units are motivated to some extent. The reinterpretation of hamburger as a compound is motivated in being related to other compounds within the system of English , and the formmeaning pairing of spam is motivated in being related to other words of English with unpleasant connotations. As aptly described by Taylor (this volume) in his analysis of bang, this onomatopoeic word is motivated "by the cumulative effect of a network of associations pertaining to the word's phonological components."
lntroduction:
R吃flections
on motivation
25
The notion of ecology in c1udes the idea of a system that contains slots or "niches" that are filled or fillable by linguistic units. This view of language was already put forward by Meillet (1 903: 407) , who regards language (French langue) as a "système où tout se tient".14 An important consequence of the ecological view of language is that a local change is not only seen as affecting its immediate vicinity , but ultimately the system at l缸ge. A phonological example of the impact of an ecological change 扭曲e Great Vowel Shift at the end of the Middle English period. According to Akmajian, Demers and Harnish (1984: 368f.) , the long mid vowels /e:/ (as infeet) and /u:/ (as in mood) were raised to /i:/ and /u:/, respectively. As a result, the original long high 仕ont and back vowels were diphthongized to /ai/ and /au/, respectively, and the lower /æ:/ and /'J:/ were raised to /e:/ and /0:/, respectively (see Figure 8). The chain reactions 往iggered by these shifts have traditionally been described as "push and pull" effects, suggesting that a system as such exerts some linguistic pressure. In this case, the raised mid vowels may have 平ushed" the high vowels into a lower position, and the low vowels may have been "pulled" up into the vacated locations left behind by the raised mid vowels. The reason for these chain reactions might have been to restore the equilibrium of a system that seemed to be ecologically out of balance. u:
~. ",) 飞缸。 :J
al
au
Figure 8. The Great Vowel Shift
The same applies to changes affecting the semantic system. The change of meaning of an existing lexeme or the adoption of a new lexeme typically leads to a restructuring of the semantic field to which the lexeme belongs. 14. For a discussion of the origin of the quote "un système où tout se tient" , see Linguist List 14.1 954 , July 17 , 2003. It is found in Mei1let' s writings but apparently ultimately goes back to de Sausssure's lectures.
26
G现nter
Radden and Klaus-Uwe Panther
A well-known example is the meaning extension of the Old English word bryd ‘ young bird' to ‘ bird' in general at the cost of the established Old English word 卢gol， which then developed the specialized meaning ‘ fowl\ The gap left behind by the shift of meaning of bryd, 'young bird' , has partially been filled by fledgling. Two contributions to this volume demonstrate that the notion of ecology also applies to the grammatical system. Both Taylor's and Foolen's chapters are concerned with constructions that are associated with expressive meanings such as Bang goes the weekend! and a bear 01 a man. On the one hand , both constructions have pointers to other more neutral constructions such as Here comes my bus and a wheel 01 αcαr， respectively , on the other hand, they gain their expressivity by "deviating" from these constructions and occupy a niche of their own.
6.2.
Genetic motivation
The term ‘ genetic motivation' is due to Bernd Heine (this volume); it relates to diachrony. Just as present-day human behavior is the result of past motivations , present-day linguistic behavior (and one might add , the product of this behavior, language structure) is motivated by factors that were operative a long time ago but whose effects are still visible and relevant to an adequate understanding of language structure today. Among the many examples Heine (1 997) discusses we single out one case of genetic motivation. The schematic notion of comparison in the sense of ‘ comparative' has a variety of source schemas cross-linguistically (p. 112), e.g. the Action Schema, the Location Schema, the Polarity Schema, etc. Let us consider the last schema, which may be coded by two independent clauses expressing positive and negative polarity. Thus X is Y, Z is not Y has the meaning ‘ X is Y-er than Z' (p. 117). For example , in the Carib language Hixkaf)咄la， the idea ‘ Kaywerye is taller than Waraka' is literally rendered as ‘ tall-not he.is Waraka , tall he.is Kaywerye\The factors that must have been operative at the time when this construction was used in Hixkaryana in the past are still relevant today. The construction is mainly motivated by the languageindependent factor ‘ implicature': the fact that ‘ person A is tall and person B is not tall' implies that, provided the standard of comparison is the same , person A is taller than person B. The logic of conversational implicature must have been the same for Hixkaryana speakers of old times as for present-day speakers of Hixkaryana as well as for speakers of English , i.e. if
Introduction: R笔flections on motivation
someone said, Bill is not tall;
Har.η is
27
tall, we would draw the same infer-
CDCe.
In this volume , the contributions by Bemd Heine and Christian Koops are mostly devoted to genetically motivated grammatical phenomena. Heine establishes the foundations of genetic motivation , and Koops studies . a particular type of emergent grammaticalization.
6.3.
Experiential motivation
The experientialist approach to language advocated by many cognitive scholars, in particular Johnson (1 987) and Lakoff (1 987), draws attention to 也e fundamental impact of embodied meaning. The notion of embodiment is most perspicuous in the sensori-motor experiences that give rise to kin自由etic image schemas such as the container schema, the part-whole schema, etc. Image schemas relate to our earliest experiences in childhood and are assumed to be directly meaningful 一 Lakoff and Johnson even claim that they are preconcep阳al in nature. For most image schemas , it is , bowever, intuitively more plausible to assume that, only after experiencing a number of image-schematic situations , a child can form abstract imagescbematic concepts. Irrespective of whether image schemata are preconcep阳al or only preverbal , they are powerful motivating factors because of their embodied basis. As Beitel, Gibbs and Sanders (1997) have shown in their experiments on the polysemy of the spatial preposition on , the imageschematic basis of sense relations is psychologically real and motivates sense extens lOns. Parts of our early experiences are also basic events and primary scenes (see Grady 1997). Basic events are simple, goal-oriented interactions with the world, such as walking , sitting or jumping, while primary scenes relate ωthe subjective experience of basic events , such as the experience of strain or discomfort involved in lifting a heavy object. According to Lakoff (1987: 206ft), embodied experiences are not just to be understood in the sense of what happens to an individual, but in the much broader sense of "也e totality of human experience and everything that plays a role in it." They feed into our basic conceptual structure and allow us to access abstract concepts by metonymic and metaphorical projection. In this volume , two studies demonstrate the impact of experiential motivation on language structure. In their analysis of the polysemy of the particle in , Vyvyan Evans and Andrea Tyler show that its sense extensions are
28
Günter Radden and Klaus-Uwe Panther
motivated from the spatio-physical experience of the proto-scene of in , which is associated with the image schema of ‘ containment\In his study on "Motivating the uses ofbasic verbs" , John Newman shows that the morphosyntax of verbs of sitting, standing and lying in some languages is motivated by the experiential basicness ofthese states and acts.
6.4.
Perceptual motivation
Our perception of the world is inseparable from our experience and cognition , or, as expressed by the philosopher Immanuel Kant, "we see things not as they are but as we are." Sensory stimuli are meaningless and only become meaningful by associating them with something familiar , i.e. "seeing typically involves categorization" (Lakoff 1987: 126). Principles of perception allow us to filter out irrelevant information , supply information not present , and thereby structure the sensory stimulus into a meaningful gestalt. Many of the organizing principles that are pertinent in the structuring of perception also motivate language structure. Three perceptual capacities seem to be particularly relevant as motivating factors for language: attention to salience, recognition of similarity, and viewing arrangement. Attention to things that are salient shows up in many guises. In perception , salient entities are , amongst others , the figure (especially when moving) as opposed to the ground , a good gestalt as opposed to a poor gestalt , and a whole as opposed to its parts or "active zones". Especially Talmy and Langacker have shown that perceptual principles operate in language. For example, the figure/ground alignment of conceptual entities determines how they are coded in language (t he book on the table vs. ?the table under the book). Recognition of similarity is an important gestalt-perceptual principle. For example , an aπay of equidistant identical figures as in Figure 9a appears to us as unstructured , but when different-Iooking figures are added in a regular fashion , we tend to see the similar-Iooking things as belonging together and , as the circles altemating with the squares in Figure 9b , as forming colurnns and hence a regular pa忧em. The human ability of viewing different things as similar and , as a result , grouping them together is also of vital importance to language. The most important linguistic areas where recognition of similarity is relevant and may hence be said to contribute to motivating linguistic structure are , amongst others , the following: (i) categorization and generalization: con-
Introduction:
R吃flections
on motivation
29
ceiving separate things as being similar enough to be grouped together as members of the same category or abstract schema; (ii) iconicity: perceiving a simil町ity between phenomena in conceived reality and the linguistic expressions describing them.
F郁lre
口口口口
口口口口
a. b. 0000 0000 0000 0000 0000 0000
9. Similarity as gestalt-perceptual principle
The impact of viewing arrangements on language structure has been extensively studied by Langacker. Facets of the viewing arrangement include the 甲eaker's adoption ofa vantage point (e.g. taking the hearer's point ofview as in I'll come with you) , viewing frame , subjective and objective construal , 也anning， and fictive motion (see Matlock, this volume).
6.5.
Cognitive motivation
Cognitive factors that may have a motivational impact on language structure include the human ability for developing and accessing knowledge 归lctures (such as categories , frames and mental spaces) and performing cognitive operations on them (such as relating concepts , blending concepts , mappings within a domain and across domains , and drawing inferences). Cognitive factors almost always interact with other factors. For example , 因.tegorization and frame construction are to a large extent based on experience, mental spaces are tied to communicative interaction, and metaphor and metonymy are often triggered by experiential, cultural and ecological factors. In this volume , six contributions are concerned with aspects of cognitive motivation. Teenie Matlock shows how the human ability to mentally simulate motion (fictive motion) is reflected in linguistic structure. Anatol Stefanowitsch and Ada Rohde compare two possible explanations for the preference of coding goal over source. Gerhard van Huyssteen investigates
30
Günter Radden and Klaus-Uwe Panther
the motivation of reduplicated expressions in A仕ikaans. Three contribution address the impact of metonymy on grammatical structure: Ruiz de Mendoza and Díez Velasco demonstrate how the choice of anaphora is motivated by the conceptual properties of the metonymic antecedent; BrdarSzabó and Brdar explore the role of metonymy in explaining crosslinguistic differences in predicative struct1,lres; Barcelona investigates the grammatical consequences of metonymically motivated shifts of proper names to common nouns.
6.6.
Communicative motivation: Economy of coding
Since one of the main pu甲oses of language is communication, it comes as no surprise that there is a strong 由ive for making communicative acts as economical and perspicuous as possible. In formulating the framework of his Cognitive F oundations 01 Grammar, Heine (1 997: 3) describes the first assumption about language structure as follows: "The main function of language is to convey meaning. The question why language is used and structured the way it is must therefore be answered first and foremost with reference to this function." The content of the message must hence be presented with c1 arity and, at the same time , it should be coded in such a way that the hearer can interpret it with minimal processing effort (economic motivation). The requirement of c1 arity has a long history in the prescriptive and descriptive traditions of language studies. Suffice it here to mention Grice's (1975) maxims of manner such as Be perspicuous and Avoid ambiguity and vagueness. As to economic motivation , we have already seen an example of coding motivated by economy in Section 2: compounds name (almost by necessity) only certain aspects of the complex ICM they evoke. More generally , it has long been recognized that there is a universal tendency in language to code conceptual content in an economical way. George Zipf (1949) discovered an inverse relationship between the token 仕equency of a word and its length: the more frequent a word , the shorter its phonetic form. Even in a language like German, which has many more polysyllabic words than monosyllabic words , almost 50% of the word tokens actually used in texts are monosyllables (Crystal 1997: 87). Speakers' tendency to choose monosyllabic words rather than polysyllabic words reflects the principle of least effort in phonological coding, which is motivated by the speaker's wish for economy in speech. Pustet (2004: 2) points out that Zipf discovered the
Introduction: R吃flections on motivation
31
same kind of statistical correlation as in language in various other domains such as demographics and economics. The observed statistical regularity is 也us by no means restricted to language but is a "language-independent factor" in the sense discussed in Section 1. On the semantic-pragmatic level economy principles have been pro- . posed by neo-Gricean scholars (Horn 1989, Levinson 2000). Horn and Levinson develop the idea that (generalized) conversational implicatures serve to minimize coding effort: a meaning does not have to be coded as a separate lexical item if it. can be inferred on the basis of general pragmatic principles. Thus , there is no need for a word *nall ‘ not all' because it is pragmatically inferable from some. However, the logical contrary of αII is coded as a separate word, i.e. none, because its meaning ‘ all [...] no t' is not derivable by implicature. Similarly, the deontic expression permitted (= deontic may) implicates the default meaning ‘ permitted [...] not', i.e. if a person is permitted to do something, then s/he is also , by default, permitted not do it; thus no separate coding ofthis idea is necessary. Economy has also been recognized as guiding communicative principle in relevance the。可. According to Sperber and Wilson (1 995) , linguistic ∞mmunication is geared towards achieving maximal contex阳al effects with a minimum of processing effort. An utterance is relevant to the extent 由at its contextual effects in the given context are large and the cognitive effort needed to process the message is small.
6.7.
Other motivations
A full-fledged theory of motivation would, of course , have to distinguish many more language-independent factors of motivation than the six discussed above. These would, amongst others, include cultural, social , psychological and anthropological factors as well as biological and neurological determinants , which, however, are not yet sufficiently known. A good example of the impact of culture on language is the different ∞iflcep阳alization of anger in English and Chinese. Yu (1 998: Chapter 3) has shown that English and Chinese share the metaphor ANGER IS HEAT and one of its subtypes, ANGER IS FIRE , but they differ with respect to the other subtype: English has the metaphor ANGER IS A HOT FLUID IN A CONTAINER as in You make my blood bo i!, where Chinese uses the metaphor ANGER IS HOT GAS IN A CONTAINER as in ‘ He's ballooned with gas (i.e. inflated with anger)' (p. 55). Yu argues that the GAS metaphor is motivated by the philo-
32
Günter Radden and Klaus-Uwe Panther
sophical theory of yin-yang , in which things in the universe are categorized by "the law of the unity of opposites" (p. 72f.). In this the。可， fluids (y in) are categorized with cold, while gas (y ang) is categorized with heat because heat is understood as a necessaη， condition for the occurrence of gas. The GAS metaphor is therefore motivated in Chinese culture , while the HOT FLUID metaphor, though understood by Chinese-speaking people , is not made use of. The HOT FLUID metaphor is , as shown by Kövecses (1 995) , very widespread cross-linguistically; however, it may not be exclusively motivated physiologically, i.e. by experiential motivation , but as argued by Geeraerts and Grondelaers (1 995) it may be a legacy of the medieval the。可 of the four humors , i.e. be an instance of cultural motivation. The humoral interpretation of our emotional vocabulary would account for the fact that only liquids , but not solids , serve as the source domain of the ANGER IS HEAT metaphor. The role of socio-psychological factors in triggering language change was first observed by William Labov in 1962. In his seminal study social motivation of a sound change旷"， Labov (1 972) showed that the local fishermen of the island Martha's Vineyard emotionally reacted against the influx of vacationers from the mainland by reintroducing a feature of pronunciation typical of their traditional speech: a centralized pronunciation of the vowel /a/ in the diphthongs /ay/ as in light and /aw/ as in town. This pronunciation had a positive connotation and became a marker of identity and loyalty to the island. In a follow-up study carried out by Bl ake and Josey (2003) "40 years after Labov" , the social situation on the island has changed: tourism is no longer seen as threatening but as sustaining the local communities. As a result , /ay/ centralization has lost its social meaning and is replaced by decentralization. 15 Another 臼ctor that is of vital importance to language is the probably anthropological universal that humans rank higher than non-humans. This preference for humans is reflected in many domains of linguistic structure. For example , in English human participants make a better reference point than non-human participants in possessive constructions as in the president's dog versus *the dog's president (see Langacker's 2000 and Taylor's 1996 work on possessives). 15. We should note in passing that, in general , sociolinguistic correlations between social status and linguistic variables do not , as such , indicate a causal , and hence motivated, relation.
Introduction:
R拼éctions
on motivation
33
7. Multiple and competing motivations Motivation is a multifactorial phenomenon. In a given case, several factors may either jointly motivate a linguistic unit or, more commonly, compete with each other. In their analysis of binomial 丘eezes， Cooper and Ross . (1 975) identify several semantic and phonological cons仕aints， i.e. motivating factors , goveming the ordering of conjuncts. In the freeze bow and arrow , semantic and phonological factors jointly motivate this order: the semantic factor ‘ power source' motivates the position of bow as the first element, and the phonological factor ‘ more syllables' motivates the position of arrow as the second element. The order in the equivalent German expression, Pfeil und Bogen , is reversed. It is in accordance with the phonological constraint (the monosyllabic element Pfeil ‘ aηow' precedes the more "weighty" two-syllabic element Bogen ‘ bow') , but it competes with the semantic constraint ‘ power source to be mentioned first' and overrides it. Also Cooper and Ross's example Trick or treat (p. 72) illustrates competing motivations and phonological constraints overriding semantic constraints. The freeze is in con岛口nity with the phonological constraint 也at the second element, /tri: tI, should contain a more resonant nucleus than 也e first element /tri k/, but is in conflict with the semantic ordering principle A or B ('If not A , B will occur') as in Hands up, or I'll shoot. The resolution of competing motivations results in an element of arbitrariness in this case, the ordering principle A or B no longer applies without exceptI ons. Some motivational principles compete with each other by their very nature. A prime example is the conflict between economic and isomorphic motivation discussed by Croft (1990: 192匠， who follows Haiman , 1985).16 With respect to the lexicon , the principle of economy motivates a minimal vocabulary, while the principle of isomorphism requires a distinct word for every distinct concept. One resolution to the economy-isomorphism conflict is polysemy, where the principle of economy predominates. An other resolution is the use of different forms, i.e. a predominance of isomorphism. A third possibility, namely the use of more than one form for a given concept, i.e. synonymy, is both un-economic and "un-isomorphic" and, therefore , unmotivated and not likely to be found in naturallanguage. At a more general pragmatic level , there is a constant need to resolve competing 16. Croft and Haiman refer to what we caIl isomorphism as iconicity.
34
Günter Radden and }.乙laus-Uwe Panther
metonymy as a conflict-resolving device as suggested by Langacker (2000: 199) in his paper on reference-point constructions. He observes: Metonymy allows an efficient reconciliation of two conflicting factors: the need to be accurate , i.e. of being sure that the addressee's attention is directed to the target; and our natural inclination to think and talk explicitly about those entities that have the greatest cognitive salience for us. We might add economy as a third factor. As opposed to the explicit expression The water in the kettle is boiling, the metonymic expression The kettle is boiling is clear enough for directing the hearer's attention to the intended target , construes the salient container as the figure entity , and is considerably shorter than the explicit version.
8. The articles in this volume The twelve articles collected in this volume can be grouped into four of the types of motivation discussed in Section 7: ecological motivation , genetic motivation , experiential motivation and cognitive motivation. It should be kept in mind, however, that the issues presented in the papers often also touch upon other types of motivation , i.e. , the allocation of the papers to a specific category is to a certain degree "arbitrary".
8.1.
Ecological motivation
The basic units of grammar are syntactic constructions. According to John Taylor, three kinds of constructions can be identified in Cognitive Grammar: phonological constructions , semantic constructions, and symbolic constructions. Only the last type is usually considered a construction. In his paper "The ecology of constructions", Taylor argues that a linguistic structure may be motivated with respect to all three kinds of construction. In his understanding , a construction is motivated to the extent that it is related to other units in the language. The structure he uses to illustrate this point is the constructional idiom bang goes as in Bang goes my weekend. As a phonological construction, the sound shape of the ideophone [bæIJ] is motivated within a network of associations with words sharing one or more of these sounds such as slam and is therefore felt to be an appropriate forr口 to designate the sound of a sudden impact. As a semantic construction , the
Introduction:
R吃flections
on motivation
35
bang goes construction conjures up a specific scenario in which a person feels mild annoyance at having things interfere with their plans. As a symbolic construction, the bang goes construction instantiates the schematic subject-final construction [X V NP Subj ] and constructions with go such as Bang went the balloon, when it burst. Constructions do not exist in . isolation but occupy an ecological niche within a network of relations. In 也is sense , any construction is motivated to some extent. In his chapter "Expressive binominal NPs in Germanic and Romance languages", Ad Foolen analyzes constructions such as an angel of a child, which exist in various Germanic and Romance languages. The construction consists of two noun phrases linked by the preposition of in English and similar "neutral" prepositions in the other languages considered. The binominal cons位uction as in She is an angel of a child is emotionally "marked": it conveys a strong expressive force. 17 Foolen discusses the problem of which of the two NPs is to be considered the head of the construction. At first sight, the construction seems to confirm the hypothesis that the relation between linguistic form and content is arbitrary; by extension , it seems to support a modular view of language, i. e. the autonomy of syntax from semantics , because the conceptual modifier, i.e. the metaphorized prope时y of being an angel, shows up as the syntactic head of the noun phrase, whereas 也e concep阳al head , i.e. child, is syntacticized in a subordinate prepositional phrase. However, on closer inspection, the expressive binominal noun phrase construction turns out to be highly motivated: the direction of the motivational relation proceeds 仕om concep阳al content to syntactic 岛rm (or alternatively, schematic grammatical meaning). The metaphoric property ‘ angel (like) , is foregrounded and consequently given the most prominent position (head) in the grammatical construction. Foolen calls this position the "expressive" head of the construction. In contrast, the conceptual head child, which assumes the function of "referential head", is syntactically relegated to a non-prominent position although it is topical in sentences like An angel of a child entered the room , which is referentially clearly about a child, not about an angel. Foolen concludes that the doubleheadedness of the construction is motivated by its double function: the first NP is the expressive head of the c
17. See also Taylor (1 996: 329f.), who also notes the "distinctive affective overtones" (329) of what he calls ‘ appositive q严constructions " since the two noun phrases refer to the same entity.
36
8.2.
Günter Radden and Klaus-Uwe Panther
Genetic m'O tivati 'On
In his c'Ontributi 'O n "On genetic m'O tivati 'O n in grammar", Bernd Heine c'Ontrasts tw'O 'Opp'O sing views 'Of motivation: ‘ structural m'O tivation' and ‘ genetic motivation'. Structural motivati 'On pertains t'O the relationship between structures within a given language or between linguistic and extralinguistic structures. Genetic motivation refers to language structure as a product of human behavior: it accounts for the motivating forces underlying linguistic change. In this view, language structure is basically motivated and arbitrariness constitutes the exception and is in need of explanation. By way of illustrati'O n of this type of motivation, Heine l'Ooks at the cognitive forces that are responsible for the emergence of three grammatical categories: numerals , indefinite articles and possession. The numeral systems found in the languages of the w'Orld have ev'Olved 仕om our experience of body parts and a set 'O f basic arithmetical operati'Ons. Indefinite articles typically derive from the numeral ‘ one' , which is still reflected in their predominant use with singular count n'Ouns. The concept ‘ possession' derives from different source schemas in different languages and may, in turn , give rise to different non-possessive meanings such as tense and aspect. In combining hist'Orical evidence with cr'O ss-linguistic generalizations, the search for genetic motivation pr'O vides a new promising approach to the study of grammatical structure. Christian Koops' chapter "Emergent aspect constructions in PresentDay English" can also be regarded as a contribution to the t'Opic 'O f genetic m'O tivati 'O n. It is concerned with the emergence of grammatical structures at the incipient stage, c'Oncentrating on constructi'O ns in Present-Day English (PDE) that convey progressive meaning. Studies on grammaticalization have sh'O wn that the progressive aspect tends to 'Originate 仕'Om a restricted set of s'Ource concepts, in particular ‘ l'O cati'O n' , ‘ posture' and ‘ motion\ Paths of grammaticalization attested across languages are probably wellmotivated and may be assumed to 'O ccur in any given language. As Koops convincingly demonstrates , Present-Day English als 'O has aspectual c'O nstructions emerging from these source n'O tions. The source n'O ti 'On ‘ locati'On' is found in the PDE aspectual locative c'O nstructi 'O n with in the middle/midst 01 as in 1 was in the middle 01 getting my hair cut. The source notion 'pos阳re' underlies
lntroduction: R笔flections on motivation
37
especially with the basic verb go (around) as in You can 't go around testing everybo命 for everything. The emergent aspec阳al senses associated with these three constructions are subtypes of the imperfective aspec t: specifically, they are progressive, durative or repetitive in meaning. The aspectual meaning characterizing each of the three constructions as well as . additional aspects of meaning and grammatical constraints are shown to 岛llow from the lexical meanings of their source notions.
8.3.
Experiential motivation
Two chapters are devoted to aspects of experiential motivation: the first investigates motivation in lexical structure, the second the role ofbasic verbs in grammaticalization. The lexicon , in contrast to grammar, has traditionally been viewed as the idiosyncratic and arbitrary component of language and , therefore , as lacking structure and systematicity. This position is challenged by Vyvyan Evans and Andrea Tyler, who argue that lexical structure, too , is highly organized and conceptually motivated. The motivation of lexical structure can most clearly be seen in the meaning extensions of polysemous lexemes. In their paper "Spatial experience, lexical structure and motivation: The case of in" , Evans and Tyler illustrate their approach by way of analyzing the "principled polysemy" of the English particle in. The central meaning , or "sanctioning sense", of a given preposition is a highly abstract representation of a spatial configuration of two objects, which Evans and Tyler call 'proto-scene\In the case of in , the proto-scene is characterized by the complex spatio-functional relation of ‘ containment'. Our experience and interaction with aspects of containment has consequences for certain associations, which, via pragmatic strengthening, motivate meaning extensions of in. Thus , as Evans and Tyler argue , our common experience of a bounded location and being in a state of restricted freedom give rise to a conventionalized State Sense as in We 're in a state of war. Bounded landmarks may also be associated with specific activities , thus giving rise to an Activity Sense of in as in She 泣的 medicine. The fifteen senses of in distinguished by Evans and Tyler form a radial network of interrelated senses. What makes this cognitive-linguistic study particularly relevant, however, is the insight that each of the extended senses is motivated from our spatiophysical experience with different aspects of containment. The experiential ∞ntribution
38
Günter Radden and Klaus-Uwe Panther
approach adopted here can most certainly also be fruitfully applied to other instances of lexical polysemy. John Newman investigates the impact that our bodily experience, in conjunction with the grammatical patteming of a given language, has on linguistic structure. In his paper "Motivating the uses of basic verbs: Linguistic and extralinguistic considerations", Newman looks at selected "basic" verbs in various languages. Basic verbs express ordinary, body-based experiences and play a fundamental role in language: in particular, they are subject to figurative extensions and adopt grammatical functions such as serialization, tense and aspect markers , noun c1 assifiers , and case marking. Basic verbs typically appear in pairs such as ‘ come' and ‘ go' ,‘ eat' and ‘ drink' , and ‘ give' and ‘ take'. These verb pairs reflect natural groupings of states or acts that we experience as complement町y. In contrast,‘ at rest' positions are typically experienced in three ways: standing, si忧ing and lying. As a resuIt, pos阳re is expressed by the three basic verbs ‘ stand' ,‘ sit' , and ‘ lie\In their grammaticalized forms , the three pos阳re verbs may express a gradation of control 仕om standing through sitting to lying. Some morphosyntactic facts appear to be naturally correlated with this gradation. For example, in the Oceanic language Manam,‘ stand' is aligned with verbs of action and agents , while ‘ sit' and ‘ lie' are aligned with state verbs without agents.
8.4.
Cognitive motivation
The papers that we have subsumed under the heading ‘ cognitive motivation' are very heterogeneous. Moreover, all of the contributions presented in the preceding subsections discuss cognitive factors of motivation as wel l. The concept of motion and its linguistic coding have been extensively studied in cognitive linguistics and are the subject of two contributions in this volume. In her contribution "The conceptual motivation of fictive motion" , Teenie Matlock provides compelling linguistic and psychological evidence to show that fictive motion constructions (FM-constructions) are motivated by our cognitive ability to mentally simulate motion along a path. In FMconstructions, stationary scenes are construed as motional as in The road runs along the coast. The motion described in FM-constructions is purely subjective and consists in the speaker's mental scanning of the trajectory. The FM-construction displays linguistic behavior that naturally follows
Introduction: R吃flections on motivation
39
this concep阳al basis. For example, since scanning requires 也at the object is not fully visible with just one look, the 衍ajectory described by the subject of the FM-construction needs to be long and large, and 也e time of scanning as expressed by temporal adverbials should not be too short. The results of experiments measuring the decision times in understanding FM- . constructions suggest that people construct a dynamic representation of motion and process fictive motion differently depending on the type of motion verb. Also , experiments using drawings provide evidence for mentally simulated motion in processing FM-constructions. Matlock finally points out 由at FM-sentences require a contex阳al situation in which people 硝emptωestablish or maintain common ground. This communicative aspect may also contribute to motivating 由is particular structure. The chapter by Anatol Stefanowitsch and Ada Rohde is concemed with the "The goal bias in the encoding of motion events". The paper assesses two explanations that have been proposed to account for the asymme町 in the coding of the goal vs. the traject。可 and the source of 由e moving entity, i.e. for the fact 也at 由e dis衍ibution of goal-PPs is less restricted 由an that of other path-PPs. One explanation, referred to as the (psychological) salience hypothesis, assumes that the asymme位y is part of O町 conceptual system and motivated by 0町 greater interest in the goal of 缸tions 由an in their source. Another explanation, referred to as completeconceptualization hypothesis, assumes that the goal bias is motivated by 也e higher information value conveyed by goal-PPs: unlike PPs coding the initial and medial portions of a pa白， goal-PPs allow for an inferential conC叩阳a1ization of the complete tr哥 ecωry. In order to test these hypo由eses， 也.e authors conduct a corpus-based study of 由e combinations of selected motion verbs with path-PPs. Frequency analyses confirm the general bias ωIwards the goal , but they also reveal the existence of "exc叩tions": 由us， 也e motion verbs cruise and stro /l mainly occur with trajectory PPs and the motion verb escape mainly occurs with source PPs. While both explana1ions are not mutually exclusive as motivating factors , the completeωInceptualization hypothesis accounts for both the general goal bi副部 well 槌岛rexc叩tional cases. Reduplication is usually considered an example of iconic motivation p町 缸.cellence. In his paper "Motivating 也e composition of A企ikaans r 齿。m
,
40
Günter Radden and Klaus-Uwe Panther
profile of the reduplicated composite structure diverges slightly from those of the component structures. Thus , the Afrikaans time adverbial nou ‘ now' prototypically profiles present time and may , within certain contexts , also be used to profile the near future or the recent past. The reduplicated form nou-nou , on the other hand , profiles either the future or the past but not the present. The meaning of ‘ distancing from the present' may be captured by the metonymy MORE OF FORM FOR MORE OF CONTENT. Another form of profile shift discussed by van Huyssteen involves the recategorization of a perfective verb as an imperfective verb. The reduplicated past tense verb may take 阳'0 forms: complete reduplication as in geva• gevat ‘ touchedtouched' or partial reduplication as in gevat-vat ‘ touched-touch\The former structure conveys the meaning of ‘ repetition' , the latter also that of ‘ attenuation\Van Huyssteen convincingly shows how these differences in meaning can be accounted for by their different compositional structures. Onomatopoeic reduplications differ from grammatical reduplications in function and structure. Their function is referential as in the bird's c可 hoep-hoep , which metonymically stands for the bird Upupa Africana; and their structure is characterized by mutual elaboration of their elaboration sltes. The three following contributions focus on metonymy as a motivating force. In their chapter "Metonymic motivation in anaphoric reference" , Francisco Ruiz de Mendoza and Olga Díez Velasco address the thomy problem of what determines the choice of anaphoric pronouns that have metonymic antecedents. Their contribution is an excellent example of cognitive motivation of grammatical form , i.e. content-form motivation , in this case , the form of anaphoric pronouns. Ruiz de Mendoza and Díez Velasco provide a solution to the coreference problem on the basis of their theory of metonymic mapping and three general cognitive principles that guide the selection of the pronoun. In their view, metonymies are either concep阳al expansions of a given source domain into a conceptually more elaborate target domain (source-in-target metonymy) , or they involve a process of conceptual reduction of a source domain that leads to a more restricted target domain (target-in-source domain). In any case, the selection of the appropriate anaphoric pronoun is determined by what the authors call the "matri
Introduction: R写flections on motivation
41
Availability Principle", which is at work in the above example, the authors propose two additional principles: the "Domain Combinability Principle", which takes the concep阳al content of the verb phrase into account of which the referent of the pronoun is predicated, and the "Domain Precedence Principle", which accounts for cases that involve "double" metonymies . with two possible matrix domains. In their contribution "Predicative adjectives and grammatical-relational polysemy: The role of metonymic processes in motivating cross-linguistic differences" , Rita Szabó-Brdar and Mario Brdar provide a further example of content-form motivation. They argue that certain grammatical facts ofEnglish follow 企om the operation of metonymic processes, in contrast to Croatian, German, and Hungarian, where these cognitive processes are restricted or virtually absent. The authors consider a variety of English constructions, such as the "Setting (Locative) Subject Construction" (e.g. London wasfoggy today) , the "Experiencer Subject + Predicative Adjective Construction" (e.g. 1 am hot) , the "Be + A司jective Possessive Construction" (e.g. 1 was firm ofpurpose) , "Be + Adjective Manner Constructions" (e.g. One should be as clear as possible about historical facts) , and "Raising Constructions" (e.g. The editor is certain to reject it). They then compare these constructions with their semantic equivalents in Croatian, German, and Hungarian. The authors' method of analysis yields interesting insights into the four languages studied: English tends to keep the a司jectival conS往uction formally constant, relying on metonymic processes such as COMPLEX ENTITY FOR ITS LOCATION, PERMANENT STATE FOR TEMPORARY STATE , POSSESSOR FOR POSSESSED , MANNER FOR ACTIVITY , SALIENT PARTICIPANT FOR WHOLE EVENT , etc. , to rearrange argument structure. By contrast, in Croatian, German and Hungarian these metonymies are either restricted or impossible so that these languages have to resort to the strategy of overtly rearranging argument structure by selecting formally different, i.e. non-adjectival , predicative expressions. Antonio Barcelona's paper "Metonymy behind grammar: The motivation of the seemingly ‘ irregular' grammatical behavior of English paragon names" is concerned with the motivation for the use of proper names as common nouns. Proper names are characterized by a number of morphoS泸lt
42
Günter Rαdden and Klaus-Uwe Panther
as a paragon is evoked , which, by metonymy , yields a characteristic property. In the case of Shakespeare, this characteristic prope时y might be described as "writer endowed with immense literary talent". Secondly, the metonymy IDEAL MEMBER FOR THE CLASS maps the stereotypical model and its ideal property onto a whole class. Thus , the paragon Shakespeare stands for the class of writers that have an immense literary talen t. As a result , Shakespeare becomes a class name and is coded as a common noun.
9. Conclusion We as well as probably most ofthe authors in this volume regard the search for motivational explanations as a useful heuristic for linguistic research. At this stage , it is impossible to provide conclusive evidence for or against the hypothesis that all of language is motivated by language-independent factors. Thus we do not claim that every linguistic phenomenon is motivated. What we hope to have shown in this introduction is that for many linguistic phenomena motivational accounts suggest themselves very strongly - not in the sense of nomological-deductive explanations in the "hard" sciences, but more in the spirit of what the Gennan philosopher Wilhelm Dilthey characterized as ‘ understanding' (verstehen) in the humanities or cultural sciences (Geisteswissenschaften). The contributions to this volume surely testify to the fruitfulness of a motivational approach to language.
References Akmajian , Adrian , Ri chard A. Derners and Robert M. Hamish 1984 Linguistics: An lntroduction to Language and Communication. Carnbridge , MA/L ondon: MIT Press. Beitel, Dinara A., Rayrnond W. Gibbs and Paul Sanders 1997 The ernbodied approach to the polyserny of the spatial preposition on. In: Hubert Cuyckens and Britta Zawada (eds.) , Polysemy in Cognitive Linguistics , 242-260. ArnsterdarnlP hiladelphia: Benjarnins. Berg, Thornas 1998 Linguistic Structure and Change: An Explanation from Language Processing. Oxford: Oxford University Press.
Introduction:
R电flections
on motivation
Blake, Renée and Meredith Josey 2003ηle /ay/ diphthong in a Martha's Vineyard community: What can we say 40 years after Labov? Language in Society 32: 451 -485. Bolinger, Dwight 1968 Entailment and the meaning ofstructures. Glossa 2: 119-126. Claudi, Ulrike and Bemd Heine 1986 On the metaphorical base of grammar. Studies in Language 10: 297-335. Cooper, William E. and John R. Ross 1975 World order. In: R. E. Grossman et al. (eds.) , Papers from the Parasession on Functionalism , 63-109. Chicago: Chicago Linguistic Society. Croft, William 1990 Typology and Universals. Cambridge: Cambridge University Press. 2000 Explaining Language Change: An Evolutionary Approach. Harlow, Essex: Longman. Crystal, David 1997 The Cambridge Encyc/opedia of Language. Cambridge: Cambridge University Press. Cuyckens, Hubert, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds.) 2003 Motivation in Language: Studies in Honor of Günter Radden. Amsterdam/Philadelphia: Benjamins. Geerae邸，Dirk
The interaction of metaphor and meton严ny in composite expressions. In: René Dirven and Ralf PÖ由19S (eds.) , Metaphor and Metonymy in Comparison and Contrast, 435-465. BerlinINew York: Mouton de Gruyter. Geeraerts , Dirk and Stefan Grondelaers 1995 Looking back at anger: Cultural traditions and metaphorical pattems. In: John R. Taylor and Robert E. MacLaury (ed时， Language and the Cognitive Construal ofthe World, 153一 179. BerlinlNew York: Mouton de Gruyter. Goldberg, Adele E. 1995 A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Grady, Joseph 1997 Foundations of Meaning: Primary Metaphors and Primary Scenes. Berkeley, CA: U.C. Berkeley dissertation. 2002
43
44
Günter Radden and Klaus-Uwe Panther
Grice , H. Paul 1975 Logic and conversation. In: Peter Cole and Jerry L.岛10rgan (eds.)，年eech Acts (Syntax and Semantics 匀， 41-58. New York: Academic Press. Haiman, John 1980 The iconicity of grammar: Isomorphism and motivation. Language 56: 51 4-540. 1985 Natural 砂ntax: Iconicity and Erosion. Cambridge: Cambridge University Press. Hawkins, John A. 1994 A Pe价rmance Theory of Order and Constituency. Cambridge: Cambridge University Press. Heine, Bemd 1997 Cognitive Foundations ofGrammar. New YorklOxford: Oxford University Press. Heine , Bemd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Hiraga, Masako K. 1994 Diagrams and metaphors: Iconic aspects in language. Journal of Pragmatics 22: 5-21. Hoeks恼， Teun and Jan G. Kooij 1988 The Innateness Hypothesis. In: John A. Hawkins (ed.), Explaining Language Universals, 31-55. Oxford: Blackwell. Hopper, Paul J. and Elizabeth Closs Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Hom , Lawrence R. 1989 A Natural History of Negation. Chicago: University of Chicago Press. Hui忧， W.
2001
Motivation. Educational Psychology Interactive: Bill Huitt's Home Page. h句://view .showstat. com/showview .asp?user =motlvate Kövecses, Zoltán 1995 Anger: Its language, conceptualization, and physiology in the light of cross-cultural evidence. In: John R. Taylor and Robert E. MacLau可 (eds.)， Language and the Cognitive Construal of the World, 181-196. BerlinINew York: Mouton de Gruyter. Labov, William 1972 The social motivation of a sound change. In: William Labov , Sociolinguistic Patterns , 1-4 2. Philadelphia: University of Pennsylvania Press.
lntroduction: R吃flections on motivation Lakoff, George
1987
Women , Fire, and Dangerous Things: Wh at Categories Reveal about the Mind. Chicago/L ondon: University of Chicago Press.
Langacker, Ronald w.
2000
Grammar and Conceptualization. Berlin/New York: Mouton de Gru严er.
Levinson , Stephan S.
2000
Presumptive Meanings: The Th eory 01 Generalized Conversationallmplicature. Cambridge, MA: M. I. T. Press.
Meillet, Antoine
1903
lntroduction à l'étude comparative européennes. Paris: Hachette.
des
1αngues
indo-
Newman, John
1996
Give: A Cognitive Linguistic Study. Berlin/New York: Mouton
de Gruyter. Newmeyer, Frederick J.
2000
Language Form and Language Function. Cambridge , MAlLon-
don: MIT Press. Onions , C.T. (ed.)
1966
The
0;矿òrd Dictionary
01 English Etymology. Oxford: Clarendon
Press. Peirce , Charles S.
1955
Philosophical Writings 01 Peirce. Selected and Edited with an lntroduction by Justus Buchler. New York: Dover Publications.
Popper, Karl R.
1963
Conjectures and Rφltations: The Growth edge. London: Routledge and Kegan Paul.
01 Scientifìc
Knowl-
Pustet, Regina 2004 Zipfand his heirs. Language Sciences 26: 1-25. Saussure , F erdinand de 1916/59 Course in General Linguistics. New YorkIToronto/L ondon: McGraw-Hill. Shisler, Benjamin K. n.d. Dictionary 01 English Phonesthemes. www.geocities.com/ So忱。/ Studios/9783年hondl.html
Sperber, Dan and Deirdre Wilson
1995
Relevance: Communication and Cognition. Oxford: Blackwell.
Stegmüller, W olfgang
1969
Probleme und Resultate der Wissenschaflstheorie und Ana护 tischen Philosophie. Vo l. 1: Wissenschaflliche Erklärung und Begründung. Studienausgabe , Teil 3: Historische und rationale
45
46
Günter Radden and Klaus-Uwe Panther
Erklärung. Kausalitätsprobleme, Determinismus und lndeterminismus. Heidelberg/N ew York: Springer. Sweetser, Eve E. 1990 From Etymology to Pragmatics: Metaphorical and Cu /t ural Aspects 01 Semantic Structure. Cambridge: Cambridge University Press. Taylor, John R. 1995 Linguistic Categorization. Oxford: Oxford University Press. 1996 Possessives in English: An Exploration in Cognitive Grammar. Oxford: Clarendon Press. 2002 Cognitive Grammar. Oxford: Oxford University Press. U Ilmann , Stephen 1972 Semantics: An lntroduction to the Science 01 Meaning. Oxford: Basil Blackwell. Verhagen , Arie in press Construction 01 lntersubjectivity. Oxford: Oxford University Press. Waugh , Linda R. 1994 Degrees of iconicity in the lexicon. Journal 01 Pragmatics 22: 55-70. Yu , Ning 1998 The Contemporary Theory 01 Metaphor: A Perspective Irom Chinese. AmsterdamJP hiladelphia: Benjamins. Zipf, George 1935 The Psychobiology 01 Language: An lntroduction to Dynamic Philology. Cambridge , MA: M. I. T. Press.
Section 1 Ecological motivation
The ecology of constructions John R. Taylor
Abstract According to Cognitive Grammar, the grammar of a language can be characterized as a structured inventory of conventional linguistic units. The inventory is said to be structured in the sense that 由e units do not consti阳te encapsulated chunks of information; on 也econ位ary， each unit stands at 也e hub of a network of relations to other units. Three kinds of relations are discussed: the relation between a schema and its more fully specified instances; the relation between a p町t and the whole in which it features; and relations of similarity. The focus in this chapter is on constructions, defined 部 linguistic structures which are analysable into component parts. A conS位uction may be said to be motivated to 也e extent 也at it, and its pa邸， bear multiple relations to other units in the language. It is these relations which, cumulatively, create a ‘ niche' for the construction within the language system. Motivation is illustrated on the ex缸nple of a constructional idiom in English: the bang goes construction. Keywords: construction, entrenchment, ideophone , idiomaticity, productivity, schema-instance relation , syntagmatic relation
1. Introduction My aim in this chapter is twofold. First, 1 discuss , in general terms , the nature of constructions and what it might mean to say that a construction is motivated. Subsequent1y, 1 study the motivation of a ‘ constructional idiom' in English.
50
John R. Taylor
The chapter presupposes Langack町 's model of Cognitive Grammar. Two aspects of the model are of particular relevance to the discussion. The first concems the status of language as a symbolic system, that is , as a means for relating forms and meanings. Consistent with the symbolic nature of language , only three kinds of entity are recognized in Cognitive Grammar: phonological structures (these pertain to language in its perceptible form 1) , semantic structures (that is , meanings , or conceptualizations) , and symbolic structures , that is , associations of phonological and semantic structures (Langacker 1987: 76). While the study of symbolic structures subsumes the study of phonological and semantic structures, a certain degree of autonomy is nevertheless attributed to both phonology and semantics. Phonology may be organized in terms of its own specific principles , consequently, not every element of a phonological structure need participate in a symbolic relation with an element of a semantic structure. Similarly , not every element of a semantic structure has to be directly symbolized by some component of phonological structure. The second aspect of Cognitive Grammar that needs to be highlighted concems the content of linguistic knowledge. It is proposed that knowledge of a language (that is , a ‘ grammar') consists in a person's grasp of an inventory of established linguistic conventions; in accordance with the symbolic nature of language , these pertain to phonological , semantic, and symbolic structures. Established conventions, which speakers have thoroughly mastered , are referred to as ‘ units\Importantly , the inventory is claimed to be structured (Langacker 1987: 57) , in the sense that each unit participates in relations of various kinds to many other units. ‘ Doing' Cognitive Grammar consists , ve可 largely， in elucidating these relations. 1 claim that it is these relations which , cumulatively, motivate a linguistic structure , in that they create a ‘ niche' for the struc阳re within the larger language system.
2. Constructions 1 propose the fo l1 owing definition of a construction:
1.‘ Phonology' , therefore , needs to be understood rather broadly , to incorporate orthographic representations and (in the case of sign languages) signing , as well as , for spoken language , gestural accompaniments of speech.
Th e ecology 01 constructions
(1)
51
A construction is a linguistic struc阳re that is intemally complex , that is , a structure that can be analyzed into component parts. 2
Given that Cognitive Grammar recognizes three kinds of linguistic entities , we can identify three kinds of constructions in a language, phonological, semantic , and symbolic: -
一
-
phonological constructions. The phonological form [kæt] is a phonological construction, in that it can be analyzed into constituent phonological units , namely , the segments [时， [æ] and [t]; semantic constructions. The complex concept [BLACK CAT] is a semantic construction, which can be analyzed into its component concepts , [BLACK] and [CAT]; symbolic constructions. The expression black cat, understood as the association of the phonological structure [blæk kæt] and the semantic structure [BLACK CAT] , is a symbolic construction, being composed of the constituent symbolic structures [BLACK]/[blæk] and [CAT]/[kæt].
As mentioned in the Introduction, an important topic in Cognitive Grammar is the elucidation of the relations that exist amongst the structures which comprise a grammar. Two kinds of relation are especially important: 一‘ V ertical'
(or categorizing) relations between a more schematically characterized structure and its more fully specified instances; 一‘ Horizontal' (or syntagmatic) relations between a structure and a larger structure of which it is a part. Syntagmatic relations are a defining feature of constructions, as characterized in (1).
It should be noted that ‘ vertical' and ‘ horizontal' relations are independent of each other. For example, there can be syntagmatic relations between 卸lly specified structures , or between structures which are highly schematic. Similarly, the schema-instance relation can apply both to structures that are intemally simple and to those th创 are intemally complex (i.e. ‘ cons仿uc tions' , on my use ofthe term). A third kind of relation also needs to be mentioned:
2. Constructions, on this definition , correspond to tures' (Langacker 1987: 487).
Langack町、‘ composite
struc-
52 一
John R. Taylor Two
(or more) linguistic structures , whatever their degree of schematicity or intemal complexity, may be perceived to be similar.
The similarity relation is important, because it is the perception of similarity between structures that permits the emergence of a more schematic structure, which captures the commonality between the instances. 1 now consider these kinds of relation with respect to constructions. 1 focus , initially, on the relations as they pertain to phonological constructions. 1 begin with phonological examples because it is here , 1 believe, where the properties of constructions can be illustrated most clearly. Moreover, insights from phonology can provide a model for the analysis of other kinds of construction, and can thus help to clarify the methodological and conceptual problems which arise in these areas.
2.1.
Schema-instance relations amongst constructions
The perceived similarity between phonological structures of the kind [kæt] , [pet], [SI时， and many more , allows the emergence of a schematic construction which cap阳res the commonality between the instances. The schema in question may be represented as [CVC] cr. The schema specifies the kinds of entities that are eligible to occur in the construction, namely , entities of the kind ‘ consonant' (c) and ‘ vowel' (V) , as well as the manner in which they are combined. The schema also characterizes the complex structure as a ‘ syllable' , symbolized by subscript ‘ σ\These two aspects need to be examined in more detail. Consider, first , the specification of the component entities as ‘ vowels' and ‘ consonants\To be sure, there may be a language-independent basis for this classification. Thus , vowels are relatively sonorous segments , produced with minimal constriction of the airtlow through the vocal tract , while consonants, being articulated with greater degree of constriction, are less sonorous. However, since sonority is a gradient notion , it does not of itself give rise to a clear-cut distinction between vowels and consonants. What triggers the dichotomous categories is , rather, the function of segments within a syllable. (In fact , it is only because of its relevance to syllable structure that sonority is a linguistically relevant parameter of segments in the first place.) Thus , a segment counts as a vowel or as a consonant , not only because of its inherent phonetic properties , but in virtue of the slot which it fills in the syllable schema. There is a sense , then , in which it is the
The ecology ofconstructions
53
construction which defines the categories which are able to function as its parts. Although based in general phonetic aspects , the categories of vowel and consonant emerge as functions of the roles which segments play within the schematic construction. As with vowels and consonants, there may well be a general phonetic basis for syllables, namely , as phonological structures characterized by a peak of sonority flanked by segments of lower sonority. Nevertheless, a particular sequence of sounds in a given language counts as a syllable, not only because of its inherent phonetic properties, but also because it conforms with (i.e. , it can be regarded as an instance o t) a schematically characterized syllable construction. The notion of ‘ syllable' , then , also emerges at the level of the constructional schema, not at the level of the instances. The schematic construction captures the commonalities of a vast range of specific constructions which share a similar intemal structure. It is in virtue of these commonalities that a unit of the kind ‘ syllable' can be postulated. Constructions such as [kæt] , [pεt] ， etc , count as syllables because they instantiate the construction. [CVC] is not, of course, the only syllable schema in English. English allows other kinds of syllables , e.g. [CV:] , [CVV] , [CCVCC], and so on. 3 Co-occurrence restrictions within a syllable 一岛r example, the fact that a long vowel or diphthong can be followed by , at most, one non-coronal consonant4 - point to the need to recognize sub-components within the syllable, such as the onset and the rhyme constituents, as well as sub-categories of vowels and consonants , such as short vs. non-short vowels and coronal vs. non-coronal consonants. These ‘ natural c1 asses' of segments, like the notions of vowel and consonant, emerge as properties of language-specific constructional schemas. It should also be bome in mind that a full account of English syllables will need to make reference to specific segments; [h] , 岛r example , is res位icted to occurring as a syllable onset while [1)] is restricted to occurring in a syllable rhyme (and then , only after a ‘ short' vowe1 5). What this means , is that ‘ syllable' , in English, will need to be rep3. [V:] represents a ‘ long' vowel , as in see [si:] , [VV] represents a diphthong, as inso [S;)u]. 4. Thus, fiend [丑 :nd] and sound [saund] are acceptable syl1ables , whereas *[fi:mp] and *[saI1J k] are not. Coronal consonants include [t, d, e ， δ ， n， s, Z , 1, r] , in contradistinction to the labials [p , b, f, v] and the velars [k, g, 1)]. 5.η1US，
[IDI)] is an acceptable English syllable, but *[b:I)] is not.
54
John R. Taylor
resented by a large and c'O mplex netw'Ork 'O f c'O nstructi 'O ns, c 'O mprising schemas 'O f varying degrees 'O f abstracti 'On , and making reference, n 'Ot 'O nly t'O intersecting and cross-cutting gr'Oups 'Of segments, but als'O t 'O specific s'Ounds. M 'O re'Over, as is well-kn'Own, languages differ very c 'Onsiderably with respect t'O the kinds 'O f syllables 由at they all 'O w. Wh ile the pr'Operties 'O f syllables may well be based in universal ph'Onetics , the syllable netw 'O rk f'Or English (and f'Or any 'Other language) will be largely peculiar t 'O that language. 1 have illustrated at s'Ome length the schema-instance relati 'Ons am'Ongst c'O nstructi 'Ons 'O n examples fr'Om ph'O n 'O l'Ogy. The insights gained 仕om ph'On 'O l'Ogy carry 'Over t'O semantic and symb'Olic c'O nstructi 'O ns. Let us c'O nsider, s'Omewhat m 'Ore brie f1y , the case 'O f symb 'O lic c 'Onstructi 'Ons. (‘ Symb 'O lic c'O nstructi 'Ons' c'O rresp 'O nd by and large t'O what are traditi'O nally referred t 'O as ‘ syntactic c'O nstructi'Ons\) On the basis 'O f expressi'Ons such as a cat , the dog , etc. we may prop'Ose a n 'Oun phrase schema [DET N]NP.6 The schema makes referenceωthe schematic units [DET] and [N], while the c 'Onstructi 'O n itself is characterized as a n 'Oun phrase (NP). An imp 'Ortant t'Opic in C'Ognitive Grammar has been the search f'O r a c 'O nceptual basis f'Or lexical and syntactic categ'Ories , such as n 'Ouns , determiners, and n 'Oun phrases (Langacker 1987). H 'Owever, reference t'O c'Onceptualizati'Ons al 'O ne cann'O t be a sufficient basis f'O r characterizing the lexical and syntactic categ'Ories 'O f a particular language. As was the case with syllables , it is the n 'Oun phrase schema itself which defines the categ'Ories that are eligible t 'O 'Occur in it, and which, m'O re 'O ver, justifies the categ'Orizati 'On 'O f an expressi'On as a n 'Oun phrase. What makes explosion a n 'Oun , and n 'Ot a verb , is the fact that it can c'O-'O ccur with a determiner within a n 'Oun phrase. In the final analysis , n 'Ouns are identified 'On distributi 'Onal grounds , as units that can instantiate the N-sl 'O t in a n 'Oun phrase schema. [DET N] is n 'Ot the 'Only n 'Oun phrase schema in English. As with syllables , the English n 'Oun phrase will need t'O be characterized by a dense netw 'Ork 'O f c'O nstructi 'Ons , which make reference t 'O categ'O ries such as ‘ definite 6. In its status as a symb 'O lic c'Onstructi 'On, the n'Oun phrase schema wi11 need t'O be characterized in tenns 'Of its ph'On'Ol'Ogical and semantic pr'Operties. Its ph'On'O-
l'Ogical structure sancti'Ons the c'O mbinati 'O n 'O f the ph'O n'Ol'Ogical f'Onn 'Of a detenniner with the ph'O n'O l'O gical f'Onn 'Of a n'Oun , while its semantic structure sancti'Ons the c'O mbinati 'On 'Of the semantic structure 'O f a detenniner with the semantic structure 'Of n'Oun, resulting in the pro fi1 ing 'Of a grounded instance 'O f the n'O minal entity.
1页e
ecology 01 constructions
55
detenniner' ,‘ indefinite detenniner' ,‘possessor' ,‘quantifier' , ‘numeral' , ‘ mass noun' ,‘plural noun' , and so on. Co-occurrence res位ictions within a noun phr出e also need to be represented, such as the fact that some occurs only with mass nouns and plural count nouns. We shall also need to refer to 甲ecific lexical items. For example, a peculiarity of the noun whereabouts s 也at it is virtually res位icted to occurring 明白 the or with a possessive (e.g. his whereabouts); *these whereabouts and *some whereabouts are 咀rtually ruled out. 7 On the one hand, then, a full characterization of the word whereabouts will need to make reference to those noun phrase con宿uctions in which the word is able to occur. At the same time , a full char配terization of those noun phrase constructions will need to make reference , oot only to 由e schematic categories 由.at make up the constructions, but also to 由eir specific instances, such as the noun whereabouts.
2.2.
Part-whole relations amongst constructions
By definition (1) , a construction can be analyzed into component p町ts.ηle question arises , whether a construction can itself function as a part of a larger construction. In general，出is is indeed what we find. Thus , a syllable can function as a p町t of a larger phonological con由uction， namely, a 岛ot. For example, a 岛ot may consist of a strong (or 由essed) syllable followed by a weak (or unstressed) syllable. The construction can be represented 臼 [σsσw]foot. Notice, once ag创n， that it is the ∞ns位uction which defines the properties of its parts, in 也at 由e very notion 。f a strong vs. weak syllable emerges only at the level of the containing ∞nstruction. If syllables are considered in isolation, there can be no basis for categorizing them as strong or weak. Strong vs. weak are relational ootions, which emerge only when a syllable is considered 臼 part of a foot cons住uctlon.
The embedding of a construction within a larger construction is a familiar notion in s严lt缸. For example, noun phrase constructions are an escomponent of clausal constructions. Thus, we might characterize a transitive clause as [NPsu均 VTrans NPOi刷 Obj]. (白le 'fonnula' 岛r a transitive clause is meant to comprise a statement of both its phonological and semantic poles; see 岛0位lote 6.) Once again, we observe that it is only with 黯ntial
7. These observations were confmned by a search ofthe British Nationa1 Corpus , available at http://thetis.bl.ukllookup.h伽让
56
John R. Taylor
reference to the clause schema that we are able to categorize its constituents as ‘ transitive verb' ,‘ subject NP (of a transitive clause)' , and ‘ direct object NP (of a transitive clause)\ It is sometimes possible for a construction of a certain 可pe to contain , as one of its parts , a construction of the same type. A noun phrase can itself contain a noun phrase, a clause can be contained within another clause. Consider the prenominal possessive construction. A prenominal possessive has the schematic structure [NP-POSS N]NP , where POSS is the possessive morpheme. Instances of the construction include the man 's hat, the neighbours' car, some students' essays. Given that the prenominal possessive is itself a noun phrase, we should expect a prenominal possessive to be able to occupy the NP slot within a larger prenominal possessive. This is indeed the case. In the boy's father 's boat, the prenominal possessive the boy's father functions as the possessor within the larger construction. The embedding process can be iterated: the boy's father 's 卢iend's boαt. This aspect of constructions effectively handles what generative linguists have always proclaimed to be a defining property of human languages , namely , ‘ recursion' (see , e.g. Radford 1988: 128). A finite set of ‘ rules' (or, in Cognitive Grammar,‘ constructional schemas') is able to generate an indefinite number of sentences (or, in Cognitive Grammar terms , is able to sanction an indefinite number ofinstances).
2.3.
Comparison with other accounts
My understanding of ‘ construction' differs from those which are to be found in the current literature. There is , to be sure , much uncertainty in the recent Cognitive Linguistic literature on what constitutes a construction.:; Nevertheless , the dominant view appears to restrict the notion to syntactic constructions, which correspond , by and large , to schematic symbolic constructions , in my terminology. Consider Goldberg's definition: (2)
C is a CONSTRUCTION iffdef C is a form-meaning pair
such that some aspect of Fj or some aspect of Sj is not stric t1 y predictable from C' s component parts or from other previousl) established constructions. (Goldberg 1995: 4)
8. See, for example, the heterogeneous nature of the topics dealt with in the recem volume edited by Foolen and van der Leek (2000).
Th e ecology 01 constructions
57
The class identified by Goldberg's defmition intersects with that identified by (1). Thus , Goldberg's definition would include such things as schemati臼lly characterized passive constructions, possessive constructions, transitive and di位ansitive constructions, and the like. However, it would exclude 甲ecific instances of these schematic cons位uctions， provided that the properties of the instances are fully predictable 齿。m those of the schemas. Moreover, (2) would exclude the very possibility of recognizing phonological constructions, since these do not constitute form-meaning 阳出. On the other hand, monomo甲hemic lexical items would turn out to be constructions. The association of the meaning [CAT] 明白 the phonological structure [kæt] cannot be derived from other form-meaning relations in the language. An important recent development is Croft's ‘ radical construction gramm缸， (Croft 2001). Croft's theory is ‘ radical' in that, like the approach sketched out here, it regards constructions as determining the properties of 1heir constituents. Although categories such as ‘ noun' and ‘ verb' may be based in universal semantic-conceptual notions , when it comes to 由e analysis of a specific language these categories can only be defined with E回pect to 也e language-specific constructions in which 由ey OCCUf. Like Goldberg, Croft addresses syntactic issues (也at is, symbolic structures). There is, however, a brief recognition (pp. 61 -62) that the ‘ radical' approach might also be relevant to the study of phonological categories, such 黯 vowel， consonant, and syllable.
2.4.
Motivation
The approach to constructions offered here enables us to tackle, in a rather systematic way, the question of whether, to what extent, and in what ways, a linguistic structure is motivated. 1 define ‘ motivation' as follows: 。)
A linguistic structure is motivated to the extent that it is related to 。由er structures in the language.
All structures in a language are, 1 would maintain, motivated to some exThis amounts to the claim that a linguistic structure - whether this be a phonological, semantic, or symbolic struωre - does not exist in isolation 量。m other facts about a language. While each of the conventional struc阳回 in a language have to be learned and stored 一由is， after all , is what 能nt
58
John R. Taylor
makes them conventional , and which confers on them unit status - it is not the case that each structure occupies a self-contained pigeon hole. Each unit has pointers to other units and can therefore be ‘ filed' under a number of different ‘ addresses\1 refer to this aspect of a linguistic unit as its ‘ ecology'. A linguistic unit occupies a ‘ niche' which is made available by the rest of the language. This serves to introduce a high degree of entropy into a language system. To be sure , some structures might be more highly motivated than others , that is , their position in the language is supported by numerous links to structures of different kinds , while others might be relatively isolated. Just about everything in a language , though, will be motivated , to some extent. A structure which lacked motivation - that is , one which did not participate in relations with anything else in the language would very likely be perceived as not belonging to the language at all. 9 Like my approach to constructions , my definition of motivation is much broader than the one that seems to underlie current discussions. Current approaches to motivation are largely restricted to a忧empts to account for the syntagmatic combination of symbolic units in terms of the properties of the associated semantic construction; altematively , the combinatorial possibilities of a symbolic unit are addressed with an eye on the semantic contribution which the unit makes to the semantic constructions in which it occurs. Wh ilst 1 do not wish to underestimate the possibility that formal aspects of an expression can be related to the symbolized semantic struc阳re， to restrict the notion of motivation to this relation may be undesirable. for at least two reasons: (i) If we restrict our attention to the semantic motivation of syntactic (and morphological) combinations, we exc1 ude in principle the possibility that phonological structures might be motivated , or that phonological structures might motivate syntactic structures. And indeed , phonological structures have been largely neglected in the Cognitive Linguistics literature. 1 am inc1 ined to see this neglect as a symptom of a more general feature of Cognitive Linguistics - at least, of the manner in which Cognitive Linguistics has developed over the past couple of decades. 1 suspect that the ‘ cognitive' m ‘ Cognitive Linguistics' has been taken to be synonymous with ‘ conceptual' , and ‘ conceptual' has been taken to refer only to semantic structure. 9. An example of a fully unmotivated structure would be a word or a phrase of unknown meaning from an unknown language and with an unpronounceable phonology.
Th e ecology 01 constructions
59
Since phonological units , such as segments, syllables, and feet , cannot be characterized in semantic terms , they lie outside the scope of a semanticsbased investigation. On the other hand, as Langacker (1987: 78-9) has rightly obse凹ed， phonological structure is no less a ‘ cognitive' phenomenon than semantic structure. (ii) The approach can be methodologically problematic, to the extent that it sets out to relate observable aspects of a language (namely, syntactic and morphological organization) to what is not observable, namely , semantic structure. Consequently, proposals to motivate syntactic organization in terms of semantic structure often have an air of circularity. One infers a semantic structure from a syntactic organization, which , in tum, is claimed ωbe motivated by the semantic structure. 10
The suspicion of circularity can be countered by appeal to more general pattems in a language. Consider, as an example, the fact that over can designate either a path (He walked over the hill) or a place (He lives over the hill). The two uses are related, in that He lives over the hill construes the place as the endpoint of an imaginary path. We might therefore say that the relation between the path and the place readings (the relation, in fact , is one ofmetonymy) motivates the place reading. This may well be a valid observation , as far as it goes. But it fails to do justice to the real extent to which 也e place-as-endpoint-of-a-path reading is motivated. The motivation ∞mes， crucially , from the fact that the pathlendpoint polysemy is quite well-established in English. It is exhibited by the prepositions past (We drove past the post-o.ffice / The police station is past the post-o.ffic啡， down (We walked down the hill / The cottage is down the hill), through (Go 10. It is in these tenns that Hudson (1990) queried Lang邵阳's (1987) account of the conceptual basis ofthe categories ofnoun and verb. For another example of possible circularity, consider Radden's (1989: 571-2) account of the preposition at. Radden maintains that at, in its temporal sense, is used of points in time , which lack intemal duration. Hence we say at noon , at seven 0 'clock, at sunrise. This use is taken to be a metaphorical extension of spatial at. Problematic for this account could be expressions such as at Christmas , at Easter, at night, and at the weekend, which refer to temporally extended time segments. Radden's answer is that in these cases the periods oftime are "viewed as having no duration". But the only evidence that they are indeed ‘ viewed as having no duration' is , precisely, the fact that they are construed with the proposition at.
60
John R. Taylor
through the kitchen / The pantη1 is through the kitchen) , and several more. Similarity between these examples of the path/endpoint polysemy justifies the recognition of a schema which cap阳res the polysemy relation. The polysemy of over is motivated precisely because it instantiates this wellestablished schema. The schema - and the instances which support it create a ‘ niche' for the endpoint sense of over. Imagine a language whose path prepositions did not , as a group , exhibit an endpoint reading. If, in such a language, a word which symbolized the path sense of over also symbolized the endpoint sense , we should be inclined to regard this as an unexpected , idiosyncratic fact about the lexical item, and not particularly well motivated. There would not be a niche in the language system for this particular reading. The relevance of phonology in discussions of motivation can be illus仕ated on the word hamburger. According to the OED , the expression Hamburger steak was current around the turn of the nineteenth cen阳可 The expression denoted a steak, or something resembling a steak, but made out of minced beef. Hamburger was derived , by -er suffixation, from Hamburg, similar to how Franlifurter is derived from Fran所川， Londoner from London , and villager fromνillage. Hamburger steaks were , namely , associated with the city of Hamburg. At a certain point, it would seem, speakers began to analyze the word as ham+burger. We can be confident of this because of the emergence of words like cheeseburger, eggburger, and even the noun burger. But what could have triggered the reanalysis? The first person to have reanalyzed Hamburger could not have been driven by semantic considerations. At the time the reanalysis was first made , -burger was not yet a morpheme , and therefore did not have a meaning. And while ham did have a meaning , its meaning was not strictly relevant to the issue. Hamburgers , after all , are not made of ham. If semantics were the only consideration, it would be just as plausible (p erhaps even more plausible) to analyze meat as m+eat (meat is something that you eat) , or hear as h+ear (hearing is something that you do with your ears). These last examples are , of course , bizarre. They are bizarre , not because of their semantic implausibility (as 1 have suggested, to recognize the mo甲heme eat in meat is , if anything , somewhat more plausible than recognizing the morpheme ham in Hamburger) , but because the analyses
The ecology 01 constructions
61
is another way of saying that the analysis of meat as m+eat lacks motivatJ. on. What, then , did trigger the reanalysis of Hamburger? The answer, 1 suggest, lies in the phonology. The second syllable of Hamburger contains the long vowel [3:]; being long, this vowel is likely to attract secondary stress: . ['hæm,b3:g:}]. 11 The word has the stress pa忧em appropriate for a compound noun: ['σ] Laσ]. The stress 阴阳m is exhibited by compound nouns such 部 dog-loveκ man-hater， horse-breeding, and count1 ess more. Moreover, [b3:g:}] , with a strong-weak stress pa忧em， is able to constitute a 岛ot. As Cut1er (1990) reports , there is a marked tendency in English for polysyllabic words to coincide with a stress-initial foo t. These aspects of the language conspire to motivate the emergence of burger as a meaning-bearing unit (a process which might well have been supported by the fact that the residual syllable, [hæm] , constituted a meaningful word, even though its meaning was not directly relevant to the case). A comparable reanalysis of Londoner would not have been possible. The second syllable of Londoner ['lAnd:}n:}] contains the inherent1 y unstressed vowel [:}]; there is no way in which -doner [d:}n:}] could have emerged as a mo甲heme， even less as a word, in English. Rather than being driven by semantic considerations, 也en， the reanalysis of Hamburger was dependent on its phonological properties and was motivated by phonological schemas for polysyllabic words and noun compounds.
3. Idiomaticity, productivity, and entrenchment As already pointed out, constructions can differ in terms of the schematicity with which they are specified. In this section 1 consider, briefly, three fur1her dimensions along which constructions can differ. 12 These concem the idiomaticity, productivity, and entrenchment of a construction. (i) Idiomaticity. The idiomaticity of a construction has to do with the extent to which its properties cannot be predicted from other aspects of the language. Goldberg (1 995) , it will be recalled, applied this criterion to her
11. My transcription presupposes a non-rhotic accent. Rhoticity, however, is not pertinent to the point under discussion. 12. For a fuller account of these issues , see Taylor (2002).
62
John R. Taylor
definition of a construction, restricting the tenn to those assemblies of fonnal and semantic structures whose properties are not predicable from other facts about the language. On Goldberg's definition , The farmer shot the rabbit would not be regarded as a construction, since the properties of this expression are fully predictable from other facts about English. These ‘ other facts' include (i) the transitive clause schema [NPsubj VTrans NPOirect Obj] with its associated phonological and semantic specifications , (ii) a noun phrase schema sanctioning the expressions the farmer and the rabbit and their associated semantics, (i ii) a schema for the grounding of a clause by means of a past-tense verb fonn , (iv) the status of shot as the ‘ idiomatic' past tense of shoot, as well as (v) the fonn-meaning relations symbolized by the lexical items farmer , shoot, and rabbit. Given a knowledge of all these , a speaker can surely predict that The farmer shot the rabbit will be a fully acceptable expression in English and will be able to compute its meaning. On the other hand , a speaker will not be able to compute the (idiomatic) meanings of The farmer /d cked the bucket, The farmer 再pilled the beα肘， or The farmer pulled my leg. Although these expressions instantiate fonnal aspects of the transitive clause schema, their semantic properties need to be specifically stated, at the level of the instances. (ii) Productivity. This notion applies primarily to schematically characterized constructions. A schematic construction is productive to the extent that instances can be freely generated in accordance with the specifications of the schema. A moderately productive schema is the [X by X] construction , which designates the rate at which a process unfolds , whereby [X] designates the unit by which the rate is measured. The schema is instantiated by such expressions as one by one , day by day , page by page. There are perhaps a couple of dozen such expressions in regular use. While a speaker may be able to create new instances in accordance with the schema, not every candidate for the X-slot resu 1ts in an acceptable expression. We could not say that people entered *seνeral by several, and it would be odd to say that someone spent their inheritance ?cheque by cheque. Knowledge of the schematic construction involves , inter alia, knowledge of the range of its potential instances. A non-productive construction is exemplified by the phrase by and large. There is no schema in English which sanctions the coordination of a preposition and an adjective. We cannot have *in and big, *out and small, *at and hungr予人 In a sense, by and large is sanctioned by itself, and by its own high degree of entrenchment. Finally, for a highly productive schema, we can tum to the transitive clause construction [NPs向
1页e
ecology 01 constructions
63
VTrans NPDirect Obj]. Even this construction, however, is not fully productive. One issue concerns the range of verbs which are able to fill the VTrans-slot, another concerns the semantic roles of the subject and direct object nominals. Prototypical examples of transitive verbs include hit, kill, and shoot. These verbs designate an intentional action of the part of the subject refer- . ent which materially affects the direct object referent. Verbs of perception (see , hear) , verbs of cognition (remember , know, forget) , and verbs of cognitive attitude (like , regret , want) are also assimilated to the transitive verb categ。可， whereby the perceiver, cognizer, or experiencer are construed as subject, the perceived, cognized, and experienced as direct object. Even more distant from the transitive prototype are clauses of the kind The fifth day saw their departure , The last decade witnessed many changes. Here , 也e subject nominal designates the setting , or context, for what is strictly 甲eaking a one-participant event. The number of verbs which can appear in setting-subject clauses is strictly limited. For example, we cannot say that 叮ne Ji..仙 day heard the explosion. Setting-subject expressions - as a subcatego叩 of the transitive clause construction - need to be specified with respect to those verbs which are eligible to occur in it (and the kinds of subjects and direct objects that they can take). Contemporaneously, a full characterization of the verbs see and witness would need to state, not only the verbs' availability to designate perception (1 saw/witnessed the accident) , but also their availability to function in a setting-subject construction. (iii) Entrenchment. A construction is entrenched to the extent that it is repin the speaker's mental grammar. Entrenchment is a function of frequency of successful use. 1 dare say that for most English speakers, the 缸pression Have a nice day is highly entrenched. This is in spite of the fact 也就 the expression is not particularly idiomatic - at least, it is not idiomatic 扭曲e way in which kick the bucket is idiomatic. Its meaning is predictable from the semantics of the imperative construction which it instantiates and from the meanings of its component parts. Yet it is doubtful whether a 笃陀aker who says Have a nice day is constructing the expression in accordance with the imperative schema. Although it conforms with the schema, it is accessed as a pre-formed unit. We can also speak of the entrenchment of a schema. A schema is entrenched to the extent that it is frequently used , 也at is , to the extent that it is imminent in a wide range of different inslances. An entrenched schema will be readily available to sanction new ÏDstances. The re-analyses of Hamburg+er as ham+burger was possible in virtue of the entrenchment of the relevant phonological schemas. E臼ented
64
John R. Taylor
4. The ‘ Bang goes' construction In this section 1 exemplify the concept of motivation with reference to a specific construction in English. Consider the following expression: (4)
Bang goes my weekend!
An analysis of this expression involves , first of all , a statement of its formal and semantic aspects. We shall also need to consider the place of this expression within the broader context of the English language. This will involve an examination of the parts of the construction, the schematic units of which the parts are instances, as well as the schematic constructions of which the complex expression in (4) is an instance. The expression in (4) will be motivated to the extent that it can be related to these various other facts about English , and thereby occupies an ecological niche within the language. Let us begin with the semantics. The expression in (4) conjures up a rather specific scenario. You have planned some activity for the weekend and are looking forward to it. Just before the weekend , something comes up which interferes with your plans. It could be that your boss requires you to submit a report first thing on Monday moming, or that family commitments require you to visit an aged relative. The interfering activity takes precedence and you need to cancel your plans. You feel annoyed , but cannot reasonably renege on your duties. The expression has a familiar , even jocular tone - nothing of great importance is actually at stake. If your grandmother had died and the funeral was on Saturday, it would be in very bad taste to utter (4). It would also be inappropriate, on leaming that your spouse , whom you love dearly , has set up home with another person , to comment: Bang goes my marriage! or, on leaming that you have a terminal illness , to declare: Bang goes my life! The possibility of substituting my weekend with other nominals - Bang goes my Saturday afternoon! , Bang goes my nice new theo刷， Bang go our chances 01 making a million! 一 suggests that (4) should be regarded as an instance of a more schematically characterized construction. We might represent this as [bang go NPSubj ] , where NP designates the cherished object. The nominal functions as the subject of the verb , as shown by number agreement on go. The construction is productive , in that very many nominals can occupy the NP slot. It is not the case , however, that any NP is equally appropriate in the construction. Neither is it the case that the clause
The ecology 01 constructions
65
can be grounded by the full range of possibilities that are available for other kinds of c1 auses.
Table 1. Number of intemet hits for collocations involving bang 缸" bang goes my banggomy bang goes your banggo your bang goes his bang go his bang goes her bang go her bang goes its bang go its bang goes the bang go the bangwentmy bang went your bang went his bang went her bang went its 坦E皇 went the
132 22 2626 9 31 3 2
bang goes our bang go our
31 2
bang goes their bang go their
8
bang went our
10
bang went their
3
O O
604 191 42 O
59 8 O
493
Number ofhits exemplifying: Unambiguously singular subjects: Unambiguously plural subjects:
3435 37
Subjects with 1st person possessors: Subjects with 2nd person possessors: Subjects with 3rd person possessors: Subjects with definite article:
239 2635 117 1288
Present tense verb: Past tense verb:
3663 615
66
John R. Taylor
These topics were investigated using the intemet search engine www.AllTheWeb.com. The search scored 4072 hits for the collocation ‘ bang goes' against only 587 for ‘ bang go' and 1012 for ‘ bang went\13 It will be noted that present tense is preferred over past tense , and that singular subjects far outnumber plural subjects. Some further properties of the construction can be gleaned from the search results summarized in Table 1. It will be noted, once again , that singu1ar subjects outnumber plura1 subje邸， also that the cherished 。同 ect is overwhelmingly associated with a speech act participant (mainly the addressee) rather than with a third party. The preference for the present tense is also apparent. Not shown in the Table is the fact that perfective ‘ bang has /h ave gone' ，自lture ‘ bang will go\ and various modals , e.g. ‘ bang might go' , as well as negated expressions such as ‘ bang doesn't go' , scored no hits at all. These features of the bang goes construction indicate that the construction is firmly anchored in the speech act situation. The speaker expresses her current annoyance at the prospect of having to abandon a cherished plan, or empathizes - somewhat ironically , perhaps - with the addressee at their having to abandon their plans. Past tense is appropriate to the extent that a speaker reports a past experience. Others tenses , the various modalities , and of course negatives , would be incompatible with this characterizat lOn. Let us next consider the place of the construction within a broader linguistic context. A notable feature of the construction is that the subject nominal follows the verb. Some other constructions in English share this property. These include the deictic constructions in (匀， the preposed directionals in (6) , and the preposed locatives in (7). (5)
a. There's Har吵" wearing his new jacket. b. There goes Harη" with his new girlfriend. c. Here comes the bus.
(6)
a. Away ran the children. b. Out 01 the hole crept a timid mouse. c. Along this road marched the army.
13. 1t is likely that not all the intemet hits pertain to the bang goes constructioc However, there is no reason to doubt that the numbers reflect real pa忧ems L. the construction's use.
The ecology 01 constructions (7)
67
a. Up on the hill used to stand the governor 's residence. b. On the bed was lying the sleeping child. c. On the other side 01 the lake stands the war memorial.
The deictics (5) are introduced by here or there and serve to direct the . hearer's attention to an entity currently in the speaker's percep阳al field; 可pically， the verb is one of be, come , and go , and in the present tense (Lakoff 1987). The directionals (6) designate the path of a moving entity; they are introduced by a directional phrase while the verb designates motion or manner of motion (run , climb , crawl, etc.). Directionals may be used in narratives , therefore , unlike the deictics, they can be associated with the p臼t tense. The locatives (7) are introduced by an expression of place , while the verb is a locational verb，句'pically be , stand, or lie. Since the locative constructions serve to describe the place of an entity, they are compatible with a variety of tenses. The deictic , directional , and locative constructions in (5) - (7) are each associated with their own construction-specific properties, concerning their semantics and the kinds of items that are eligible to fill the X and V slots. Their similarities, however, point to the existence of a schematic subjectfinal construction in English, namely , [X V NPSubj ] , of which the deictic , directional , and locative constructions are instances. The bang goes construction also instantiates the [X V NP Subj ] construction, and it, too , is associated with special properties. These concern , not only its semantics, but also the specification of the initial constituent as the lexical item bang and 也e verb as go , in either the present or the past tense. The bang goes con由uction is special in another way. The constructions in (5) - (7) may take a pronominal subject, in which case the subject appears in its normal position before the verb: There he is , Away they ran , On the bed he lay. The position of a pronominal subject does not arise with the bang goes construction, since the construction seems not to tolerate a pronominal subject at all: .Bang it goes , *Bang goes it. The fact that the bang goes construction permits only the verb go invites us to look more closely at this word (cf. Radden 1996). The verb can designate not only motion through space (1 went to the airport) , but also a change of state over time: The milk went sour, The light went red. A further 圃e refers to a static situation , which the conceptualizer scans through conceived time (Langacker 1987: 172; Mat1 0ck This volume): The road goes tluough the mountains. In addition, the verb can evoke the sequential structure of an activity (lt takes only
68
John R. Taylor
bet) , as well as the sequential structure of a text, either a linguistic text , such as a poem or a joke, or a musical text: (8)
a. Do you know how the poem goes? b. 1forget how the joke went. c. The tune goes like this.
Observe that (8c) would normally be followed up by the speaker 邵阳ally humming or singing the tune in question. This brings us to the ‘ quotative' use of go. The verb introduces a mention of a perceptible event, either acoustic or visual (9); it cannot, however, introduce a report or a description (10). (9)
a. He went [f:::]. b. He went [OBSCENE GESTURE]
(1 0) a. *He went a hushing sound. b. *He went an obscene gesture.
(11)
'Yes, of course, ' went the young man.
The quoted event can be a linguistic utterance (11); moreover, as this example shows , the quoted element can be fronted. When the quoted element is fronted , as in (11) , the expression instantiates the [X V NP subj] schema that we invoked in connection with the bang goes construction. Interestingly , quotative go can be associated, not only with a quotation as such , but with an ideophone, that is , with a word which conventionally designates an audible event. Ideophones which are compatible with go include pop, crash, crack, bump, thud, slap, whoosh , whiz喃 swish , as well as bang. The ideophone can be in initial or final position.
(1 2) a. The light bulb went pop. b. Crash went the Heffalump's head against the tree-root. C. Wh oosh, whoosh, whoosh went the blood through her veins. d. The ba l/oon went bang when it burst. This brings us to the word bang itself. Like go , the word has a variety of uses (Riemer 1998). As a noun it designates a kind of sound (There was a loud bang) , either the sound of an explosion or the sound of one object
Th e ecology 01 constructions
69
coming into sudden contact with another. The idea of a sudden and noisy impact predominates in verbal uses: They banged on the door (i.e. knocked loudly on it) , He banged the door shut (i. e. closed it noisily). A further use of bang is 臼 a kind of adverbial modifier (It was bang on target; Your remark was bang on). Here, the idea is ofprecision - a pr'咱 ectile hit the tar-. get at 由e precise intended point, the remark was highly relevant to the discusslon m progress. These uses of bang appe缸 to be motivated by the word's onomatopoeic character. One feels 也at the pronunciation of bang iconically resembles the sound 由at the word designates; in fact , Collins English Dictiona叨， as well 副也e Shorter OED , claims that the word, like its Germanic cognates, is ‘ of imitative origin'. The word's onomatopoeic character comes to 也e fore in (13):
(13) a. He shut the door, bang! b. Hefired the rifle, bang! Yet if one considers the matter dispassionately, 1 think one would have to agree 也at 也e word's pronunciation doesn't really resemble a banging sound. How, then, can we account for the very strong intuition 也at the word is imitative? The reason, 1 suggest, lies in the phonetic and semantic associations triggered by the word. There are quite a few monosyllabic words which designate a noisy impact and/or sudden movement and which contain the [æ] vowel: (14)
slam, slap, crack, clap, flap, crash, bash, clang, spank, smack, wham, prang
There are several words commencing in [b] which are associated with a sudden event: (1 5)
boo, beat, bat, batter, bump, binge, bingo
In a number of words a final nasal (especially a velar nasal) is associated with sound or movement: (16)
sin，ι ring，
ping, zin，ι flinι
sling，
dong, gon，ι
hum，
vroom , boom
70
John R. Taylor
Moreover, several words , such as prang, clang, slam , wham , and batter. can be listed under more than one of these categories. All these associations serve to reinforce the appropriateness of the sound-meaning relation symbolized by the word bang. It is because of this network of associations , 1 suspect, that English speakers perceive the phonological structure [bælJ] to be an appropriate form to designate the sound of sudden impact. The sound-meaning relation symbolized by the word is motivated, not so much by an objective similarity between the meaning and the pronunciation, but by the cumulative effect of a network of associations pertaining to the word's phonological components. 14 As noted, the ideophone bang can occur with quotative go , with either a preposed or a postposed subject nominal. (17) a. The balloon went bang, when it burst. b. Bang went the balloon, when it burst.
It looks like we have come full circle - Bang went the balloon exhibits exactly the same structure as Bang goes my weekend. Semantically, however, the expressions differ in important respects. Bang in Bang went the balloon represents the sound that the balloon made as it burst, but there is no suggestion that my weekend makes a sound when my plans for the weekend have to be abandoned. The two uses of bang are not entirely unrelated, however. If something ‘ goes bang' it is destroyed, either through an explosion or sudden impact. The bang goes construction perspectivizes this aspect - my plans for the weekend are destroyed by obligations which interfere with those plans. The constructions also differ in their formal aspects. The ideophone can occur after quotative go (The balloon went bang) , a possibility not available for bang goes expressions 俨My weekend wenr bang). The quotative construction is also compatible with a pronominal subject (It went bang, Bang it went) , something which is again not possible with the bang goes construction. These considerations point to the need to recognize the bang goes construction as a distinct constructional idiom of English, whose properties cannot be fully predicted from other aspects of the language , and which therefore needs to be specifically learned by speakers of English. Nevertheless , in spite of its idiomatic properties, the construction is not totally 14. Associations between the sound and meaning of(monomorphemic) words have been extensively explored by Rhodes and Lawler (1 981).
1页e
ecology 01 constructions
71
encapsulated 企om
other facts about the English language. The construction is motivated by numerous affinities , both formal and semantic, with other structures in the language户 The construction exists, and survives, because of the ecological niche which it occupies in the inventory of linguistic units which constitute the English language. 16 As a matter of fact, similar remarks apply to the word bang itself. The sound-meaning relation symbolized by the word cannot be predicted from other facts of the language. Nevertheless, other phenomena in the language serve to motivate the sound-meaning relation, and contribute to a speaker's impression 由at the sound of 也e word is somehow appropriate for the meaning which it symbolizes. Bang also has a niche amongst the cluster of words designating movement, sound, and impact. In a sense, the motivation of the word bang is a microcosm of the manner in which the bang goes cons位uction itself is motivated.
5. Conclusion languageωa "structured inventory of units" (Langacker 1987: 57). On any reckoning, the number of units which a fluent speaker of a language needs to master is immense. 白白， one might suppose, would impose an intolerable burden on a speaker, in terms of learnit毡， storage, and retrieval. It is not the case, however, that each unit in the inventory constitutes an encapsulated piece of information. The entropy of the grammar is reduced - hence, the learning, storage, and re住ieval of linguistic units is facilitated - by a dense network of relations which structures 由e inventory. It is these relations which, α皿lUlatively， serve to ‘ motivate' a linguistic structure, and which create a
Langacker defined the grammar of a ∞nventionallinguistic
15. Even the jocular, light-hearted tone of bang goes expressions may not be peculi缸 to the construction. For my part, adverbial uses of bang (It was bang on!) have a somewhat jocular feel to them. 16.1 have encountered English speakers who claim to be unfamiliar with the bang goes construction. Nevertheless, if asked, even these speakers 缸e quickly able ωdescribe， with considerable accuracy, the kinds of situations in which Bang goes my weekend! might be uttered. The让 ability to understand the expression is presumably a function of the activation of the numerous similarity relations betw回n 部pects of the construction and other, more familiar facts about English.
72
John R. Taylor
‘ niche' for a unit within the language system. Each structure stands , as it were, at the hub of a network of relations to other units. Three kinds of relation were addressed. These are the relation of a part to a larger configuration , the relation of a more schematically characterized unit to its more 卸lly specified instances , and the relation of similarity between the instances. In the case of symbolic constructions 一 traditionally referred to as ‘ syntactic constructions' 一 the relations pertain to the phonological , semantic , and symbolic constituents , as well as to the status of a construction itself, both as a schema which sanctions instances , and as an instance of more schematic constructions. These relations were explored on the example of the bang goes construction in English. The construction is idiomatic. in the sense that its properties - as well as its very existence - cannot be predicted from other facts about the language. However, as 1 have tried to show, the parts of the construction - the words bang and go - have uses outside the construction which have affinities to their values within the construction. Moreover, the form of the construction suggests that it be regarded as an instance of a more schematically characterized [X V NPSubJ construction, while the construction's special semantic value determines the properties of the nominals that are able to occupy the NP slot, as well as the restricted range of tenses and modalities that can serve to ground the clause As a consequence of all these , bang go expressions strike users of English as a particularly felicitous and economical way to express mild annoyance at the frustration of their plans. It is in this sense that we can speak of the motivation ofthe expression, Bang goes my weekend!
References Croft, William 2001 Radical Construction Grammar: 砂ntactic Theory in Typological Per，再pective. Oxford: Oxford University Press. Cutler, Anne 1990 Exploiting prosodic probabilities in speech segmentation. In: "町 Altmann (ed.) , Cognitive Models 01 年eech Processing: Psycho!inguistic and Computational Perspectives , 105-121. Cambridge, Mass.: MIT Press. Foolen, Ad and Frederike van der Leek (eds.) 2000 Constructions in Cognitive Linguistics: Se!ected Papers 卢'om the Ft斤h lnternational Cognitive Linguistics Conference, Amsterdam , 1997. AmsterdamJPhiladelphia: Benjamins.
1到e
ecology of constructions
Goldberg, Adele 1995 Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Hudson, Richard 1990 Review ofLangacker (1 987). Lingua 81: 272-284. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar, Vol. 1: Th eoretical Prerequisites. Stanford: Stanford University Press. Lakoff, George 1987 Women, Fire, and Dangerous Th ings: Wh at Categories Reveal about the Mind. Chicago: University of Chicago Press. Radden, Günter 1989 Figurative uses of prepositions. In: René Dirven and Richard Geiger (eds.), A User's Grammar of English: Word, Sentence, Text, Interaction , Part B: Th e Structure of Sentences , 551-576. Frankfurt: Lang. 1996 Motion metaphorized: The case of coming and going. In: Eugene Casad (ed.), Cognitive Linguistics in the Redwoods, 423-458. Ber1in: Mouton de Gruyter. Radford, Andrew 1988 1协lsformational Grammar: A First Course. Cambridge: Cambridge University Press. Rhodes, Richard and John Lawler 1981 Athematic metaphors. Proceedings ofthe Annual Meeting ofthe Chicago Linguistic Society 17: 318-342. Riemer, Nick 1998 The grammaticalization of impact: Bang and slap in English. Australian Journal ofLinguistics 18: 169-83. Taylor, John 2002 Cognitive Grammar. Oxford: Oxford University Press.
73
Expressive binominal NPs in Germanic and Romance languages AdFoolen
Abstract
This paper analyses constructions such as ‘ an angel of a child' in different Germanic and Romance languages. The construction consists of two NPs linked by the preposition 01 in English and simi1ar ‘ neutral' prepositions in the other languages considered. It conveys expressive meaning and is , therefore, referredωas ‘ Expressive Binominal NP' (EBNP). The paper raises the question which of the two NPs is to be considered the head of the construction. Both NPs qualify to a certain extent. The ωmparison with simi1ar cons往uctions in Oceanic and other non-IndoEuropean languages is illuminating here. The construction has two functions: a referential function and an expressive function. It is argued that the double-headedness of the construction is motivated by its double 岛nction: 也e first NP is the expressive head of the construction, whereas the second NP is the referential head. Keywords: attribute , blending, expressivity, Germanic languages , head, metaphor, Noun Phrase, of, Romance languages
m缸ke命less，
1. Introduction Present-day English , Dutch and German , and Romance languages have an construction in common. In (1) , some instantiations of the conslruction are presented:
E阳'esting
76
( 1)
AdFoolen
a. English:α bearofa man b. Dutch: een boom van een kerel ‘ a tree ofa man' c. French: un diable d 'homme ‘ a devil of a man'
In the Germanic and Romance languages, the construction has three obligatory slots, namely two noun phrase slots and a slot for a preposition that relates the 阳'0 noun phrases. This preposition is invariably the semantically most ‘ neutral' preposition available in the language , namely of (English) , van (Dutch) , de (French) , etc. There are strong constraints on the determiner of the second NP and less strong conditions on that of the first determiner (an unmarked definite article like the in English is not allowed). For ease of reference , 1 will use the shorthand EBNP for ‘ Expressive Binominal NP' for the construction, combining the functional (expressivi可) and formal (two NPs) characteristics of the construction in the name. 1 will refer to the first determiner as Det1, to the first noun as N 1, to the preposition as P , to the second determiner as Det2 and to the second noun as N2. Besides these main constituents, the construction is optionally expandable by adjectival and PP modifiers (see section 2 .4). The analytical goal of this paper is to determine the formal properties of the EBNP in the Germanic and Romance languages; in particular, 1 will go into the properties that are common to all these languages and point out language-specific differences with regard to constraints on the way the syntactic slots can be filled. The main theoretical objective of this paper is to decide whether the EBNP construction can be considered a motivated cons位uctíon.
1.1.
Expressivity in language
Language users show their involvement in what they say by using certain words , morphemes, constructions or intonation pa忧ems. The expressive function (Ausdruck) is one of the three semiotic functions that Bühler (1934) distinguished with regard to natural language, the other two being the representational (Darstellung) and directive (Appell) functions. Bühler considered the representational (conceptual) function to be central to natural language, in line with an old and still prevailing tradition in linguistics. Since the 1970s, research on the directive function has been stimulated by speech act theory. Only in recent years has interest in the expressive function developed a certain momentum (cf. the overview in Besnier (1990) and
Expressive binominal NPs in Germanic and Romance languages
77
the special issue of the Journal 01 Pragmatics , edited by Caffi and Janney (1994)). The literature on expressive syntax is sti11 rather scarce, but see the papers by Michaelis and Lambrecht (1 996a, 1996b) on exc1 amative sentence 可pes. The present paper, which is part of a longer project on the expressive function oflanguage (cf. Foolen 1997), intends to contribute to the research topic of expressive syntax by concentrating on the EBNP , a conS位uction on the subc1 ausallevel. 1 The intuition that the EBNP has an expressive value is a linguistic intuition that is shared by at least some of the linguists that have written on the construction (cf. the subtitle of Leys (1997): ‘ Emotionality as constructional meaning'). Supporting evidence for this intuition can be found in the fact that the construction a忧racts nouns meaning things like ‘ scoundrel' , ‘ hell' , etc. , and in the impression that it typically occurs in text types where emotional involvement is to be expected: informal dialogue, written reviews of books and films , reports of sports events , etc. For the pu叩ose of this study , expressivity is understood in the sense of ‘ having emotional content'.
1.2.
Motivation
Taylor (this volume) discusses the notion of ‘ motivation' in grammar (see in particular his section 1.4). According to Taylor, "a linguistic structure is motivated to the extent that it is related to other structures in the language." These relations constitute the ‘ ecology' of the structure. The structure occupies a ‘ niche' which is made available by the rest ofthe language.
1. At the HamburgConference on Motivation in Grammar, 1 presented a general paper on expressive syntax. For the present volume, 1 decided to restrict the scope of the paper to one construction. 1 presented parts of the present paper at the Intemational Pragmatics Conference in Budapest (July 2000) , the Linguistics Colloquium of Ri ce University, Houston (March 2001) , and in a lecture for the Department ofGermanic Languages at UCLA, Los Angeles (May 2001).1 thank the audiences for their stimulating discussion, in particular Jan-Ola Östman (Helsinki), Bob Kirsner (UCLA) , Suzanne Kemmer, Michel Achard and Sebastian Ross-Hagebaum (Rice). Special thanks to Knud Lambrecht (Austin) , Frederike van der Leek (Amsterdam) and Pieter Seuren (Nijmegen) for their constructive suggestions and support. All remaining weaknesses of the paper aremme.
78
AdFoolen
With respect to the EBNP construction, there has been a lot of uncertainty in the literature as to its proper analysis and, thus , to its ‘ ecology'. In particular, the question whether Nl or N2 is the ‘ head' of the construction is a bone of contention. If N 1 is the head , then the construction is relateè primarily to an NP schema with a PP postmodifier. If N2 is the head, then the construction can be related to an NP 可pe with an adjectival premodifier. A third option would be , to consider both Nl and N2 as heads because the head features are distributed among them. In that case the construction would have a ‘ hybrid' status, being motivated by the two different I\'P schemas just mentioned. It seems that something of the latler type indeed holds for the EBNP construction. If that is the case , then this might explain why linguists could not come to an agreement about which noun should be considered the head of the construction. As Taylor points out, there is a second, and more commonly accepted. view on what motivation in grammar means. In this perspective, a construction is motivated if its formal properties can be related in a nonarbitrary way to the semantics or function of the construction. If we call motivation as discussed above ‘ structural motivation' , then we could call this second type of motivation ‘ iconic motivation'. In my view , this second type ofmotivation is relevant to the EBNP construction as well: the doubleheadedness of the construction can be considered to be motivated by the two functions which the construction simultaneously fulfills. More specifically, the head status of N2 has to do with the representational function of the NP , whereas N 1 is the head in relation to the expressive function of the construcÌl on.
1.3.
Data
When native speakers of English are presented with a few instances of the EBNP construction, they usually have no difficulty in providing additional examples. Typically, they use a restricted set of examples that seems to be wholly or partly stored in their mentallexicon: 2
2. 1 thank Patricia Zukow-Goldring (English) , Amalia Mendes (Portuguese 1. Andrei Filtchenko (Russian) , Svitlana Zhabotynska and Jewgenija Sokolinskaja (Russian) , Esther Pasual (Catalan) , and others for providing me with useful and entertaining examples.
Expressive binominal NPs 切 (2)
Germanic
a hell 01 ajob, a dream 01 a caκ
a
and Romance languages
79
dud 01 a film , a blast 01 a party
Aco甲us
search shows , however, that the construction is more productive than might be conc1uded on the basis of spontaneous speaker reactions , i.e. it is not infrequent and is lexical1y fi l1ed with a variety of nouns: 3
(3)
a barn 01 a place, a Battenburg cake 01 a building, a battered shell 01 a building, a bear hug 01 a welcome， α beast 01 a bestseller, a beautiful shrug 01 a pe功rmance， α belly flop 01 a play, a belly-laugh 01 a novel, a butterball 01 a kid
Similar examples were provided by native speakers of the other languages or found in publications on the construction: (4)
Dutch: een juweeltje van een universiteit ‘ a jewel-DIM of a university' jouw leestneus van een broer ‘ your party-nose of a brother' zijn 阳lee apen van kinderen ‘ his two apes of children' die sigaren rokende duivel van een echtgenoot ‘ that cigarsmoking devil of a husband'
(5)
German: ein Engel von einer Frau ‘ an angel of a woman' dieses Nest von einer Hauptstadt ‘ this nest of a capital' dieser Windhund von Fausto ‘ this greyhound (= careless and unreliable sort of person) of a Fausto' so ein Apparat von Karton ‘ such a monster of a box'
(6)
French: unj坤on de valet ‘ a scoundrel of a servant' un diable d 'homme ‘ a devil of a man' cet imbécile de Jules César ‘ this idiot of a Julius Ces町，4 quelle drôle de planète ‘ what a funny planet'
3. 1 thank Michael Barlow (Rice University), who did a co叩us search on the (English) Sunday Times , using the frame ‘ a N of a N' as search string. 1 present only a few examples , starting with the le忧er 币， after the first deterrniner ‘ a\ 4. Note that in the English translation the proper noun gets a determiner. The logic behind the occurrence of Det2 is not yet clear to me, cf: that idiot 01 a brother, but: your idiot 01 brother; in German: dieser Idiot von Bruder.
80
(7)
AdFoolen
Other Romance languages: Italian: una bestia di avvocato ‘ a beast of a laywer' Spanish: el imbécil de tu hermano ‘ this idiot of your brother' Catalan: Aquest cony de cotxe no卢nciona. This bush [colloquial, swearword for female sex organ] of car doesn 't work. ‘ This damn car doesn't work'. Portuguese: una joia de universidade ‘ a jewel of a university' Romanian: idiota a~at de Maria idiot-the that ofMaria ‘ that idiot ofMaria ,5
When asked to translate expressions such as the ones above into their native language , speakers of Russian and Finnish usually c1 aim that a literal translation in their own language is not possible. Instead, they offer examples like the following as equivalents with the same expressive value: (8)
Russian: a. medved', a ne chelovek bear, but not man ‘ a bear ofa man ,6 b. angel, a ne zhenschina angel , but not woman ‘ angel of a woman'
(9) Finnish: a. pojan nulikka boy-GEN rascal ‘ a rascal of a boy' b. piru mieheksi devil man-TRA (translative case) ‘ a devil ofa man'
5. Norbert Corver, Taal in zicht. Utrecht University 2002 , p. 29. 6. Russian also has another way of rendering the same expressive meaning by compounding the two nouns as in medved-chelvek ‘ bear-man' , which may also be reversed as chelvek-medved.
Expressive binominal NPs in Germanic and Romance languages
81
Thus , Russian and Finnish do not seem to have the EBNP construction in form of an NP-P-NP-structure, but use other constructions to express involvement of the speaker with regard to 由e referent ofthe NP. 7 With regard to diachrony, there seems to be no study on the history of 也e construction. 1 only 岛und the 岛llowing global sketch in Leys (1997). The Romance languages inherited the construction from Latin, which used the genitive, cf. monstrum mulieris, monster of a woman-GEN,‘ a monster of a woman'. The cons位uction was then taken over by the Germanic languages. According to Leys, the indefinite 缸ticle in the Det2 position in the Germanic languages was, app缸'ently， an innovation after 也e introduction of the construction in the Germanic languages. For the Scandinavian languages, Leys (1997: 33) reports that Swedish has en idiot till vakmästare ‘ an idiot of a guard' , in which 由e preposition till ‘ to' is used,8 and he adds 由at Old High German similarly used the ‘ goal' preposition zu , e.g. ein helt ze manne, ein wunder zeinem weib. Later, this preposition was replaced by von under the inf1uence of French de. So it seems that a similar construction was already present 扭曲e earlier history of the Germanic languages and that it wω ‘ remodeled' ， in p缸ticul町 with regard to 也e preposition, under the in f1uence of Romance languages.
由e
2. Analyzing the EBNP construction 2.1.
Reference grammars
The main reference grammars of English, German and Dutch show some awareness of the existence of the EBNP construction, but their treatment is , 。n 也e whole, not very convincing. Quirk et al. (1985: 1284) discuss the construction in section 17.47 ‘ Apposition with oJ二phrases' ， together wi由 other NPs containing appositive postmodifiers, like the city 01 Rome. The main 岛cus of Quirk et al. is on 也e question which NP is the head of the construction. The EBNP con-
7. Turkish seems to behave like Russian and Finnish in pu忧ing two nouns in juxtaposition: Nuri serseri-si, Nuri vagabond-3sg，‘白at vagabond of Nuri' (Ineke van der Craats, p.c.). 8. In Danish, the exclamation helvedes til v，矿r ‘ hell of weather' , literally ‘ hell to weather' occurs, but the construction is very rare. (Lene Schøsler, p.c.)
82
Ad Foolen
struction represents , in their view , "a special case of prepositional apposition [...]. This structure [...] is not a regular prepositional postrnodification , since N2 is notionally the head, as can be seen in the paraphrases: [...] an angel of a child - an angelic child". Any mentioning of a special constructional meaning is absent. The German reference grammar Duden (1995: 249) does not regard the EBNP as a separate construction. It restricts itself to a morphological remark , saying that if the first noun is in the nominative and if in the appositive prepositional postrnodifier the indefinite article is left out (a possibility in German , see 2 .4), then this second noun can take over the nominative case of the first noun , instead of having the dative case , as would be the unmarked case after von: ein armer Teufel von Philologe , ‘ a poor devil of a philologist' , instead of Philologen , ein Lump von Angestellter, ‘ a scoundrel ofan employee' , instead ofthe dative case ofthis word. 9 The Dutch reference grammar ANS differs in a positive way 仕om its English and German counterparts, in that it does consider the EBNP as a separate construction that deserves discussion in a subsection of its own (p. 853-854). This is most certainly due to Paardekooper (1956) , the ti t1 e of whose article , Een schat van een kind ‘ A darling of a child' , became the standard name in Dutch grammar for this construction. The ANS discusses the construction in much more detail than the English and German reference grammars. It points out that: 一
一
N2 is the head , Nl specifies certain properties ofN2 The construction is called a ‘ metaphorical construction' , i.e. the construction contains a metaphor as in a bear of a man ‘ the man is a bear' , etc. Nl must be affectively loaded , and this affective meaning is then transferred to the construction as a whole.
However, there are some problems with regard to each of these three claims. First, the claim that N2 is the head is controversial. It seems to be mainly based on semantic intuition and paraphrase (see further 2.5). Sec9. In the discussion of the paper that 1 presented on the EBNP at the Germanic Department of UCLA , it tumed out that there is uncertainty among native speakers of German what exac t1 y the dative form of Angestellter in this context would be like: Angestellten oder Angestelltem. For further discussion on such nominalized participles in the EBNP construction see Vui1l aume (1993: 170).
Expressive binominal NPs in Germanic and Romance languages
83
ond, to call the construction metaphorical suggests that there is always a built-in metaphor. This is indeed often the case, but not always (see 3 .1). Finally, as 1 will argue in section 3.2 , Nl is not necessarily affectively loaded, so we have to account for the expressive meaning of the construction in a different way.
2.2.
Construction Grammar
The first step that should be taken to improve the existing descriptions is to consider the EBNP as a genuine construction in the sense of Construction Grammar: a construction with its own set of formal properties and with its own non-compositional constructional meaning. 1 will not apply any Construction Grammar formalism here, but what authors like Goldberg (1 995) and Michaelis and Lambrecht (1996a, 1996b) speci马， as general properties of constructions, seems to fit the expressive binominal NPs very well: 一
Constructions
are units sui generis in a grammar, with their own mean-
mg. 一
一
Constructions
have a range of properties involving different levels: prosodic, morphological , syntactic, lexical. Insofar as a construction shares a certain property with another construction, there is a relation between these constructions. The grammar of a language is thus conceived of as a family of related constructions. In Construction Grammar, there is no strict dividing line between lexicon and grammar. Constructions that are often considered marginal by other grammarians , can teach us much about language. The core，如lly generalized, unmarked constructions might be the endpoint of a continuum of generalization, they are not necessarily the only point of departure for a proper insight in the structure of the language.
The type of analysis 1 will follow for the EBNP construction is similar to the way Taylor (this volume) analyses the Bang goes my weekend! construction. A main question to be answered is , therefore , which other constructions in the language is the EBNP construction related to.
84
2.3.
AdFoolen
The constructional meaning ofthe EBNP
As pointed out in section 1.1., speakers can use the EBNP constructior. when 由ey want to show a feeling about someone or something to which they refer. Note that they do not describe that feeling , as they would do b~ saying 1 am impressed by the size 01 that man , instead of He is a bear 01 c man. The feeling is not expressed by one of the words in the cons位uction. but by the construction as such. However, ifthe language provides a choice between a neutral word and a word with an emotional connotation, the 1\ 1 position is typically filled with the connotational word: ♂ty instead of man. Dutch vent or ke时， instead of man , w旷instead of vrouw. So, instead of saying, as the Dutch reference grammar ANS does , thar the first noun must contain an affective meaning , which is then transposed to the construction as a whole , 1 would prefer to see it the other wa~ around: the construction has an expressive meaning , and this constructional meaning ‘a忧racts' words that ‘ resonate' with 由is affective meaning. Which feeling it is exactly that the speaker expresses, is not part of the constructional meaning. On the basis of the words used in the utterance. combined with properties of the textual and situational context, the hearer has to recons位uct the feeling that the speaker intended to express. There is no guarantee that this interpretive process leads to the intended outcome, as is illustrated by the following example. When 1 told a Chinese woman thar in Dutch you can say a dragon 01 a film , her spontaneous reaction was: ‘ Oh, that means that the speaker finds the film good'. The actual meaning of this expression in Dutch is , however, that the film was annoyingly sentimental. Particularly striking is an observation that has been made by several authors with regard to Det1. This can be any determiner that is allowed in a normal NP , whereby exclamative determiners like what a or such a are more frequent in this context than within other NPs. Furthermore, it is remarkable that the unmarked definite article the is ve叩 rare in this position. at least in the Romance and Germanic languages 1 looked at. It does not seem adequate to postulate a syntactic rule that stipulates a restriction on the use of the definite article here. It is not clear why the definite article does not occur in this position. Is it that a demonstrative like that gives better possibilities for a ‘四sonating' use? That can be used for 'distancing'. whereas the definite article is ‘ neutral\If there is a choic
Expressive binominal NPs in Germanic and Romance languages 2 .4.
85
Formal properties ofthe construction
All the formal properties of the construction, be they phonological, syntactic or morphological , point to a strong integration ofNl and N2. In the NP plus PP postmodifier counte甲art of the EBNP , the formal properties of the two NPs are independent of each other. The formal properties of the EBNP 由at contribute to the integration ofNl and N2 are the following: 一
一
Intonation: one overarching intonation pa忧em， a flat head pa忧em which does not focus on one of the nouns. In contrast, each NP of the normal ‘ NP ofNP' construction is stressed; it is , however, possible to stress one ofthem more than the other: a whéél of a bike versus a wheel of a bíke. Syntax: adjectival modifiers precede the first noun and prepositional modifiers follow the second noun. Modifiers immediately preceding N2 are rare , probably because this would interfere with the strong unity between Nl and N2.
(10) a. that beaut{向1 angel of a child of six years b. that beautiful six years old angel of a child c. ?that beautiful angel ofa six year old child A second syntactic prope此y that points to strong integration is that it is not possible to pronominalize the second noun, or to move the prepositional phrase by extraposition or other movement rules , as is possible in the ‘ NP ofNP' counterpart construction: (11)
Dutch: *Ik heb die schat van hem gezien. (in the EBNP reading) 1 have that darling ofhim seen b. *Ik heb die schat gezien van een kind (in the EBNP reading) 1 have that darling seen of a child c. Ik heb het wiel gezien van die fiets (in the non-EBNP reading) 1 have the wheel seen of that bike ‘ 1 have seen the wheel ofthat bike' a.
-
Morphological: Number agreement is obligatory in all languages considered. In contrast to Dutch and German, English does not allow to use the construction in the plural (c f. Dutch Schatten van kinderen , ‘ treasures of children', German Engel von Kinder ‘ angels of children').
86
Ad Foo/en
Case agreement: As was pointed out in 2.1 , N2 in German does not ha飞 E to take the dative case , as would be expected on the basis of the prep 0sition von , but instead can take over the case of the first noun. This car. only be observed in certain examples in the nominative case: diese' Lump von Angestellter ‘ this scoundrel of servant\If the first noun J~ oblique , the second noun is also oblique , but in that case you cannot see whether this is the dative triggered by von or whether there is case agreement, triggered by the first noun (see Vuillaume et a l. 1993 fo~ further details). 一 Gender agreement: in Dutch the first determiner may agree in gende with the second noun: dat dijk van een wijf ‘ that dike of a woman. meaning a strongly reliable woman) , instead of ?die dijk van een wly Wijf is neuter and takes that as the distal demonstrative pronoun. whereas dijk is non-neuter and would require die. There is , however. uncertainty on this point among native speakers , and they seem to avoiζ the construction in case of gender clash. Die wolk van een baby ‘ th二 cloud of a baby' is fine because both nouns are non-neuter, but ?Die/d.::.: schat van een kind both sound less good , probably because of the difference in gender between the two nouns. 10 一 Definiteness agreement: In Portuguese , when the first determiner :5 indefinite , then the second is null , cf. (12时， which is not different fro r:: what we see in the other Romance languages , but when Nl is definite the second is also definite , cf. (1 2b): (1 2)
Portuguese:
a. Ela é uma beleza de mulher ‘ She is a beauty of a woman'. b. Encontrei 0 estúpido do (= de + 0) teu marido ontem ‘ 1 met that idiot ofyour husband yesterday'.
The s仕ong formal integration of the two noun phrases might be seen as ζ iconic reflection of the fact that they relate to the same referent: if the refe:-ent is called an angel of a child, then the referent is both a child and (meGphorically) an ange l. In the unmarked counterpart constructionα wheel 可 J
10. Everaert (1 992: 52) elicited native speaker judgments on examples with a.:-..二 without gender clash between Nl and N2. He found considerable variation 士 judgments.
Expressive binominal NPs 切
Germanic
and Romance languages
87
bike, there are two different (indefinite) referents, al也ough 也ey are related other, and although one might even be part of the 0也er. In addition to 也.e properties just mentioned, we could 岛rmulate s严ltac tic-semantic constraints like the 岛llowing: The construction cannot occur in negated form (*no angel 01 a chil，功， or in indefinite quantified 岛m (*sommige schatten van kinderen ‘ some darlings of chi1dren'). But such cons衍aints probably follow from 由e function of the construction: If not a specific referent is involved, then expressive evaluation does not make muchsense.
ωeach
2.5.
What is the head ofthe construction?
''Though treatment [...] has been varied, depending on basic approaches , a
few aspects of why constructions like those of (1) are problematic have been recognized widely: perhaps the most prominent of those is that of beadedness. Is the first noun the head, or the second? Or perhaps both?" (Verhaar 1997: 264). The question of which part of the construction is to be considered the bead was the subject of fo町 articles that all appeared in 1997 and 1998: Leys (1997) , Verhaar (1997) , Den Dikken (1998) and Abraham (1998) , and 也.e question was also touched upon in a fifth paper, De Schutter (1997). All these authors assume that there must be a head, and only one head. The options of no head or two heads are not discussed. The au由ors agreed on the question, but 也.ey differed strongly with regard to the answer. If we diSI毡gard the first determiner for the moment,11 the other candidates, oamely the first noun，由e preposition and the second noun each had 由eir defenders. But the arguments are, in each case, not very convincing and often highly theory-dependent. The tests that are normally applied for deciding on head status within an NP, unfortunately, do not work for 也is construction: Number agreement with the verb: In the wheels 01 the bike were expensive, wheels is the head, because 也is noun determines the agreement
Jl.Everaert
(1992: 53) proposes a DP analysis, with Det1 as the head, but in the he considers N2 as the head.
∞mplement NP
88
一
Ad Foolen
with the verb. But as the 阳10 nouns in the EBNP construction must agree with each other, the test cannot differentiate between them. 12 Optionality: In the unmarked construction the wheel of a bike , the prepositional postmodifier can be left out without loss of the nominal construction, so the first noun or noun phrase is the head. But in case the prepositional phrase in the EBNP is left out, a bear of a man 今 α beαr， it is not an EBNP construction anymore , so this does not work either.
Because the traditional tests are not decisive , other criteria were used to determine the head status, mainly paraphrase and analogy. Leys (1 997) uses analogy as an argument in favor of N 1. In his view, the construction ein Bär von Mann ‘ a bear of a man' strongly resembles the construction Lehrer von Beruf ‘ teacher by profession'. In both cases , the first noun specifies the more general second noun. He calls this relation ‘ Relation des Betreffs' ,‘ relation of regarding': regarding his profession, he is a teacher, regarding ‘ man' , he is a bear. In Leys' view , the relation between NI and N2 in the two constructions is very similar. If we consider the first noun as the head in the Lehrer von Beruf case and the second noun as part of the prepositional postmodifier, then we should do that in the case of the EBNP as well. The defenders of N2 as the head use paraphrase as their main argument, cf. Quirk et al. (1 985). Besides the paraphrase argument, it is pointed out that adjectives preceding the first noun are interpreted as adjectives pertaining to the second noun (her cigar smoking devil of a husband: it is the husband who smokes cigars , not the devil). If we consider the first noun , together with the functional categories of a/von einem/van een as a kind of 硝 ective， than we get a 气latural result'. The integration of hell of a into helluva would support this view , as would the possibility in French to use an adjective in this position: ce drôle d 百omme ‘ this funny man'. De Schutter (1 997) takes a similar view, and McCawley (1 998: 765) calls this phenomenon of a ‘ noun mimicking an a司jective' ‘ syntactic mimicry'. Verhaar's proposal (1997) can be seen as a variant on this position: he considers Nl an evaluative classifier, in analogy to classifiers in classifier languages. Whereas the latter subdivide categories according to objective 12. Italian has gender agreement between subject and predicate. Here the agreement is with the second noun: Quel pezzo di una donna era belissima. Quella bestia di un professore era deciso (Pieter Seuren, p.c.). This would be an argument in favor of considering N2 as the head.
Expressive binominal NPs 切
Germanic
and Romance languagl臼
89
properties of form and function , evaluative classifiers in the EBNP subdivide 也e category ofN2 in terms of positive or negative evaluation. 13 A generative view on the EBNP construction is presented by Den Dikl osition ken (1998), cf. also Abraham (1 998). Den Dikken considers the pr叩 臼 the functional head, which takes a small clause as its complement: 01 [man bea矿. Normally, it is the subject of a small clause th创 is raised, as happens in predicational sententional constructions: is [John illJ 今 John is ρ illJ. But in the EBNP construction, we have a ‘ marked' derivation: instead ofthe subject, the predicate ofthe small clause is raised to the subject position, which leads to bear 01 [(a) man 矿. The construction is thus a case of predicate inversion, a type of movement which manifests itself also in certain constructions on the sentence level , cf. The biggest idiot in town is our doctor, where, in the generative view, the predicate is in subject position. Theoretically, this solution has some attractive features: The marked derivation accounts in a natural way for the marked status of the construction, and the predicational relation that is intuitively felt between the nouns has been given a syntactic basis. But a double price has to be paid: Firstly, instead of a monostratal analysis we have to accept a derivational analysis with more than one level of representation. One could argue 也at 由is price has to be paid anyway somewhere in the description because, intuitively, the predicational paraphrase ('the child is an angel') is ‘na阳ral'. If the predicational relation is not built into the syn细X，也en it will show up in the semantic-cognitive inte甲retation. But, one could argue , cognition provides a powerful inte甲retational appara阳s， cf. Fauconnier (1999: 96): "Language is only the tip of a spectacul町 cognitive iceberg", so, all in all , it seems attractive to let the predicational relation be p缸t of the cognitive interpretation of the cons位uction. The second disadvantage of the generative approach has to do with the preposition q卢 In the summary ofDen Dikken's analysis 1 gave above, the F叩ositional head was specified by the item oJ, but Den Dikken (1998: 190) proposes that 由is 01 is only a spell-out form of an abstract functional 臼tegory:
Just confining ourselvesωthe linear string of words found in the N 01 a N cons位uction， there is one element that 1 so far ignored in my discussion of
13. Milner (1 978: 174-197) was an early study, in which 也eNlw部 analyzed as a ‘ specifier' in relation to N2.
90
AdFoolen 也e construction but which should also be assigned a place in the final alysis-the word qfseparating the fronted predicate from its subject.Just like 也e copula be, this 伽n创 is per岛呻 meaningless. Although it looks llke a preposition, it does not behave like one.In German, fbr instance, von does not ?bliptodly entertain a Case relationship with the noun phrase following it.
So the pdce we have to pay in thiswe of analysis is accepting qf as a meaninghss padofthe conshiction-For a cogIitive linguist such a price is high (see the next section fbrhIther discussion).Meanwhiln一 1 conclude with 吨ard to the present section，也atthe literatlmdoesIn1Otle 牛ad tωo a final answ町 reg 伊 a时ing the head-s t"ioIn1.If we reshct ourselves to N1and N2 , it seems that they both have some rights to be called the head. In section 4, 1 will make 由is view more concrete.
2而.
The meaning of oland the relation between Nl and N2
"Le tour ce J坤on de valet est parmi ceux qui intègrent DE, peut-ê仕e celui qula amen6le plus de commentaims gammaticauLet6gdement le plus de d臼accords" (Englebert 1992: 40) (‘ The construction this scoundrel of ι servant is one of the constructions in which ‘ of ispartofthe COIlstmcthrL 削 perhaps the one 阳 has led to the la耶st number of gramm即正 comments, and at the same time to the most controversies. ') The meaning of the pr叩osition 01 and its counte甲arts in other languages IS not easy to pin down, probably because the meaning is rath e:general, like that ofthe copula be, the verb do or the noun thing. Langacker(1999)acknowledges that qfindicates a great vadety ofrelations between a trEUectory and a landmark, like the chiming qfbirds, a rikZ 01 gold, the state 01 Cal伪rnia， an acqaintance 01 Bill 's. 01 is thus polys二 mous. The p町Vwhole relation as in the bottom qfthejar he regards 臼丘 prototypleal use, "one with special cognitive salience,, (1999:77).As to 止主 schematic value of oJ, he makes the following proposal: "It seems qui:: accurate to describe them [the senses of oJ, A卫] as desi伊ating m intrir15:二 relationship of some kind between the Wo participants,, (1999:78)·TE= case of 01 in the EBNP construction does not occur in Lang邵阳's sur\"号 。fpossible senses , however. Stefanowitsch (1998) contrasted the use of 01 with the s-genitive, like 工 the bolt 01 the door vs. the door's bolt. He analysed a co甲us askir:豆
Expressive binominal NPs 切
Germanic
and Romance languages
91
whether there is some pa伽:m in the distribution of the two 岛rms， and if 也at is indeed the use, whether that should imply a different semantic analysis 岛r 'Ofand 泣白白 led him to postulating the Possessed/Possessor relation to be the central sense of the s-genitive. As for 'Of, he subscribes to the abs回ct meaning of intrinsic relationship as proposed by Langacker, but on 也e level of the prototypical meanings , he assumes , besides Langack町、 PartlWhole relation, an additional prototypical meaning, namely the Originary/Origin relation, which is present in cases like This ρying pan is made q所'Om steel, People die q所''Om AIDS. Historically, this meaning was 也e central one of 'Of Like Langacker, he does not take the EBNP construction into consideration. A third proposal was made by Campe (1997) , who contrasted postmodifiers introduced by German v'On ‘ of with postmodifiers in the genitive case, cf.: der Abschied v'Om Lehrer vs. der Abschied des Lehrers, both meaning ‘ the leaving of the teacher'. Like Stefanowitsch, Campe analysed a co甲us and in addition she used a questionnaire to elicit native speaker judgments. Campe came to the following conclusion (1997: 181): "The most basic semantic difference between the two altematives is thus that the genitive tends to focus on a stable, objective and intrinsic relation of interdependence between the referents of two entities, whereas the v'O n-phrase concentrates more on a [subjective] , d:严lamic， sep町ational， extrinsic relation of relative independence between the referents of two entities." In contrast to Langacker and Stefanowitsch, Campe did take the EBNP construction into consideration. She observed that the EBNP construction does not allow free variation between v'On and the genitive: The use of the genitive instead of v'On is considered bad to very bad by 95% of the native speakers she interviewed. Campe calls the type of v'On relation in this construction ‘ subcategorizing': In dieser Tr'O ttel von einem Mann ， ‘白is idiot of a man' , the speaker characterizes the man as a certain type of man. And she adds: "The use of von [in this construction] creates a distance between the two components of the relation. It is possible that by means of this linguistic distance, a c'Onceptual distance is highlighted, viz. the one between the o句éctive individual and the su句éctive (p ositive or negative) value judgment someone else passes on that individual" (1 997: 172). Intuitively, Campe's analysis of v'On fits the use
92
Ad Foolen
the premodifying genitive , like English: Peters Buch, ‘ Peter's book'). Bm as Langacker and Stefanowitsch did not take the EBNP construction int0 consideration, they should at least feel challenged by Campe' s analysis t0 ask themselves how they would accommodate the construction. Defenders of a ‘ meaningless' of might comment that this type of semantic analysis of the preposition leads to such a general , vague meaning that the step to assuming a meaningless purely formal relator is only a small one. The step in itself indeed seems a small one , but involves a principle. namely the principle of allowing for meaningless elements in linguistic analysis. Cognitive linguistics prefers not to do so , and therefore 1 propose to adopt Campe's analysis of von as the meaning of the preposition as it occurs in the EBNP construction. In this view , the preposition relates the referential meaning ofN2 to the evaluative meaning of N 1. Normally , a semantic relation of the type ‘ evaluation - reference' results in an attributive construction. The EBNP is special, in that in this case it results in a construction where the evaluative part does not have attributive status with relation to the referential part.
3. Two types of expressive binominal NPs 3.1.
Impression versus attitude
Up to now , 1 have treated all examples of the construction in the differem languages as equal. But we have seen that the instantiations of the construction do show variation on more than one point. At least the following slots can vaηin a systematic way: 一
-
一
Det1 can be definite or indefinite. Nl can be a metaphor for the referent it applies to (bear of a man, castle of a house) or an expressive word (French idiot, j坤on， German Lump Dutchfeestneus) , functioning as an insulting or endearing term with respect to the referent. A test for terms of the latter type is the exc1 amative frame X that you are! If a word can take the position of X in this frame. it should be able to function as the first noun in the EBNP construction. Det2 can be definite , indefinite or zero. In Dutch and English , it is invariably indefinite, in German it can vary between indefinite and zero. and in the Romance languages it can vary between definite and zero.
Expressive binominal NPs in Germanic and Romance languages
93
Although there is no absolute co-occurrence between the values these parameters take in the different languages , there are some correlations. The correlational pa忧em seems to be as follows: indefinite Detl - metaphoric Nl 一 indefinite Det2 , versus definite Det1 - insulting/endearing Nl - definite/null Det2. Let us call the two subtypes resulting from these correlations . Type 1 and Type 11 respectively. Type 1 seems to occur more frequently in the Germanic languages whereas Type 11 is typical for the Romance languages. Moreover, Type 1 occurs in predicate position (He is an X of a Y) and Type 11 in argument position (subject, object or other) , cf. the followìng contrast: (13)
a. He is a bear of a man. b. Cet idiot de ton j始re a oublié de me r再pondre. ‘ That idiot of your brother has forgotten to answer me' .
There seems to be a pragmatic difference too: Type 1 is used when the speaker is impressed by someone or something, as a result of a perception, Type 11 is used to express a certain feeling towards a referent, a feeling that might already exist in the speaker. As Dutch and English do not have a possible variation in the second determiner position, these languages have to use the indefinite determiner in both subtypes: (1 4)
a. Hij is een boom van een kerel. ‘ He is a tree of a man\ b. Jouw feestneus van een broer wilde deze mooie dag nodig naar Lembang toe. ‘ Your party-nose of a brother urgently wanted to go to Lembang at this nice day'. (Helga Ruebsamen , Het lied en de waarheid , p. 52)
German is an interesting test case because it is the only language with the possibility of va可ing between the indefinite article and zero in the Det2 position. This variation has been regarded by most authors as stylistic or dialectal (cf. Abraham (1998: 342) and Campe (1997 , her example 12 b)). But Vuillaume et a l. (1993) have looked at the actual dis位ibution of the two variants in language use, be it in a small co甲us， with only 70 examples. Their finding was that there is more to it than style. They discuss several examples, among which the following:
94 (15)
AdFoolen Und nun sah er wirklich Kröger auf sich zukommen , den grossen Genossen Kröger , einen Ochsen von einem Mann [...]. ‘And now he really saw Kröger approaching him, the big Kröger. the former comrade Kröger, an ox of a man [...]\
Krögel飞 den früheren
In this context ‘ ox' is c1 early used metaphorically for the big size of the man, which made a strong impression on the subject. It is thus a Type 1 example. If Det2 had been left out, the construction would have an insulting meaning: the subject calls the man an ox , which means he considers him stupid. Then it would constitute a Type 11 case.
3.2.
The built-in metaphor
A few words should be said about the metaphor that is part of the Type 1 EBNP. We should be well aware that this is not a case of conceptual metaphor. The latter type of metaphor pertains to general categories: TIME IS MONEY , LOVE IS A JOURNEY, etc. In that case , several aspects of the source domain are mapped onto the more abstract target domain. The target domain is understood in terms of the more accessible source domain. But in the EBNP case, only one aspect or dimension of the two related categories is relevant, cf. Ruiz (1998: 263): "[...] the mapping is used to single out and place in focus one aspect of the target which is , by convention, particularly relevant; the rest of the domain remains unaffected". According to Ostergard (1998: 92) , one should analyze such cases of metaphor (his example is This surgeon is a butcher, but it holds for ~ butcher of a surgeon as well) in terms of blending. Not 由e whole butcher scenario is mapped onto the target, "only selected parts of the butcher space [are] mapped into the blend" , in this case for example the lack ofprecision in cutting. This leads to the desired inference that the surgeon is incompetent. In such comparisons, the dimension , in this case cutting precision, must belong to the range of properties or relevant dimensions of both categories And at the same time the two categories should hold prototypically different values on the intended dimension: The cutting precision of the surgeon is prototypically precise, that of the butcher prototypically imprecise. In the blend we preserve the surgical scenario, but we substitute the surgical value of cutting precision with the value ofthe butcher scenario, leading to a non-
Expressive binominal NPs in Germanic and Romance languages
95
prototypical, extreme value in this dimension for the surgeon scenario, which is suggested to hold 如r the referent we are talking about. Note that the relevant dimension is not mentioned explicitly in the construction，也e hearer must reconstruct which quality or dimension the speaker has in mind. So, it is not simply a calculus of semantic features 也at is involved in the interpretational confrontation between the two nouns, it is a whole complex of experience that 扭曲e input for the blending process. The construction is a guide in the interpretational process, insofar as it requires an outcome of a certain type: a quality of the referent must come out as representing an extreme case. If a man is compared to a tree, a possible outcome pe此ains to extreme size, tallness in particular, and it is indeed this outcome 也at is the conventional meaning of Dutch een boom van een kerel ‘a 位'ee ofa man'. But false interpretation and misunderstanding are quite well possible. If you don 't know what English speakers conventionally mean by a bear 01 a man you have to deduce 仕om the context whether personality, physical size or a strong appetite for sweets is meant. This type of analysis seems to be adequate 岛r metaphors as they occur in this surgeon is a butcher as well as in the EBNP construction a butcher 01 a surgeon , but also for certain adjectival compounds. Consider Dutch Dit laken is sneeuwwit ‘This blanket is snow white'. In也is case, the relevant dimension, namely ‘ white' , is explicitly mentioned in the construction, which makes the interpretation less of a puzzle. The ex创nple implies that, normally, the value for blankets on the scale ofwhiteness is less high than the one for snow. And the ac阳al blanket we are talking about is exceptional in that its value 扭曲is dimension takes an extreme value, a value 由at is 也e normal value ofwhiteness for snow. 4. EBNP, PLAC, and DRNA With the help ofthe studies by Ross (1998) and Malchukov (2000) , we can put the EBNP into a wider perspective. Ross shows 由at in different Oceanic languages of Northwest Melanesia 由e adjective-noun construction does not have the ‘ normal' form with the noun as the head. Instead, the relation between at往ibute and noun is constructed in analogy to 由e possessive construction a wheel 01 a bike. This leads to constructions like: a big one 01 a house. Ross calls this phenomenon 'Possessive-like attribute conS住uction' ， PLAC for short. This construction has also been a岱ested in
96
AdFoolen
Hausa , Aleut, and some Tibeto-Burman languages. Ross discusses extensively the difficulty of determining the head of the construction. His conclusion is that in some of the Oceanic languages the attribute is the head , whereas in other Oceanic languages the second noun has gained head status. It seems that diachronically , a process of head shift is taking place, leading to ‘ normal' attribute-nominal head constructions. How did the Oceanic languages develop the marked PLAC construction in the first place? Ross hypothesizes that Proto-Oceanic had an EBNP construction (he calls it a whopper construction, from his example a whopper 01 a house) (p. 271): "[T]here was a (marked) attributive structure, a whopper construction that was pattemed on a possessive-construction and that had a marked discourse function." This construction functioned as a kind of ‘ catalysator' for the development of PLACs. Ross explains the head shift occurring later as follows (p. 273): "Once [PLACs] became the basic attribute construction, it seems to have been a natural process for speakers to reinterpret the referring element (the modified noun) as the head and the attribute as a modifier, doing away with the mismatch." By ‘ mismatch' he means the discrepancy between the syntactic and the notional side of the constructlon. Independently ofRoss's work , Malchukov (2000) discovered pa忧emsof this type in other languages (Tungusic languages , Aleut , Gude , Chinook). His name for the construction is DRNA (Dependency Reversal in NounAttribute constructions). Malchukov notes that in languages where the DRNA exists besides a ‘ normal' attributive construction, the DRNA is used to give the attribute high "discourse-pragmatic salience" (p. 45). As a general explanation, he invokes "the notion of iconicity or rather isomorphism between pragmatic and syntactic structures: a pragmatically salient NPconstituent tends to be syntactically salient as well , that is , tends to reveal head properties." (p. 49). This view is particularly helpful in relation to the EBNP construction The ‘ upgrading' of the attribute to a noun which is the head of the construction can be understood against this background. Without the upgrading , the attribute (for example in ‘ an angel of a child') would surface in an adjective ('angelic' or in a adverbial constituent ‘ like an angel'). With the upgrading , there is the further advantage that the attribute c
Expressive binominal NPs in Germanic and Romance languages
97
‘ head shift'. We should not exclude the possibility that the different Euro-
pean languages we discussed differ with regard to the position they have on this ‘ cline\And even within a language, not all instantiations necessarily take the same position on the cline, cf. the English helluva job case, which seems to be an example in which the head shift has progressed to near completion. It is well known that expressive constructions are diachronically ‘ weak' , i.e. they lose their expressive force , which creates the need for new expressive forms. The life cycle of expressive forms is shorter than that of representational forms. Head shift might be simply a symptom of the loss of expressivity. The construction ‘ slides back' into normal patteming. In future research , we should look at the different European languages again and try to determine which place the construction in the different languages takes on the cline. Such an approach might resolve the difficulties that have tumed up again and again with regard to the headedness questíon.
s.
Conclusions
1. The general approach of Construction Grammar is a fruitful framework for the analysis of the expressive binominal construction. At the same time , the findings can be seen as support for a constructional view on grammar. 2. Two subtypes of the EBNP can be distinguished , the ‘ impressive' one , and the evaluative one. They differ lexically (first noun) and syntactically (the determiners). 3. We did not find an unequivocal answer to the question ofwhich element in the EBNP construction should be considered the head of the construction. The head properties seem to be distributed over the two NPs. Further research is necessary on this point, however. 4. The EBNP construction is motivated in the iconic sense: the salience of the expressive meaning is ‘ mirrored' in the salience given to the notionally attributive p町t of the construction. 5. With regard to the ‘ ecology' ofthe construction, the EBNP is motivated by the possessive NP-of-NP construction and by the attributive A-N construction. This double motivation leads to diachronic instability, the latter construction being the stronger ‘ attractor' in the long run.
98
AdFoo/en
References i\braharn, VVerner
1998
Ein Schatz von einem Kind. Zur Prädikatsyntax binominaler Nominalkonstituenten. Deutsche Sprache 26: 337-347.
i\NS
1997
A/gemene Neder/an伽年raa胁nst. VV. Haeseryn et al. (redaction). Groningen: Nijhoff. Deurne: Plantyn. 2th revised edition.
Besnier, Nico 1990 Language and affect. Annua/ Review 01 Anthrop%gy 19: 41945 1. Bühler, Karl 1934 Sprachtheorie. Jena: Fischer. Caffi, Claudia and Ri chard VV. Janney (eds.) 1994 Involvement in language. Special issue of Journa/ 01 Pragmatics 22: 3-4. Campe, Petra 1997 Genitives and von-Datives in German: i\ case oflree variation? In: M. Verspoor et al. (eds.) , Lexical and 砂ntactical Constructions and the Construction 01 Meaning, 165-185. i\msterdam: Benjamins Den Dikken, Marcel 1998 Predicate inversion in DP. In: A i\lexiadou and C. VVilder (eds.) , Movement in the Determiner Phrase, 177-214. i\msterdam: Benjamins. De Schutter, G. 1997 The Noun Phrase in Dutch. Leuvense Bijdragen 86: 309-356. Duden 1995 Grammatik der deutschen Gegenwarts，早 rache. Mannheim: Bibliographisches Institut. Englebert, J\nnick 1992 Le 'p etit mot' DE. Genève-Paris: Librairie Droz. Everaert, Martin 1992 Nogmaals: ‘ Een schat van een kind'. In: H. Bennis and J.VV. de Vries (eds.) , De binnenbouw van het Neder/ands. Een bundel artike/en voor Piet Paardekooper, 45-54. Dordrecht: Foris. Fauconnier, Gilles 1999 Methods and generalization. In: Th. Janssen and G. Redeker, (eds.) Cognitive Linguistics: Foundations, Scope, and Methodo/ogy, 95-127. Berlin: Mouton de Gruyter.
E再pressive
binominal NPs 切
Germanic
and Romance languages
Foolen, Ad 1997
The expressive function of language: Towards a cognitive semantic approach. In: Susanne Niemeier and René Dirven (eds.) , 1页e Language 01 Emotions , 15-31. Amsterdam: Ber飞jamins. Goldberg, Adele
1995
Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press.
Langacker, Ronald
1999 Leys, Odo 1997
Grammar and Conceptualization. Ber1in: Mouton de Gruyter. "‘ Ein Engel von (einer) Frau'. Emotionalität als konstruktionale Bedeutung". Leuvense Bijdragen 86: 27-52.
Malchukov, Andr吗 L
2000
D甲endency Reversal in Noun-attributive Constructions: pology. München: Lincom.
A η
McCawley, James D. 1998 Th e Syntactic Phenomena 01 English. Second edition. ChicagolLondon: The University of Chicago Press. Michaelis, Laura A. and Knud Lambrecht 1996 a The exclamative sentence type in English. In: A. Goldberg (ed.), Conceptual Strucωre， Discourse and Language, 375-389. Stanfo时， CA: CSLI Publications. 1996 b Toward a construction-based theory of language function:ηle case of nominal ex位aposition. Language 72: 215-247. Milner, Jean-Claude
1978
De la 砂ntaxe à 1'interprétation: Quantités, Insultes, Exclamations. Paris: 剧itions du Seuil.
Ostergard, Svend 1998 Content and speaker's viewpoint. Versus 80/81: 79-98. Paardekooper, P.C. 1956 Een schat van een kind. De Nieuwe Taalgids 49: 93-99. Qu尬， Randolph et a l.
01 the English Language. London: Longman. Ruiz de Mendoza Ib姐钮， Francisco José On 也e nature of blending as a cognitive phenomenon. Journal 01 1998 Pragmatics 30: 259-274. Ross, Malcolm 1998 Possessive-like attributive constructions in the oceanic languages ofNorthwest Melanesia. Oceanic Linguistics 37: 234一276. 1985
A Comprehensive Grammar
99
100 Ad Foolen Stefanowitsch, Anatol 1998 Possession and partition: The two genitives 01 English. Hamburg: Seminar für Englische Sprache und Kultur, Universität Hamburg. Taylor, John this vo1. The ecology of constructions. Verh嗣r， John W.M. 1997 Evaluative nominal c1assifiers in Dutch. Leuvense Bijdragen 86: 263-290. Vuil1aume , Marcel , Jean-François Marillier and Irmtraud Behr 1993 Dieser Schuft von einem Hausmeister. Uber1 egungen zu den morphosyntaktischen und semantischen Eigenschaften von Strukturen nach dem Muster ''N-von-N''. In: M. Vuil1aume, J.-F. Mari1lier and 1. Behr (eds.) , Studien zur 砂ntax und Semantik der Nominalgruppe, 167一 184. Tübingen: Gunter Narr.
Section 11 Genetic motivation
On genetic motivation in grammar BerndHeine
Abstract Comparative studies in grammaticalization suggest 也就 it is possible to reconstruct some of the motivations underlying the creation of grammatical categories. The paper summarizes recent fmdings 也at have been made within this paradigm (see Heine 1997b; Heine and Kuteva 2002 for more details). By looking crosslinguistically at three di fIerent conceptual domains , it is 町'gued 由at grammaticalization theory is able to account for some of the properties characterizing linguistic expressions for 也.ese domains with reference to 也e cognitive forces 也at can be held responsible for the genesis and evolution of grammatical categories. The domains studied are numerals, indefinite reference , and predicative possession.
Keywords: Action Schema, Companion Schema, concretization,
contex阳
alization, genetic motivation, Goal Schema, grammaticalization, indefmite article, Location Schema, numerals , possession, Source Schema, structural motivation
1.
Iotroduc创00
The term motivation appears in a number of di fIerent ways in the linguistic literature of the past decades. For some it is a key notion for understanding Ianguage struc阳re， for others again it is an auxiliary term used to make weak, rather 由an strong, claims about observed facts , or to refer to some-
104 Bernd Heine
thing that lacks theoretical or explanatory status. I Among the many views that have been voiced on the notion motivation , there are in particular two contrasting ones. The first can be sketched as in (1). (1)
Something is motivated only if it can be shown not to be arbiIn this view, language structure is essentially arbitrary. Motivated structures constitute special cases, absence of motivation being the expected case. tra可.
The relationship between the linguistic structures A and B, or between linguistic structure A and an extra-linguistic structure B, is motivated if both share properties whose presence is not coincidental or arbitrary. Motivation, as understood in this view, has been referred to as structural motivation (Heine 1997b: 5). Saussure (1922: 31) was arguing in terms ofstructural motivation when he noted that forms such as French dix-neuf ‘ 19' 的 "relatively motivated" since they are syntactically related to other forms having the same pa忧ern. and each component of this 岛口n is related to some other component. Similar positions have been voiced time and again ever since. For example. Wierzbicka's (1985: 334f.) understanding ofsemantic motivation is ofthis kind when she observes that "the system of formal distinctions and the system of conceptual distinctions are mutually isomorphic." Structural motivation is invoked typically to draw attention to systematic correlations between different structures even if no hypothesis to explain the presence of such correlations is volunteered. 2 But 1 wil1 also talk of structural motivation if there is some account whose exact explanatoη significance remains unclear. Lyons (1977), for example, defines Saussure's term motivated as "any non-arbitrary form-meaning or meaningmeaning relationship that can be shown to be based on some general principle" (Lyons 1977: 105). Lyons does not elaborate on what kind ofprinci-
1. Hopper and Traugott (1993: 63), for example, "speak more cautiously ofmoti-
vations or enabling factors" , rather than of"causes" or "explanations". 2. One may wonder whether in such cases one is really justified to talk of "motivation". This is an issue that is not further pursued here.
On genetic motivation in grammar 105
ples he has in mind; hence , 1 will take this definition to be suggestive of structural motivation. 3 The second view of motivation can be sketched as in (2): (2)
Human behavior is not arbitrary but is driven by motivations. Language structure, being one of the products of human behavior, therefore must also be motivated. lf we find "arbitrary" language structures then these are in need of explanation.
1 refer to the kind of motivation alluded to in (2) as genetic motivation (Heine 1997b). According to (2) , present-day human behavior is the resu1t of past motivations. Genetic motivation thus is a diachronic notion , and it is explanatory: lt accounts for the fact that structures A and B are causally related in that B can be historically derived from A , or that both A and B can be derived 仕om C , by means of a set of identifiable motivating forces. Thus , genetic motivation can be described typically by means of formula (3): (3)
A + motivating force X > B , where A = a linguistic form or conS位uction
(1) and (2) are contrasting positions, and they have given rise to contrasting ways of looking at language and, consequently, also of doing linguistics. They are , however, in no way incompatible with one another. Take for example the much-discussed notion iconicity. ln many instances where this notion has been invoked, the reader is left guessing what the rationale underlying this notion may be or may have been: Not infrequently, terms such as iconicity or isomorphism appear to be hardly more than labels for observed correlations，出at is , they stand for what 1 refer to as structural motivation. But for some authors (cf. Haiman 1985a, 1985b; Givón 1990; Croft 1990; Kuteva 1994), they may also be suggestive of genetic motivation, e.g. , when iconicity is used as an explanatory concept to account for s住时' tural correlations e.g. between certain linguistic and cognitive, or any other extra-linguistic, phenomena.
3. The only example given by Lyons concems the relationship between what is assumed to be the central meaning of a lexeme and "its subsequent more general meaning or meanings" (Lyons 1977: 264).
106 Bernd Heine
In a similar fashion , observations on economic motivation may be phrased in terms of structural motivation , e.g. , by saying that 企equent1: used linguistic expressions are shorter than less frequently used ones. Bm economy can also be looked at in terms of genetic motivation , e.g. , whe r. observing that people will shorten the linguistic expressions that are useè most commonly for economy (Haiman 1985a; Croft 1990: 156). That both views , (1) and (2) , are jointly re1evant to understand 1anguage structure is argued for perhaps most pronouncedly by Giv∞(1 985 ， 19901 Language structure, and syntactic structure in particular, Givón observes. are partly motivated (iconic) and partly arbitrary: "[...] transparent iconic principles interact with more arbitrary ‘ symbolic' elements in making up syntactic structure." (Givón 1990: 968). My interest in this paper is ex c1 usively with genetic motivation , that is , 1 will be concerned essentially with substantiating the relevance of the view sketched in (2). More narrowly , 1 will be dealing with findings on the evolution of grammatical forms , in short, with grammaticalization. The position maintained here is a strong one , it is based on the observations summarized in (4). (4)
a. There is evidence to suggest that some grammatical forms are (genetically) motivated. b. There is no convincing evidence to suggest that grammatical forms are unmotivated, e.g. , that they are invented arbitrarily. c. In view of (4a) and (4b) , the most plausible position would seem to be one according to which grammatical forms are motivated (rather than arbitrary).
Adopting the position sketched in (4c) means that the burden is on me to account for seemingly unmotivated grammatical forms and structures. Support for this position comes from a rich literature on grammaticalization processes that has accumulated over the last two centuries , in particular i口 the course of the last two decades (see e.g. Traugott and Heine 1991: Heine , Claudi and Hünnemeyer 1991; Hopper and Traugott 1993; Bybee. Perkins and Pagliuca 1994; Giacalone Ramat and Hopper 1998; Heine and Kuteva 2002). The motivating principles proposed here are (a) concretization and (b) contextualization. Concretization serves to describe meanings that are more difficult to understand in terms of expressions for things that are easier to understand. A paradigm example of concretization is expressing grammatical meanings by means of items used for lexical meanings Contextualization serves to structure texts in such a way that they represem
On genetic motivation in grammar 107
best the speaker's intentions and are easy for the hearer to decode. For example, using spatial or temporal concepts to express relations between diι ferent parts of a text is a common contextualization strategy, serving to create or improve text coherence. There are two main ways to demonstrate that the position sketched in . (4) is empirically justified. One would be to present some maximally convincing example and to argue that other examples that one can think of can be expected to behave the same way. An alternative way, the one adopted here , is to offer as many examples as possible within the limited scope of a paper. Accordingly, the following discussion will be a kind of tour de force , presenting generalizations on many languages without offering appropriate evidence. 1 will be dealing with three different kinds of linguistic categories in a cross-linguistic perspective. Evidence supporting the generalizations proposed here can be found in Heine (1997a, 1997b). The categories looked at are cardinal numerals (section 2) , indefinite articles (section 匀， and predicative possession (section 4). It goes without saying that no more than a few selected properties of the relevant categories can be examined in the subsequent overview. This paper will take the following format: First, some crosslinguistic properties of the category concerned are highlighted (cf. Dl , D2 , etc. below)4. Second, the experiential basis and cognitive forces that can be assumed to be causally, that is , diachronically, responsible for these properties are sketched (cf. Ml , M2 , etc. beloW)5.
2. Numerals 1 will confine myself to cardinal numerals used in attributive construction, as for example ‘ two' in ‘ two apples' , that is , ordinal numerals or numerals as used in counting are not considered. On the basis of crosslinguistic findings (see especially Greenberg 1978a; Heine 1997b), the following general observations can be made:
4. "D" is a shorthand for "descriptive statement". 5. "M" stands for "hypothesized motivating force".
108 Bernd Heine
(5)
Cardinal numerals D 1: D2: D3: D4: D5:
In the languages of the world, numeral systems having ‘ 5' ,‘ 10' , or ‘ 20' as their numeral base are statistically predominant. Ofthese,‘ 10' appears to be the most widespread numeral base. Numerals for 呀， or ‘ 10' not infrequently have nominal characterístícs. Numerals from 飞， to ‘ 9' have a propositional , clause-like struc阳re in a number of languages. Higher numerals, like ‘ 100' and ‘ 1000' , tend to be nominal in S位ucture.
D6:
Markers used for the arithmetical operation of addition (i.e. , ‘ plus') in complex numerals frequently resemble function words used 岛r the expression of either comitative ('with') or locative participants (‘ on， upon丁
It would seem that generalizations like the ones proposed in (5) can be related to the following observations: Ml: The human hand provides the most salient model for structuring the numeral system. The numeral ‘ five' therefore crosslinguistically constitutes the smallest recurrent base number (from which counting st町ts over). Hence, the most common structure is one where 丁， is derived 仕om the word for ‘ hand' ，‘ 10' 企om ‘ two hands' , and ‘ 20' from either something like ‘ hands and feet' or ‘ whole person\This is likely to account for the fact that numeral systems having ‘ 5' ,‘ 10' , or ‘ 20' as their numeral base are statistically predominant. M2: ‘ Hand' provides a far more widespread template than ‘ foot'. There are many languages 也at have ‘ hand' but not ‘ foo t' figuring in numerals while the opposite appears to be rare. Furthermore, the percep阳al difference is larger between hands and feet than between one hand and another. It would seem that this may account for the fact that there are more languages that have ‘ 10' as their base number than languages having either 哇， and/or ‘ 20\ M3: ‘ 5' and ‘ 10' tend to be derived from nouns for ‘ hand\Even if these numerals no longer have any nominal meaning , they may have retained some morphosyntactic relics ofnouns. Accordingly,
On genetic motivation in grammar 109
we may not be surprised to find languages where numerals for ‘ 5' or ‘ 10' exhibit some nominal characteristics. M4: Numerals from 飞， to ‘ 9' are likely to be created as predications about fingers and hands: They tend to refer to individual fingers and to be expressed by means of propositional structures having predications like ‘ take the index finger off ,‘ add the big finger' , ‘ put the thumb on top of X' ,‘jump from one hand to the other' as a source. It may therefore not be surprising that numerals 仕om 飞' to ‘ 9' not seldom have a propositional , clause-like structure. M5: Wh ile the body-part model accounts for a number of linguistic features characterizing lower numerals , the situation is different when higher numerals such as ‘ 100' and ‘ 1000' are involved, occasionally also numerals from ‘ 20' onwards. One reason may be that higher numerals are more likely to be borrowed and to be treated as loan nouns. Perhaps more commonly, however, an altemative model is made use of, whereby inanimate collective items , i.e. nouns meaning 吁1eap' ，‘ group' ，‘ bundle' ,‘ hair' , etc. , are used as terms for higher base numbers. In accordance with their conceptual source, higher numerals , like ‘ 100' or ‘ 1000' , therefore tend to be nominal in structure. M6: That markers used for the arithmetical operation of addition in complex numerals 仕equently resemble comitative or locative markers can be related to the fact that arithmetical operations are fairly abstract. We will therefore expect that their linguistic expression will draw on more concrete concepts. Crosslinguistically, location ('on top, above') and accompaniment ('together with') are among the most common means of encoding more abstract relations; it is therefore not surprising that they tend to be recruited to also express the concept of arithmetical addition. It goes without saying that not all languages conform to the generalizations just proposed. For example, there are languages that have 丁， (tem町y) ，吨' (quatemary) ,‘6' (senary) , or ‘ 9' (nonary) as a numeral base, and there are also languages using ‘ 12' , rather than ‘ 10' , as their base. In most cases, such "deviant" situations can be accounted for in terms of altemative strategies (see Heine 1997b: 24ff.). One strategy is to take the highest numeral distinguished as a numeral base. If the highest numeral is ‘ 6' , then ‘7' would be rendered as 飞+ 1' , etc. An other strategy involves the reanalysis of an existing numeral as some other numeral. Among the Mata-
110 Bernd Heine
pato Maasai on the northem slopes of Mt. Kilimandjaro , all decade value~ have been doubled - that is, the numerals tomon and tíkítam mean , not ‘ 10' and ‘ 20' , as they do in other Maasai dialects , but rather ‘ 20' and ‘ 40'. Conceivably this can be related to the fact 由at the Matapato are more strongl~ exposed to tourism than other Maasai sections. The overall crosslinguistic analysis of cardinal numerals leads to the following more general conclusions: (i)
There are many numerals whose motivation is opaque rather thar: transparent. (ii) Nevertheless, 1 am not aware of a single case where a numeral h~ been invented arbitrarily. (iii)ηlere 町'e many examples suggesting that the structure andlor the form numerals take are motivated (in the way sketched in the present section). (iv) The most obvious conclusion therefore is that numerals are motivated, and that in cases where no motivation has been found S0 far, this is due to our still insufficient knowledge of their evolutI on.
3. Indefinite articles In English, as in a number of other languages, there is an asymmet叩 in use between definite and indefinite articles: One can utter (6时， (6b) , and (7al. but not (7b) - that is , the indefinite article may determine singular nOUll5 but not plural nouns: English (6)
a. 1 see the child. b. 1 see the children.
(7)
a. 1 see a child. b. *1 see a children.
The question that now arises is: What accounts why is (7b) not an acceptable utterance?
岛r
this asymmetry , that is.
On genetic motivation in grammar 111
To begin with , let us say that indefinite articles are nominal detenniners whose functions include that of marking indefinite specific reference , where the referent of the noun phrase is identifiable for the speaker but is presented in such a way that it is left unidentified for the hearer. Indefinite articles may be independent words , particles, clitics, or affixes; they may be segmental or suprasegmental. The following is a catalogue of structural properties that crosslinguistically tend to be associated with indefinite articles (Heine 1997b: 68ff上 (8)
Indefinite articles D 1:
D2: D3:
D4:
In many genetically and areally unrelated languages , indefinite articles have the same or a similar fonn as the numeral ‘ one' , and both are likely to employ the same position in the clause. Ifthey are similar to the numeral ‘ one' , not infrequentI y indefinite articles are phonologically shorter an d/or less complex. They tend to be confined to detennining the singular of count nouns. In this respect, they differ from definite articles , which do not show such a constraint. If they detennine mass nouns, then they also detennine plural nouns. If they detennine plural nouns , they also detennine singular nouns , but not the other way round.
In a survey of 31 languages having indefinite articles , such an article was found to be used in 81 % of singular nouns , 23% of plural nouns , and only 10% of mass nouns (Heine 1997b: 68). Definite articles do not show structural correlations of this kind. Anya忧empt to explain these properties must take account of the historical development of the articles in question, in particular of the following crosslinguistic observations: Ml: Most indefinite articles for which sufficient evidence is available are derived from the numeral ‘ one'. Hence they are likely to exhibit much the same mo甲hosyntactic properties as the numeral , such as occurring in the same syntactic slot. M2: Once they are pressed into service to express indefinite reference, they are likely to undergo canonical grammaticalization processes: (a) They tend to be eroded , that is , to be phonologically reduced.
112 Bernd Heine
(b) This entails that they become phonologically simplified, e.g by losing the ability to carry stress. (c) They increasingly become dependent in their occurrence or. the noun they determine , to the extent that they cannot occur on their own , e.g. , (9) English a. 1 see one. b. *1 see a. The result is that indefinite articles , while still resembling the numera: ‘ one' , tend to become increasingly different from the numeral , by losing ir. phonetic substance and in morphosyntactic autonomy. M3:
M4:
The numeral ‘ one' inherently determines singular count nouns and is likely to retain this property even in its grammaticalizeé uses as an indefinite article. For example , the English indefinitë article (a)n can be traced back to the numeral one. Obviously. numerals for ‘ one' are inappropriate as modifiers of plural nouns (e.g. *one children). Although a(n) is no longer a numeral , the semantic property of incompatibility with plural head nouns has survived its development into an indefinite article. For obvious reasons , such constraints were absent in the genesis of the Englist definite article: Demonstratives , from which it is derived, do no: show any number-specific constraints of this kind. In the course of time , however, when the indefinite article becomes increasingly independent of its lexical source (the m;meral) , it may happen that the use of the article is extended tc non-singular count nouns , and eventually to mass nouns. This means that there are languages where the indefinite article sprea d5 to plural nouns. Catalan , Spanish, and Portuguese are languages where such a spread has occurred, that is , where the use 0: uno/una, etc. came to be extended to plural nouns. In the end , the indefinite article may spread to all kinds of nouns , including mass nouns.
Considering this sequence of evolution, we will expect that all language三 having an indefinite article derived from ‘ one' , use it with singular nouns.
On genetic motivation in grammar 113
while a few languages also use it with plural nouns , and the smallest number of languages have reached the stage where indefinite arti c1 es can also be used with mass nouns. Hence , there is an implicational scale as sketched above (D4): If indefinite arti c1 es determine mass nouns , then they are also used for plural nouns. If they determine plural nouns, then they also determine singular nouns , but not the other way round. Predictably, definite arti c1 es lack such properties. To conc1ude , in a similar way as we saw in the case of numerals , the following may be observed for indefinite arti c1 es: There are some cases where no motivating factors have been identified so far. There is no evidence to suggest that an indefinite arti c1 e has ever been invented arbitrarily. On the other hand, in most cases studied so far , the structure and/or the form indefinite arti c1 es take can be shown to be motivated. The most obvious conc1usion therefore is that indefinite arti c1 es are motivated, and that in cases where no motivation has been found so f泣， this is due to our sti11 insufficient knowledge.
4. Possession A more complex situation is found when one looks e.g. at linguistic expressions for possession. All research that has been carried out in the course of the past decades suggests that possession is a cross-culturally stable concept. This means , for example, that in all1anguages for which a reasonably detailed description exists there are conventionalized forms to express predications such as ‘ 1 have a dog'. At the same time , the term possession refers to a number of different structures and contents. First, there is a distinction between attributive possession (my dog) and predicative possession (1 have a dog). Second, there is also a distinction between belong-constructions (The dog belongs ω me or The dog is mine) and have-constructions (1 have a dog). The main , though not the only, diι ference between these two is that in the former construction the possessee is definite while in the latter construction it is 叩pically indefinite (cf. Seiler 1983). Third, the term possession inc1udes a range of different meanings (see Heine 1997a). Perhaps the most salient ones are the following (the English examples added in parentheses, involving various uses of have, are meant to illustrate the respective notions):
114 Bernd Heine
Physical possession Temporary possession
(1 want to fill in this form; do you have a pen?) (1 have a car thαt 1 use ω go to the office but it
Permanent possession Inalienable possession Abstract possession Inanimate possession
belongs to Judy) (Judy has a car but 1 use it all the time) (1 have blue eyes) (He has no time/no mercy) (My study has three windows)
My concern here is exclusively with expressions for predicative possession , and here again with have-constructions , that is, with expressions coπe sponding e.g. to English 1 have a dog or 1 have two sisters. Consider the following crosslinguistic observations (for more details. see Heine 1997a):
(1 0) Possession Dl:
D2: D3:
There is no universal linguistic structure common to all possessive constructions. For example, the possessor may be encoded as the clausal subject, a locative phrase, a dative adjunct, or a genitival modifier, and the possessee as the object, the subject, or as an adjunct. Possessive cons仕uctions are likely to also express concepts other than possessive ones. The way possessive relations are encoded not seldom con衍adicts some cross-linguistic generalizations on iconicity, according to which human participants are more strongly associated with subject marking , while objects and adverbial phrases are more likel) to be associated with inanimate participants; cf. the Latin construction exemplified in (11):
(11) Latin Mihi liber est. to:me book is ‘ I have a book. '
A crosslinguistic study suggests that have-constructions are almost invariably derived from any ofthe source schemas listed in Table 1.
On genetic motivation in grammar 115 Table 1. A fonnulaic description of schemas used as sources for have-possession
(see Heine 1997a for details) F'ormu/a X takes/gets/holds Y Y is located at X Y exists for/to X X'sYe况ists
X is with Y As for X , Y exists
Source Schema Action Lοcation
Goal Genitive Companion E匹
One may wonder why possession should be derived 仕om concepts that bear no immediate relationship to possession, that is, from concepts relating to what one does (Action), where one is located (Location), who one is accompanied by (Companion) , etc. There is an obvious answer: Possession is a fairly abs位act concept; a stolen apple is hard to distinguish from one that is not (Miller and Johnson-Laird 1976). It is much easier to describe where a given item is, what happens to it, or who it is accompanied by, e.g. by using body movements. Thus, action, location, accompaniment, etc. provide convenient structural templates to express possession: An item that 1 have taken or received, that is habitually located where 1 am, or that 1 am regularly accompanied by may be interpreted as belonging to me. In the course of time, a linguistic expression defining an action, a location, or a companion may then become conventionalized as a possessive expression. Each of the schemas listed in Table 1 is associated with a particular structure of morphosyntactic encoding. The kinds of associations that are most likely to be expected are presented in Table 2 (see Heine 1997a). Table 2.
Typical participant encoding in have-constructions according to source schema
S'ource Schema Action Location Goal Genitive Companion Topic
P'ossessor (=刀
P 'ossessee
Subject
Object Subject Subject Subject Comitative adjunct Subject
Lοcative
complement Dative adjunct Genitive modifier Subject Theme, subject
(=η
116 Bernd Heine
We can now try to account for the structural characteristics of possessio:: mentioned earlier: Ml:
Each of the source schemas listed in Table 2 is associated with :. different kind of participant encoding. Languages drawing on th;; Action Schema, like most languages of Western Europe , ar;; likely to develop a transitive construction where the possessor 1 encoded as the subject and the possessee as the object of tL clause. Conversely, in languages that have recruited the Goa: Schema, the possessor will appear as a dative or related particlpant while it is the possessee that forms the subject. Obviousl:. there cannot be a universal structure of morphological or syntactj~ marking of possession. M2: Perhaps one of the most confusing experiences linguists dealin; with possession are confronted with is that possessive constructions are also used for meanings that clearly are non-possessive 1三 nature. Essentially there are two kinds of processes that can be held responsible for this fact. First, as we noted earlier, expre~ sions for possession are historically derived from constructio匹 denoting non-possessive functions , like action , location , and accompaniment. Not infrequently, the source meaning survives i~_ certain contexts. It is therefore to be expected that one and the same expression is used, for example , for location and possession
,
The second kind of process has to do with the fact that possessive co r.structions, while being derived from expressions for non-possessive meanings , may themselves give rise to other kinds of non-possessive meanings. like marking verbal aspect and tense. European languages offer ample e\idence for such an evolution: In most languages of Western Europe , Possessive Schemas of the form X has/owns Y have given rise to tense-aspe c: categories such as perfect, perfective , and past (X has done η ， or fu阳re \- ì.has to do Y; see Heine and Kuteva 2002 for details). Furthermore , this schema may also lead to the emergence of other grammatical concepts. such as markers of existence, equation, etc. A few salient evolutions ir.volving possessive constructions are presented in Table 3.
On genetic motivation in grammar 117 Table 3. Domains of conceptual expansion involving possession (cf. Heine 1997a:207)
Action, Location, etc.
>
Possession
>
Tense-Aspect, Existence, E,quation , etc.
M3: In languages where have-constructions are derived 企'Om the Location or Goal Schemas, the (typical1y inanimate) possessee is likely 阳 be enc'Oded 臼 the subject and the (typical1y human) possessor as either a locative or dative participant. Thus the expected correlation between animacy and case encoding is violated in such cases.
s.
Conclusions
In the preceding discussion, a number of factors were ignored. First, 1 ignored the fact that the notion motivation has quite different meanings, depending on the perspective one wishes to adopt vis-à-vis grammar. 1 was arguing exclusively in terms of genetic motivation; it goes without saying 由at this is not the only position one may adopt. Second, my contribution rested on a crosslinguistic review of three different categories only; the reader is referred to further works on grammaticalization dealing with other grammatical categories (see e.g. Haspelmath 1990 on passives; Heine 1997b on domains such as spatial orientation and comp町ison; Bybee, Perkins and Pagliuca 1994 'On tense , aspect, and modality; for a more general account, see also Heine and Kuteva 2002). On the basis of the observations made, the 岛110wing c'Onclusions can be drawn: (a)
Quite a number of properties characterizing numerals, indefinite 町ticles，
or possessive constructions can be accounted for satisfactorily only by relating such functional domains to constructions for semantic contents belonging to other functional domains. (b)咀lis means that a synchronic explanation on the basis of, say , S严ltactic or semantic facts may be redundant once it has already been shown 由at such observations can be explained exhaustively with reference to their evolution. On the basis of a purely syn-
118 Bernd Heine
chronic analysis it would be hard to account for structural properties such as those described in (5), (8), and (10). It would therefore seem most economical to look for motivations in terms of diachronic evolution before embarking on synchronic (or ahistorical) explanations. One might argue that the situation described here is exceptiona l. that is , that more often than not, no motivation can be reconstructed. In such cases, either of the following stances can be taken:
(c)
(d)
(1) (11)
We have to look for more information in order to reconstruct the kind of motivation involved. Since no motivation has been found , there is no motivation.
As 1 hope to have shown , alternative (1) is the one to be adopted: In a number of cases, at a first glance no motivation could be found, yet after a more detailed analysis , combining historical findings with crosslinguistic generalizations , it turned out possible to reconstruct the motivation involved Accordingly , the second alternative (11) would offer a less plausible solution; a conclusion like "1 cannot see any motivation, hence, there is no motivation" , runs the risk ofturning ignorance into a scientific dogma.
References Bybee, Joan , Revere Perkins and William Pagliuca 1994
The Evolution 01 Grammar: Tense, Aspect and Modality in the Languages 01 the World. ChicagolLondon: University of Chi-
cago Press. Cohen, David (ed.) 1974
Explaining Linguistic Phenomena. New YorkIL ondonlSydneyl
Toronto: Wiley and Sons. Croft, William Typology and Universals. Cambridge: Cambridge University Press. DuBois , John W. 1985 Competing motivations. In: John Haiman (ed.) , Iconici砂 in Grammar , 343-365. AmsterdamIPhiladelphia: Benjamins. 1990
On genetic motivation in grammar 119 Giacalone Ramat, Anna and Paul J. Hopper (eds.) 1998 The Limits of Grammaticalization. Amsterdam, Philadelphia: Be对 amins.
Givón, Talmy 1985 Iconicity, isomorphism, and non-arbitrary coding in s归tax. In: John Haiman (ed.) , Iconicity in Grammar, 187一219. Amsterdam/ Philadelphia: Benjamins. 1990 砂ntax: A Functional-Typological Introduction. Volume 2. AmsterdamIPhiladelphia: Benjamins. Greenberg, Joseph H. 1978 a Generalizations about numeral systems. In: Joseph H. Greenberg, Universals of Human Language. Volume 3: Word Structure , 249-295. Stanford: Stanford University Press. 1978 b Universals of Human Language. Volume 3: Word Structure. Stanford: Stanford University Press. Haiman , John 1980 The iconicity ofgrammar. Language 56: 515-540. 1985 a Natural Syntax. Cambridge: Cambridge University Press. Haiman, John (ed.) 1985 b Iconicity 切 Grammar. AmsterdamIPhiladelphia: Benjamins. Haspelmath, Martin 1990 The grammaticalization of passive morphology. Studi臼 in Language 14: 25-72. Hawkins , John A. (ed.) 1988 Explaining Language Universals. Oxford: Blackwell. Heine , Bemd 1994 Grammaticalization as an explanatory parameter. In: William Pagliuca (ed.) , Per.再pectives on Grammaticalization , 255-287. AmsterdamIPhiladelphia: Benjamins 1997 a Possession: Cognitive Sources, Forces , and Grammaticalization. Cambridge: Cambridge University Press. 1997 b Cognitive Foundations ofGrammar. New York/Oxford: Oxford University Press. Heine , Bemd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. ChicagolLondon: Chicago University Press. Heine , Bemd and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hopper, Paul J. and Elizabeth Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press.
120 Bernd Heine Kuteva, Tania 1994 Iconicity and auxiliation. Journal 01Pragmatics 22: 71-81. Lass , Roger 1980 On Explaining Language Change. Cambridge: Cambridge University Press. Lyons , John 1977 Semantics. 2 Volumes. Cambridge: Cambridge University Press. Miller, George A. and Philip N. Johnson-Laird 1976 Language and Perception. Cambridge , Mass.: Harvard University Press. Moravcsik, Edith A. 1969 Determination. Working Papers on Language Universals (Stanford) 1: 64-98. Pagliuca, William (ed.) 1994 Perspectives on Grammaticalization. AmsterdamlPhiladelphia: Benjamins. Sanders , Gerald A. 1974 Introduction. In: David Cohen (ed.) , Explaining Linguistic Phenomena , 1-20. New York, London, Sydney, Toronto: Wileyand Sons. Saussure, Ferdinand de 1922 Cours de Linguistique Générale. Paris: Payo t. Seiler, Hansjakob 1983 Possession as an Operational Dimension 01 Language. Tübingen: Narr. Traugott, El izabeth C. and Bemd Heine (eds.) 1991 Approaches ω Grammaticalization. Two volumes. Amsterdaml Philadelphia: Benjamins. Wierzbicka, Anna 1985 "Oats" and "wheat": The fallacy of arbitrariness. In: John Haiman (ed.) , Iconicity 切 Grammar， 311-342. Amsterdaml Philadelphia: Benjamins.
Emergent aspect constructions in Present-Day English Christian Koops
Abstract This study analyzes the emergence of aspect constructions in Present-Day English. The constructions under investigation correspond to the three best attested types of progressive constructions across languages: locative constructions, posture verb cons位uctions， and motion verb constructions. The study comes to the conclusion that grammatical structures echoing wellknown cross-linguistic grammaticalization phenomena can be detected in English at a rudimentary and less conspicuous level. The emergent grammatical meanings of these English constructions cover various types of imperfective aspect. 1 Keywords: aspect, emergent grammar, grammaticalization, grammaticalization path, habitual aspect, locative constructions, motion verbs , pos阳re verbs , progressive aspect
1. Introduction
This paper presents an analysis of Present-Day English constructions conveying aspectual meanings at an incipient stage of grammaticalization. It is 1. 1 would like to thank the editors ofthis volume , Günter Radden and Klaus-Uwe
Panther, for their comments and helpful suggestions regarding the ideas contained in this paper and for discussing the relevant issues with me on various occasions. All remaining errors or inadequacies are , of course , my responsibillty.
122 Christian Koops specifically concemed with grammatical phenomena at a nascent level , i.e. with constructions that are in the process of changing their status from lexical to grammatical and have not yet reached the degree of grammaticalization which we generally think of as "true" grammar. The reason for investigating incipient grammar is to explore the factors that motivate the first appearance of grammatical meanings and the structures they are associated with. The assumption underlying this approach is that the early stages of grammaticalization provide a window on the conditions giving rise to grammatical meanings and can thereby contribute to our understanding of how and why grammatical constructions originate. The approach adopted in this paper relies on the findings of many previous grammaticalization studies, which have revealed striking crosslinguistic similarities in the evolution of grammatical meanings and structures from l~ical sources (e.g. , Bybee et al. 1994, Heine and Kuteva 2002). These findings strongly suggest that the frequently attested paths of grammaticalization are not haphazard but conceptually motivated. On the assumption that the same kinds of grammaticalization processes that have occurred in many languages may also occur in any given language at a more rudimentary and less conspicuous level , we can utilize this body of research in order to derive hypotheses about incipient grammar. The methodology followed here is that of drawing on the most frequent crosslinguistic lexical sources of a given grammatical category in order to anticipate the emergence of particular constructions instantiating that grammatical category in a given language. Following this method, the present study focuses on constructions in Present-Day English (PDE) that are 巾， veloping progressive aspect or related aspectual notions. It is generally agreed that emergent grammatical phenomena will be more clearly detectable in spoken language. Therefore , the data base for this study consists exclusively of spoken English. As sources of linguistic data，队'0 electronic co甲ora of spoken American English were used: the Switchboard corpus oftelephone conversations between strangers (Godfrey et al. 1992) and the CSPAE corpus of spoken, professional American English (Barlow 1998).2
2. The Switchboard corpus was compiled in the early 1990s and contains about three million words spoken by over five hundred speakers from various regions of the United States. The CSP AE corpus consists of transcriptions of academic committee meetings and White House press briefings from the late 1990s, comprising approximately two million words. The version of Switchboard used
Emergenta再pect
constructions in Present-Day English 123
This paper is structured as fo l1ows. Section 2 discusses the meaning of progressive aspect and characteristic properties of progressive cons位uc tions. Section 3 su凹eys the lexical sources of progressive constructions as attested across languages. Section 4 presents an analysis of three construction types in PDE corresponding to the cross-linguistical1y attested types of progressive constructions: locative constructions, pos阳re verb constructions , and motion verb constructions. Section 5 summarizes the findings and offers some conclusions regarding the motivation for the English constructions to develop extended , aspectual senses.
2. Progressive aspect constructions 2.1.
The meaning of progressive aspect
Fo l1owing Comrie (1 976), Bybee et al. (1994) view progressive aspect as a sub可pe of imperfective aspect. Like other types of imperfective aspect, the progressive imposes an intemal perspective on a situation. The situation is viewed as ongoing. More specifical1y , Bybee et al. define progressive as the aspectual value that "views an action as ongoing at reference time" so that "the action takes place simultaneously with the moment of reference." (1994: 126, 317). The following analysis will assume this definition. Given that it is a typologically based and therefore language-independent characterization of progressive aspect, it is appropriate for this study, which also takes a 可pological perspective as its starting point and applies crosslinguistic insights to a single language , Present-Day English.
2.2.
Two characteristic properties of progressive constructions
Perhaps the most frequen t1y observed property of progressive constructions is th创 they are typical1y restricted to dynamic events and incompatible with here is the one available online from the Linguistic Data Consortium (www .ldc.upenn.edu , last access: October 2001). The source of all examples taken 企om either of these two co甲ora is indicated as "SWB" or "CSPAE", respectively. If necessary, the examples have been edited to ensure greater readability: pauses, hesitation markers , false starts, and repetitions have been removed.
124 Christian Koops states. For Comrie (1 976) , the incompatibility with states is the definitionaI criterion to identi命 progressive constructions. This restriction can be observed in many languages , as for example in Mandarin Chinese. (1)
Mandarin Chinese a. Zhãngsãn zài dã Zhangsan PROG hit ‘ Zhangsan is hitting Li si. '
Li's ì. Lisi
b. *Zhãngsãn zài you qián. Zhangsan PROG exist money (l it.‘ Zhangsan is having money.') (Li and Thompson 1981: 218 , 219)
It should be noted , however, that the incompatibility of progressive constructions with state verbs is not universal. Progressive constructions in different languages exhibit differing degrees of applicability to states. It appears that especially highly grammaticalized progressives like the PDE Progressive be V -inl , which have undergone considerable generalization of meaning in the course of their historical development, are likely to allow state verbs in some contexts. For example , while it is true that the mentaI state verb know disallows the Progressive in (2胁， it is marginally compatible with it in (3). (2)
a. Sarah is reading. b. *Sarah is knowing the answer. (Bybee et al. 1994: 126)
(3)
Sarah doesn 't know all she should be knowing about this issue.
Despite these reservations , 1 will use the criterion of incompatibility with state verbs here in order to determine whether a given incipient construction is progressive in meaning. First, the constraint appears to hold as a 3. Throughout this paper the convention of capitalizing the names of languagespecific categories introduced by Cornrie (1 976) will be followed. Categories that apply to languages in general are spelled with a lower-case initial. Thus , 1 will speak of the English Progressive be V -ing while on the other hand speaking of different types of progressive constructions.
Emergenta再pect
constructions in Present-Day English 125
universal tendency. Second, it seems justified to assume its validity for cases of emergent progressive constructions because emergent constructions have not yet undergone much semantic generalization. The second characteristic property of progressive constructions relates to the fact that progressive 臼pect involves a concep阳alization of an event as ongoing at the moment of reference to it. As pointed out in the typological surveys by Dahl (1 985) and Bertinetto et al. (2000) , progressive constructions across languages tend to occur most naturally in descriptions of activities relative to a punctual temporal reference point and 缸'e typically not used in contexts in which duration is s位essed. In other words , in using a progressive construction, speakers will usually assume a punc阳al viewpoint on an event. This viewpoint may or may not be made explicit. But if the construction is additionally modified by a temporal adverbial，由is adverbial normally specifies a time point (‘ now' ,‘ still' ,‘ already') rather than a time period ('岛r an hour' ,‘ all day'). In some languages , progressive constructions 缸'e in fact incompatible with adverbials expressing duration, for example the Italian construction with stare ‘ be\ (4)
ltalian *Maria stava ballando per due ore.
(lit.‘ Maria was dancing 岛r two hours. ') (Bertinetto and Delfitto 2000: 207)
The preference for punctual reference points follows from the meaning of progressive constructions as proposed by Langacker (1991: 91-94). Ifthe cognitive effect of using a progressive construction is to de-focus the temporal boundaries of an inherently bounded event, it is to be expected 由at adverbial expressions explicitly referring to these limits are dispreferred, if not disallowed. Although the temporal boundedness of an event is not removed by using a progressive, its boundaries 町e backgrounded to such a degree that they do not represent salient reference points. Again, it should be made clear that this is not a universal cons回int on progressive constructions. It does not，岛r example, hold completely for the PDE Progressive, which can be used felicitously in a context like (4), as demonstrated by the acceptability of the English translation. Nevertheless, even in English there is a preference for punctual adverbials overall. In a corpus-based analysis of the Progressive, Scheffer (1975) analyzed the 岳equency with which different temporal adverbials are used to modi马F Progressive constructions in English fictional texts. Out of a total token
126 Christian Koops
number of 454 temporal adverbials , 73.6% were adverbs like now, still or again , which do not refer to the temporal extension of the event. Adverbials speci命ing duration , such as all day , for an hour, or for α while， only accounted for 16.2% (data from Scheffer 1975: 53 , my calculations). In summary, the two characteristic properties that best identify progressive constructions are (i) their incompatibility with states and (ii) their bias towards occurring with adverbials specifying a punctual temporal reference point rather than with adverbials expressing duration. In Section 4 , these two criteria will be applied to emergent grammatical constructions in order to determine whether they can be analyzed as progressive in meaning.
3. Cross-linguistically attested sources of progressive constructions The grammaticalization of progressive aspect proceeds from a restricted set of source notions. Progressive constructions follow a few , well-documented paths. The diachronic sources at the beginning of these paths have been widely investigated both from a typological point ofview (Anderson 1973 , Blansitt 1975 , Cornrie 1976, Traugott 1978 , Bertinetto et al. 2000) and as a grammaticalization phenomenon (Bybee and Dahl 1989, Heine et al. 1991 , Heine 1994, Bybee et al. 1994). In this section, the best attested types of grammaticalized progressive constructions will be briefly reviewed. 4
3.1.
Locative constructions
The most prominent type of progressive construction across languages what Bybee and Dahl (1 989: 81) have called the "paradigm case" 一 are locative constructions. Locative progressive constructions can often be paraphrased as literally saying that the subject engaged in an activity coded by a verb V ‘ is at V-ing\They typically consist of the combination of a non-finite main verb modified by a locative marker and an auxiliary verb ‘ be' , as illustrated by the German example in (5).
4. For a more detai1ed, quantitative ana1ysis see Koops (2001: 17-31), where the poo1ed results of twenty-one individua1 studies of progressive aspect across 1anguages are summarized.
Emergent aspect constructions in Present-Day English 127 (5)
German Ich bin am Formulieren 1 am at.the f'Ormulate.INF ‘ 1 am f'Ormulating a hyp 'O thesis. '
einer a.GEN
协pothese.
hyp 'O thesis
Altematively, l'Ocative verbs meaning ‘ be at' are used in languages that have them , as in the Mandarin example (1) ab 'Ove. In languages with l'O cative pr'Ogressives, elements such as German am (< an + dem 对 the') in (5) functi 'On as markers 'O f l'Ocati'On 'Or hist'Orically derive fr'O m a l'Ocative marker. F'Or instance, am is a spatial expressi'On in Ich bin am Flughafen ‘ I am at the airp 'Ort.' As an example 'Of a different spatial n'O ti 'O n grammaticalized in this functi 'On c'Onsider the Finnish pr'O gressive c'Onstructi 'On in (6). Here, a n'On-finite verb is marked by inessive case, a case f'Orm meaning ‘ inside 'O f. (6)
Finnish Minä olen m刀I-mä-ssä 1 am sell-3INF-INESS ‘ 1 am selling tickets. ' (l it.‘ 1 am in selling 'O f tickets. ')
lippuja. tickets.PRTV
(Bertine忧'0
et al. 2000: 522)
M'O st 'O f the spatial c'Oncepts f'Ound in l'Ocative pr'Ogressive c'Onstructi 'O ns translate int'O English as ‘ being at' 'O r ‘ being in' , i.e. as expressi'Ons 'O f c'Ontigui句， 'O r c'Ontainment (Traug 'O tt 1978). Less 仕equent but als 'O independently attested are markers translated as ‘ being with' and ‘ being 'O n\5 A variant 'O f this type are c'O nstructi 'Ons using 'O nly the verb ‘ be' and a n'O n-finite main verb , such as the English be V-ing. There has been s'Ome c'On仕oversy 'Over the status 'O f these as l'O cative c'Onstructi 'Ons. H 'Owever, whether they are als 'O essentially l'Ocative in nature , as suggested by Bybee et al. (1994) , 'O r n'O t, as argued by Heine (1994) and Bertinett'O et al. (2000) , is n'Ot directly relevant here f'Or tw'O reas 'Ons. First, clear instances 'O f them seem rare , and, sec'O nd, the pa忧em is unlikely t'O act as the m'O del f'O r new, emergent expressi'Ons 'O f pr'Ogressive aspect in English given that the Progressive be V-ing directly instantiates this type. 5. Heine (1994: 271) analyzes cases 'Ofmain verbs marked by an element meaning ‘ with' n'O t as l'Ocative but as 'O riginally stemming fr'Om an expressi'On 'Of acc 'Ompamment.
128 Christian Koops 3.2.
Posture verb constructions
In many languages, pos阳re verbs , i.e. verbs of bodily orientation in space , used in an auxiliary-like fashion to express progressive aspect (see also Newman 2002 and this volume). The set ofverbs susceptible to this process of grammaticalization appears to be basically limited to ‘ sit' ,‘ stand' , and ‘ lie'. Taken literally, such constructions may appe缸 to express nothing more than two simultaneously ongoing activities: something is done while sitting, standing, or lying. However, their translation equivalence with progressives in other languages indicates their aspec阳al meaning. 缸'e
(7)
Swedish 阿 satt och pratade. we sat and talked ‘ We were chatting. '
(Haugen 1982: 158) In languages employing posture verbs as progressive markers , the resultant constructions often take the form of coordinate structures, as in Swedish. In other languages serializing structures are found. (8)
Ngambay-Moundou m-ísi m-úsa da I-sit I-eat meat '1 am eating meat. ' (Bybee and Dahl 1989: 78)
It is not unusual for languages with pos阳re verb progressive constructions to employ several or all of the verbs ‘ sit' ,‘ stand' , and ‘ lie' as progressive markers. Furthermore, the lexical meaning of these verbs often still exerts a constraining influence on possible complement types and makes the use of either ‘ sit' ,‘ stand' , or ‘ lie' obligatory for actions typically performed in that pos阳re. Therefore, such progressive constructions may not be readily recognized in languages where they have not advanced beyond this stage.
Emergenta再pect
3.3.
constructions in Present-Day English 129
Motion verb constructions
A third well-documented type of progressive construction involves motion verbs. These can be of two types , depending on whether the verb involved in the construction literally expresses directed motion ('go' or, less frequently, ‘come' or ‘ leave') or non-directed motion (often translated as ‘ walk' ,‘ go around' , or ‘ move about'). As an example of a directed motion verb , consider how the Italian verb andare ‘肘， can be used in connection with the non-finite -ndo verb form to express a progressive situation. (9)
Italian Teresa andava scrutando l'orizzonte. 'Teresa was scanning the horizon.' (lit.‘ Teresa went scanning...') (Bertine忧02000:
579)
The Spanish verb andar ‘ walk, go around' exemplifies the second type of motion verb. In its literal meaning , it expresses non-directed motion. (10)
Spanish ... ve a la niña que anda jugando en el recreo. ‘.. she sees the child who is playing during recess' (l it.‘. ..who goes around playing...') (Torres Cacoullos 1999: 52)
If the overall degree of grammaticalization of such constructions is low, the original meaning of motion in space will still harmonize with that of the main verb activity so that the construction is frequently applied to actions actually performed in motion. That is , as with pos阳re verbs , these constructions may at first appear to express nothing more than an agent caπr ing out two activities simultaneously. But again, such an interpretation does not sufficiently characterize their aspec阳al meaning.
3.4.
Other sources
There is a small number of other sources of progressive aspect constructions besides the three discussed above. These inc1ude lexically durative verbs with meanings centering around ‘ stay' ,‘ remain' , and ‘ live' , verbs
130 Christian Koops
meaning 吁lold' or ‘ have' , and continuative constructions, i.e. constructions indicating that an activity is not merely ongoing but actively maintain民 (‘ keep doing').ηlese source 可pes are less frequent overall and will be neglected here. Nevertheless, the fact that they have been found repeatedl:in distinct languages suggests a notable degree of motivation for them as well.
4. Constructions expressing incipient aspectual meaning in PresentDay English (PDE) This section presents the results of a largely corpus-based investigation i:: search of emergent progressive aspect constructions in PDE. The choice 0: expressions to be considered here is determined by the degree to which 缸 English construction or collocation pattem resembles and therefore potentially exemplifies one of the three construction types discussed above. Th e three types of constructions in question are: locative constructions (Sectio r: 4.1) , pos阳re verb constructions (Section 4.2) , and motion verb constructions (Section 4.3). The analysis is specifically concemed with these three types because 企om a typological perspective they are the best candidates for emergent progressive meaning. For those constructions which exhibi: an emergent aspec阳al meaning , the degree to which their meaning is progressive wil1 be determined on the basis of the properties of progressi\ε constructions discussed in Section 2.2.
4.1.
PDE locative constructions
In searching for locative progressive constructions in PDE as they exist , fo了 example, in German (as seen in (5) above) , it is immediately evident tha: none of the prepositions at, in , with , or on are used in productive prepositional constructions with progressive meaning. (11)
1 was
(*atl*inl*wi由/*on)
watching TV when they called.
Such constructions existed at earlier stages of English. They are historica l1:. attested starting with the widely cited Old English ex创nple ic wæs on hu r.tunge ‘ 1 was (on) hunting' (Scheffer 1975: 244). In Middle English 缸ζ Early Modem English at least the three prepositions 归 ， on , and upon were
Emergent aspect constructions in Present-Day English 131 used in the same construction type (Jespersen 1931: 174). Today, remnants of the old locative constructions live on in some dialects of English, where constructions like He 's been a-huntin' are still used. However, locative progressives are obsolete in the standard language. An unstressed preposition in , at, with or on in a context like (11) results in unacceptabili可. The co甲us data show no evidence of such progressive constructions. 6 Nevertheless , as pointed out by Bolinger (1971) , there are a number of ways in which prepositions are used today in expressions with progressive meaning. First, the prepositions at， 的， and on combine with some action nominals to express progressive events , as in (12). Some ofthese constructions ac阳ally have morphologically related paraphrases with be V -ing, as seen in (13) , which demonstrate their progressive semantics. (12)
a. She is at work / in a conversation / on the phone.
(13)
a. They are at dinner / in afight / on tour. b. They αre dining / fighting / touring.
Second, at can be used if a progressive situation is pronominalized. In (14) , the pronoun it stands for the previously mentioned , ongoing action.
(14)
He was working an hour ago and 1 guess he 's still at it. (Bolinger 1971: 246)
Interestingly, the preposition with can also be used to pronominalize ongoing events. It functions in the continuative cons位uction with the verb keep. 6. Patterns which superficially instantiate be in V-ing do exist, for example in She is in marketing. But here the complement of in is a norninal and the expression as a whole is not progressive. The only cases that are also progressive in meaning involve stressed locative particles, as in: (i)
They rnight kind of feel like they need to go out and play all the time , not be in studying. (about sending kids to college in California, SWB)
However, the fact that both in and the non-finite verb form can be omitted individually without rendering the sentence incomplete shows that they do not form a constituent, as one would expect. (ii) (iii)
They don't feellike they need to be in. They don 't feellike they need to be studying.
132 Christian Koops (15)
1 pretty much kept with it until 1 got a job in Virginia. (about jogging, SWB)
As a third instance of a prepositional expression with progressive meaning , consider how at is in fact obligatory in a type of cleft construction - what Bolinger called "cleaving" 一 where the non-finite -的g form is syntactically isolated. (1 6)
a. Is it studying he 's at or mαking love? b. *Is it studying he is or making love? (Bolinger 1971: 247)
These phenomena show that the locative concepts found in grammaticalized progressive constructions in other languages are also employed by speakers of English to conceptualize progressive situations. Nevertheless , since (1 2) 一 (16) do not represent productive grammatical constructions , they will not be further discussed here. The only productive locative constructions with progressive meaning in PDE are be in the middle ofV-ing and be in the midst ofV-ing. (1 7)
Normally when they call you 're either in the shower or you [wereJ in the middle of cooking something and you had to stop everything 归阳nω the phone. (about telemarketing calls , SWB)
(1 8)
He ￥
in
the midst of reviewing operations here, 1 think, as well as the crime bill passed and get health care r矿orm passed and tackle all the substantive issues. (about the president, trying ω get
CSPAE)
These constructions differ from other locative progressives based on the notion of containment. Wh ereas the preposition in expresses general containment, the complex prepositional phrases in the middle of and especially in the midst ofhave much more specific meanings. Literally, they focus on the central point or central area of a container. 7 As we will see below , their narrow spatial meaning carries over into their progressive use.
7. There is some independent evidence that spatial notions as specific as ‘ in the middle of can act as sources ofprogressive markers. For example , Korean has
Emergenta，再pect
constructions in Present-Day English 133
The existence 'Of tw'O altemative spatial terms, middle and midst, raises the questi'On 'Of whether the c'Onstructi'Ons in (17) and (1 8) have basically the same aspec阳al sense 'Or whether there is a discemible semantic difference between being in the middle 'Of d'Oing s'Omething and being in the midst 'Of d'Oing it. Intuitively, the c'Onstructi 'On with midst seems t'O imply greater inv'Olvement 'Or a m'Ore c'Omplex activity. Yet n'O clear difference emerges 企om the attested instances in the C'O甲us. Their meanings are t 'O'O similar ωwarrant c'Onclusi'Ons ab 'Out semantic differences. It is, nevertheless, interesting t'O n'Ote 也at the tw'O expressi'Ons be in the middle 01 and be in the midst 01 differ c'Onsiderably in terms 'Of what types 'O f c 'Omplementati'On they 'O ccur with. Table 1 summarizes the frequency distributi 'On 'Of c'Omplement types in terms 'O f a three-way distincti 'On between (i) verbal c'Omplements, as in (17) and (18) ab 'Ove, (ii) abs忧郁t n'Ouns th创 have durati'On as part 'O f their meaning, as in (1 9) , and (iii) l'Ocative n'Ouns, as in (20). (19) a. We're right smack dab in the middle 'Of baseball season right now. (SWB) b. Our competitors are now in the midst 'Of billion dollar campaigns. (CSPAE) It 注 right
(20)
in the middle 'O f Fort Worth , Dallas. (SWB)
Table 1. Frequency of be 切 the middle ~向nd be 切 the midst o[ with different types of complementation in Switchboard and CSPAE. b也 in
V-ing
5
the middle of abstract N l'Ocative N
32
28
V-ing 5
be in the midst of abstract N l'Ocative N
5
Table 1 sh'Ows that the number 'Of abstract n'Ouns 臼 c'Omplements 'Of be in the middle 01 is ab 'Out equally high as the number 'Of l'Ocative c'Omplements. This indicates 由at its aspec阳al use is at least as well established as its spatial use.ηle be in the midst ofc'Onstructi 'On appe邵阳 have devel 'Oped even further in this directi 'On and t'O have specialized as an expressi'On 'Of aspect
a progressive construction incorporating the defective noun cwung ‘ middle , center' (Martin 1992: 463).
134 Christian Koops since no spatial use is attested at all in the co甲us. Indeed, to use in the midst ofin a purely spatial sense seems impossible. (21) a. 1 left the car in the middle ofthe parking lot. b. ??I left the car in the midst ofthe parking lo t. The next question to consider is whether be in the middlelmidst of V -ing imposes an aspec阳al contour on the main verb that can be called progressive. Aspectually, the middlelmidst-construction exhibits the properties of progressive constructions. First, it is incompatible with stative verbs , including stative verbs like sit or stand, which are compatible with the Progressive be V -ing. (22) a. *1 was knowing the answer. b. *1 was in the middle of knowing the answer. (23) a. 1 was sitting in my chair I standing over there. b. *1 was in the middle of sitting in my chair I standing over there. Second, like progressive constructions，加的 the middlelmidst of V -ing is used in punc阳al contexts. It is , in fact, incompatible with adverbial expressions of duration.
(24)
??For ten minutes she's been in the middle of studying the manual.
At the same time, an analysis of be 切 the middlelmidst ofV-ing must take account of the fact that its overall degree of grarnmaticalization is extremely low. An indication of this , besides its infrequent use in discourse (see Table 1), is its relatively low frequency of occurrence with a verbal complement. As a productive grammatical aspect construction it should be expected to occur predominantly with verbs rather than with nouns. However, the dis位ibution across different types of complementation shows that when talking about being in the middle or midst of an event, speakers seem to prefer expressing that event as an abstract noun rather than as a verb form. This indicates a rather low degree of grarnmaticalization. In this respect, be in the middlelmidst of V -ing contrasts with another incipient progressive construction which has not been dealt with so far: the construction be 切 the process of V -ing. The reason for not including this
Emergent aspect constructions in Present-Day English 135
construction here is that it incorporates the abstract, non-spatial noun process and thereby only pa此ially represents a locative progressive construction. 8 A comparison of this construction and the middle/midst-construction brings out an interesting difference that reflects their respective degrees of grammaticalization. With be in the process oJ, the patteming of complementation types is reversed: verbal complements are clearly preferred to nouns. Cases like (25a) are far more frequent than cases like (25b). (25) a. We 're in the process of moving the force 归 now. (CSPAE) b. We 're in the process of a home r，写pair right at this point. (SWB) Both indicators of degree of grammaticalization - discourse frequency and type of complementation - are summarized in Table 2.
prefeηed
Table 2. Frequency of be in the middle/midst ofV-ing and be in the process of V-ing with different types of complementation in Switchboard and CSPAE
旦拙!
Lne- r-o-c-nc-3-bF-a3 叫叫-qu
n一
机
e 一窑
V-ing 10
of abstract N 37
，。--m 句3 8
b恒的 the middle.如idst
FLV
N
Why , then, is the middle-construction used so rarely in discourse? The answer evidently lies in its highly specific meaning. As alluded to above, it seems that the literal meaning of ‘ middle' is still prevalent. Whereas the Progressive be V-ing expresses an ongoing event without speci命ing whether that event is at its temporal beginning, middle, or endpoint, the middle/midst construction refers very literally to its midpoint. The retention of spatial meaning can explain its 仕'equent occurrence with abstract nouns rather than with verbs. Abstract nouns metaphorically present events as things and allow the speaker to refer to them via the ontological metaphor EVENTS ARE OBJECTS (Lakoff and Johnson 1980). Due to its retained spatial meaning, the middle-construction is more compatible with abstract nouns , given 由at these are construed as object-like and therefore more easily thought of as having a midpoint. 8. This is not to deny 出at locative phrases with non-spatial nouns can be the source of progressive aspect markers , as for example in the case of the French cons仿uction être en train de + Infinitive.
136 Christian Koops
In order to further motivate the low overall discourse frequency of the construction be in the middle/midst of V-i略， 1 will illustrate some additional meanings typically found in connection with this construction in the corpus. These meanings can explain its low rate of occurrence because they constrain its applicability and make it inappropriate in many otherwise progressive contexts. They can be understood as pragmatic implicatures arising from the concrete meaning of being in the middle of something which have conventionalized to different degrees. Obviously, in saying that somebody is in the middle of doing something, more is conveyed than the fact that the action is ongoing. The conceptualization of the activity as having a center and of someone as being located there typically implicates deep involvement in the activity. This involvement is often interpreted as the focusing of one's attention , which explains why be in the middle ofV-ing cannot be used with activities thal involve little or no cognitive effort. (26)
??Wh en 1 saw him he was in the middle of chewing gum wearing sneakers.
Second, the constraint on mental involvement goes along with an intentionality requirement. The subject participant must be willingly involved as instigator of the activity , which explains the semantic oddity of (27). (27)
??1 was in the middle of drowning / suffocating / losing my mind
Furthermore, there appear to be general restrictions on the nature of the subject participant. All attested instances involve humans acting as agents Patientive participants that are not mentally or willingly involved are unlikely , if not impossible, to be expressed as in the middle of an activi t: This is another point of contrast with the Progressive be V -ing, which allows such participants. (28) a. 1 was being watched / ignored by everyone. b. ??1 was in the middle of being watched / ignored by everyone.
A patient can only be the subject of be in the middle/midst ofV-ing in contexts where the person is also the instigator and ultimate controller of the actívíty. (29)
1 was in the middle of getting my hair cut.
Emergent aspect constructions in Present-Day English 137 Lastly, as a result of being deeply engaged in an activity , the subject participant of be in the middle/midst ofV-ing is often understood to be markedly busy for the duration of the activity and therefore unavailable to others unless he or she is interrupted. The exchange in (30) illustrates this very vividly. (30) A: Wh en is yours on? B: 1旨n. 1 was just in the middle of watching it. A: Oh gosh , well, I'm sorry 1 interrupted. (about favorite news shows , SWB) To summarize, there is ample evidence for incipient progressive meaning in be in the middle/midst of V-ing. The construction is a c1 ear instance of an emergent progressive aspect construction. Its restriction in discourse to a small number of progressive contexts can be related to its highly specific semantics and various additional implicatures such as the notion of being mentally focused on a task and being busy doing it, which have become conventional aspects of its meaning.
4.2.
PDEpos阳re
verb constructions
As pointed out in the discussion ofposture verb progressives in Section 3.2 , at an ear1 y stage of grammaticalization these constructions superficially resemble free syntactic phrases expressing nothing more than an action carried out whi1e sitting, standing, or lying. Therefore, pa忧erns like sit and V should be expected to occur in English as well , regardless of whether they do in fact represent emergent progressive constructions. And indeed, Quirk et al. (1985) give such examples, as in (31). Interestingly, the authors discuss these as cases of "pseudo-coordination", suggesting that they involve something other than canonical coordination. (31)
a. They sat and talked about the good old times. b. Don 'tjust stand there and grin. (Quirk et al. 1985: 978)
Before entering the analysis of the syntactic and semantic properties of such constructions, it is interesting to note how common the use of some posture verbs in constructions like (31) is. To show this , 1 am comparing
138 Christian Koops their rate of occurrence in two syntactic contexts: (i) their use in simple clause structures like (32) , and (ii) their use in bi-clausal structures with a following verb that has the same subject, as in (33). (32)
a. J'll stand over here so you don 't have to turn your chairs. (CSPAE I b. 1 could just lay in thαt hammockfor hours. (SWB)
(33)
a. They made me sit and waitfor forty乒νe minutes. (SWB) b. It gets annoying when they stand right over you watching thing you do. (about co-workers , SWB)
eveη
Most cases of pos阳re verbs in bi-clausal structures are coordinate constructions like (33a). Less frequently , the second verb appears in a nonfinite adjunct clause as in (33b) (cf. Kuteva 1999 on coordinate structures) Table 3 gives the total and relative frequencies of sit, stand, and lie in these two syntactlc contexts.
Table 3. Frequencies of sit, stand, and lie in syntactic contexts like (32) and (33 I in Switchboard and CSPAE9 szt sit and V, sit v二 ing 569 593 (49%) (51%) Sit
stand stand and V, stand V-ing 120 91 (57%) (43%) stand
lie 64
(7 5%)
lie lie and V, lie V-ing 21 (25%)
The frequent appearance of posture verbs in constructions like (33) point、 to their reanalysis as incipient aspectual expressions , i.e. , this reanalysis manifests itself in the use of these posture verbs connection with another verb. 9. The count for lie includes all instances of its variant lay , as in example (32b) Excluded were all instances of transitive use and many extended meanings of posture verbs in English (This doesn 't sit well with me; 1 cannot stand cardboard food; Wh ere do you stand on gun controlη. The reason for excluding these is that their specific semantics make it unlikely or impossible for them to participate in the grammaticalization process in question. In borderline cases an extended meaning was not excluded when at least one example of the use in connection with another verb could be found.
Emergenta再pect
constructions in Present-Day English 139
The 仕equency
distribution in Table 3 shows that the posture verbs used in bi-clausal constructions are sit and stand. As we will see below, this can be accounted for by their extended u面， for example as aspec阳al verbs. Since this extended use is most conspicuous in the case of sit and, to a lesser extent, in the case of stand, the following analysis will 岛 cus on constructions involving these two verbs. There are a number of indications that posture verb constructions with sit and stand have developed unique syntactic and semantic properties. They exhibit effects which one would expect if the verbs sit and stand had in fact taken on some of the features of auxiliary verbs and if the two-verb sequence (e.g. , sit and V or sit V-ing) had in fact developed some characteristics of an auxiliary-complement verb structure. One such effect is an increase in syntactic cohesion of the verb sequence. This is a predictable result of the beginning reanalysis of a pos阳re verb as an aspectual auxiliary that forms a syntactic unit with another verb. If reanalysis occurs , we would expect the sequence sit and V to differ from a sequence like smile and V in that the la饥er is analyzable compositionally as a case of ordinary coordination, whereas the former represents a more integrated expression. One indicator of syntactic unity are extraction phenomena such as relativization and clefting, which can be found despite the fact that extraction is not generally possible out of a coordinated verb sequence. most 企'equently
(34) a. They have movies that sometimes we'll sit and enjoy. (about cable TV, SWB) b. So all we had to sit here and ask is, well, what aren 't th句1 telling us? (about news censorship, SWB) For example, the possibility of extracting the direct object movies in (34a) out of the sequence sit and enjoy suggests that the verb sequence has to some extent coalesced syntactically. Such data suggest a syntactic structure in which movies acts no longer as the direct object of enjoy only, but as the direct object of sit and enjoy. The sequence sit and V has the character of a complex predicate rather than that of two independent verbs. The semantic properties of sit and stand in posture verb constructions provide further evidence for their reanalysis as something other than verbs of bodily pos阳re. While almost all attested cases do allow a compositional reading , i.e. an interpretation of an action carried out while sitting or standing, for the majority of them such an interpretation is not sufficient. It
140 Christian Koops is inadequate in two ways. First, to interpret them literally often attach~ too much weight on the original postural meaning of the verb. As we wiL see below, pos阳re verb constructions are often not used to talk about being in a posture. Second, a literal interpretation misses important additiona: meanings implied by these constructions. 1 will discuss these two aspects 巳 阳m.
Often the meaning of being in a specific posture appe缸s bleached ou工 This can be demonstrated by negating the verb sequence. As it tums ou:. the default interpretation is that the second verb , and not the pos阳re vert. is negated. (35) a. 1'm not gonna sit and read War and Peace if1 can rent the movie b. ?I'm not gonna sit and read War and Peace if 1 can lie down ar;,; read it. More precisely, what is negated in (35a) is the time commitment implied b:. the construction. An interpretation in which the postural meaning of sit i~ negated is unlikely without a special context. 10 Bleaching of the original meaning of sit and stand can also be obser\"民 where little or no locative information follows the pos阳re verbs , i.e 1:: cases of the structure sit and V or constructions where the pos阳re verb i~ only followed by there, such as stand there and V. Given that pos阳re vert~ are also verbs of location, we would expect them to be used with some additional spatial elaboration that specifies the location where the sitting c:.standing occurs. However, often the location of the subject referent appear5 to play no role at all. (36)
You sit there and read hundreds and hundreds of cases and the ... you get one exam for the whole semester and it 's how well yω can remember it all. (about law school, SWB)
In the context in which (36) is uttered, there does not refer back to a町 previously mentioned location. Neither is the location where the sitting takes place taken up again in later discourse.
10. This is not to say that the literal meaning of ‘ si创吨， cannot be negated. Ho\\ever, doing so is only possible if the postural meaning is emphasized b:. stressing the pos阳re verb.
Emergent aspect constructions in Present-Day English 141
These syntactic and semantic effects raise the question of what the purpose of the posture verbs in these constructions is. If they are not, or at least not primarily, used to talk about being in a pos阳re or being in a location, what do sit and stand convey? An d, most important1 y, is there any evidence that sit and stand contribute progressive meaning to the conceptualization of the event coded by the following verb , as suggested by parallel developments in other languages? One of the main purposes of the widespread use of sit and stand in posture verb constructions appears to be the implication of extended duration. That is , these pos阳re verbs function to express or to emphasize that an activity takes a long time. This is particular1y relevant in the context of the present discussion, given the expectation that such constructions may develop aspec阳al meaning. (37) a. We have discovered the dog sits and watches them for hours on end. (about pet hamsters , SWB) b. A: Wh at in the world do you do with your computer that takes so much time? B: WeU, 1 sit and design cards. (SWB) c. If there is a headline that grabs my attention then 1 will stand there and look at itfor a minute then walk into my office. (SWB) The semantic contribution of the posture verbs in examples like (37a-c) is that an agent does something for an extended period of time or is spending a significant amount of time on a task. This is often reflected overtly in adverbial expressions emphasizing duration, such as for hours on end in (37a). Sometimes the notion of time is explicit1y mentioned, as in (37b). Examples like (37c) , where the time periodfor a minute is not objectively long , indicate that the duration need not be long in an absolute sense. Rather, it is conceptualized as an extended period re1 ative to the kind of activity referred to, as in this case, reading a headline. Additional justification for analyzing these constructions as involving extended duration comes from the fact that speakers use them in just this sense even with non-typical subjects of sit, namely inanimates. Although inanimate subjects are not near1 y as common as human subjects in these constructions, inanimates as subjects c1early bring out the loss of pos阳ral semantics and the added sense of duration.
142 Christian Koops (38)
There was equipment at every base 1 was ever on that was jus: going ω waste. lt was just sitting and rusting. (about militar:spending, SWB)
As a last piece of evidence showing that these constructions can impl:prolonged duration , consider (39). Here , the use of sit is particularl:. interesting because the activity referred to by the fo11owing verr. renovating a house , is incompatible with the literal meaning of sitting. (39)
1 hate it when people do that. They'lI sit there and fzx up the: r house and then they move. (about having to move in order tι transfer to a new job , SWB)
The subject referents in (39) are not ωbe understood as fixing up theL houses while sitting down. Rather, sit is best taken to imply that a considerable amount of time is spent on the task. It must be noted at this point that duration is not the only extend民 meaning implied by these constructions. The verbs sit and stand convey l wide range of other, most1y evaluative notions , which are a11 related tζ aspects of si忧ing or standing, but which are not primarily about duration (40a) and (40b) 町e ex创nples of a very common extended meaning , the notion of inactivity or idleness. With stand, there is even a sense of apath~ Speakers use these constructions to give a negative evaluation of an activi巧 or of a person carrying out an activity. Another meaning sometimes implieè by the use of sit is the notion of relaxation or pleasure, as in (40c). (40) a. They're not going ωsit there and lose money. 矿的。1 're losing money then they're going to raise rates. (about insurance companies , SWB) b. How could you stand there and watch them beat that guy? (SWB) c. lt 恒 ve.η1 convenient on afternoons when 1 wanna just get awa.'. and sit and think. (about fishing , SWB) These meanings are important to note because they account for many instances of c1 early non-literal uses of sit and stand where an interpretation 0: the activity as tempora11y prolonged, as in (37a-c), does not sufficientl:. characterize the overa11 meaning of the construction, and could therefore not explain why the pos阳re verb is used at a11. Only an analysis of a11 the various extended uses could account for the distribution in Table 3.
Emergenta司pect
constructions in Present-Day English 143
Coming now to the main question, can the extended, durational meaning associated with these English posture verb constructions be regarded as emergent progressive aspect? To begin with, it is safe to say 由at the sense of duration implied by sit and stand makes these constructions aspectual. They are aspect constructions in that 出ey clearly focus on the internal temporal constituency of an event (Comrie 1976). Like other aspec阳al expressions, they convey a concep阳alization of an event in which the temporal structure of the event is altered. Here , the temporal structure is altered in the sense that an event is conceptualized as temporally prolonged. Nevertheless, there are several reasons to assume that the English sit and stand constructions are not typical progressive constructions. First, we find cases where the second verb expresses a state, as in (41). (41)
A sign泸cant port印n of kids sit there and could give a damn. (about high school students, CSPAE)
In general , the activities referred to in the pos阳re verbs constructions tend to be activities low in dynamicity. They include mental states as in (41). Second, the temporal adverbials used with these constructions are diι ferent from those used with progressive constructions (cf. Section 2.2). The vast majority of adverbials occurring with the English sit and stand constructions express a time period ifor an hour, all day , for a while) rather than a time point. For example, there are 40 instances oftime period adverbials used to modi命 the sit-construction in the co甲us but only 3 instances of time point adverbials (now , still). Moreover, these 3 cases are all sentences which also instantiate the Progressive be V-ing, such as (42). (42)
Right now 1 was just sitting here watching TV. (SWB)
The punctual viewpoint re f1 ected in the use of the adverbial right now in (42) is likely to be due to the Progressive construction be V-ing rather than the posture verb construction. Indeed, if a punc阳al viewpoint is imposed on a situation expressed only by sit and V, no interpretation as progressive results. (43)
?When 1 entered the room she sat there and cried.
144 Christian Koops In (4月， sit and cry does not imply an ongoing event. Rather, to the exte::: that the sentence is acceptable at all, the person's crying is interpreted .:..主 starting at the moment ofthe punctual event in the adverbial clause. Third , the possibility of combining a posture verb construction with :工 t Progressive , and thereby integrating their respective construals , is in Ítse ,:' evidence that posture verb constructions do not convey progressive meι ing themselves. This is because a progressive construction cannot be a;plied to itself. This constraint clearly holds for the emergent progressl'. 主 construction be in the middle ofV-ing. (44)
a. *1 was being watching TV. b. *1 was being in the middle of watching TV.
These aspects ofthe PDE pos阳re verb constructions show that their inci仁' ent aspectual value is best analyzed as durative in the sense of expressin 豆 prolonged duration. Before concluding this discussion , however, it is inte::-esting to note that some of the special characteristics of the English sit ar: ζ stand constructions are also found for pos阳re verbs constructions in othe::languages. As pointed out by Ebert (2000) , even the more grammaticalizeζ posture verb progressives that exist in other Germanic languages such 2.5 Dutch , Danish , or Swedish (see example (7) in Section 3.2) always impl:. some noticeable duration. For example , the Danish construction in (-t 5 conveys not only a progressive situation but at the same time implies th二 the boiling of water has been in progress for some time.
(45)
Vandet står og koger. water. DEF stands and boils ‘ The water is boiling (and has done this for a while).' (Ebert 2000: 621 f
,
Furthermore, according to Ebert, progressive constructions based on posture verbs are generally exceptional in their compatibility with adverbs indicating a time period, which is exactly what we find in English. These independent observations shed a slightly different light on the Englist pos阳re verb constructions. They suggest that the state of grammaticalization found in PDE can be the basis of further extension to genuine progressive aspect , as it seems to have occurred in other languages. In summary , the analysis of the PDE posture verb constructions with 51: and stand has revealed incipient aspect constructions used to convey the
Emergent aspect constructions in Present-Day English 145 meaning of extended duration. This meaning appears to be based on pragmatic implicatures that have been partially conventionalized and are now exploited with some regularity. Furthermore, besides the aspectual notion of duration, the same constructions are also used to express other, nonaspectual meanings such as being inactive or idle.
4.3.
PDE motion verb constructions
Lastly , let us consider the meanings ofPDE motion verbs in pattems coηe sponding to grammaticalized motion verb progressives in other languages. As pointed out in Section 3 .3, the motion verbs that grammaticalize into progressive markers can , in their original, spatial senses , express both directed or non-directed motion. Therefore , both of these two 叩pes of motion will be considered here. Regarding constructions involving directed motion verbs such as go , the co甲us data provide no evidence of emergent progressive or other imperfective meaning. It appears that this is due to a semantic property of English go-constructions which prevents their extension to imperfective aspec t. To see this , consider how go can be used in an identical subject construction with another verb. There are three syntactic options to do this in English: go V , go and V , and go V -ing, as illus位ated in (46a-c). (46) a. 1 /i ke to go watch the Giants when th可 come ω Houston some times. (SWB) b. They just ldnd of /i ved out in the woods and ev的J now and then th锣 'd go and shoot their rifles and then they'd come back. (SWB) c. Her dad has a cabin and they gofishing up there. (SWB) In each example, the two activities coded by go and the following verb together constitute a complex event. However, the way in which the two subevents are related to each other differs in an important respect 仕om the concep阳al integration found , for example, in pos阳re verb constructions. Unlike in sit and V , the two activities in (46a-c) are viewed as sequential rather than as concurrent. That is , go codes a motion event and the following verb specifies an action carried out at the spatial goal of that motion event. On the other hand, grammaticalized directed motion verb constructions , as well as pos阳re verb constructions, appear to be based on a conceptualization in which the two subevents are simultaneous (cf. the Italian
146 Christian Koops example in (9) above). Since the motion events described by go in (46a-c) are not conflated with the respective second event, English directed motion verb constructions lack the prerequisite for their reanalysis as progressive constructions. 11 By contrast, non-directed motion verbs , such as go around, are easily interpreted as expressing an event that occurs simultaneously with another event when used in constructions like go around V -ing or go around and V. The interpretation as two concurrent events results naturally given that non-directed motion lacks an inherent beginning and end point. This is illustrated in (47). (47) a. He went around putting ho仰的 anything he wanted to at what ever rate ofspeed hefelt like. (about a shooting incident, SWB) b. These people that go around and kill children and women and everything else without any remorse, 1 don 't think they deserve to live. (SWB) The examples in (47) show that the two activities may be either simultaneous in a 由ict sense, i. e. an action performed in motion as in (47时， or such that one activity is inc1uded in a larger motion event forming its temporal background, as in (47b). As with pos阳re verbs , it is interesting to note how 仕'equent1y some nondirected motion verbs are used in constructions like (47a-b). The three verbs most commonly used in this way are go around, walk around, and run around. Table 4 gives their total and relative frequencies in simple c1auses like (48) and in syntactic contexts like those under (47).
11. There is a small class of directed motion verb constructions in which the two verbs can generally be interpreted as concurrent. These involve the structure go V-ing as in He went tumbling down the steps. Such constructions are interesting aspectually because go occasionally leads to an additional inceptive interpretatlOn, as 10: (i) Let's straighten out, you know, use that money and straighten ourselves out before we go 位ying to heal the world. (about giving tinancial aid to developing nations , SWB) The meaning of go as part of the phrase go trying ω heal the world in (i) can be interpreted as inceptive. It could be paraphrased as ‘ start 位ying to heal the world'.
Emergenta，再pect
constructions in Present-Day English 147
(48)
Wh en 1 walked around b矿ore the lunch the anxiety 切 the room was palpable. (CSPAE)
Table 4.
Frequencies of go around, walk around, and run around in syntactic contexts like (47) and (48) in Switchboard and CSPAE J2
革'0
around go go around and V. around go around "亿加g 41 42 (49%) (51%)
walkaround "但uc walk around and V. around walkaround v-仇g 44 21 (68%) {32些}
run around run run around and V. around run ar叨lTu:Hζ加g 78 18 (19%) 旦旦l
The frequency distribution in Table 4 shows that the verb used most often in a complex construction with another verb is go around. This suggests that go around-constructions (go around and V, go around V -ing) are most likely to have developed extended meanings, which cause them to be more 仕equent. As w i11 be shown below, such extended meanings are indeed best attested for go around. However, an emergent 部pectual sense comes out in constructions with a11 three verbs. Moreover, their respective aspectual meanings are very simi1ar overa11. For this reason 由e fo11owing discussion wi11 not be concemed wi也 the semantic differences among the constructions wi由 go around, walk around, and run around but rather focuses on their common 臼pectual meaning. P町a11el to what was noted for posture verb constructions , there is some syntactic evidence suggesting that the verb complex in go around V -ing constructions exhibits a higher degree of syntactic unity than a loose s归1tactic phrase in which the two verbs just happen to co11ocate. As with sit and V, there are instances where the object of the second verb is extracted under relativization.
12. The count includes occurrences of go around, walk around, and run around with additionallocative elements like PPs, NPs , and adverbials following them. Excluded are cases in which around describes directed motion as well as some extended meanings of go around and run around (Th ere are not enough handouts to go around; Th e house ran around a hundred thoωand dollars).
148 Christian Koops (49)
And the third theme, 1 think, throughout the advertising is the 21 st century agenda that the President is going around talking about. (about Clinton campaign a缸， CSPAE)
As argued above for sit and V , such data can be taken as evidence for an integrated construction in which go αround and talk about NP together act as a single complex predicate. Coming to the semantics of the English non-directed motion verb constructions , the majority of the cases in the co甲us can be characterized by a literal interpretation in which the subject participant actually moves from one place to another. The result is a complex event in which the action is distributed over several separate locations. In other words , the same action is performed repeatedly in various places connected by a spatial trajactory. (50) a. They started going around and looking at paint colors. (SWB) b. Wh en it only tells me that a door is open , 1 have to get out 01 the car and go around slamming doors. (about warning lights in cars, SWB) In (50a,b) the idea that the action is repeated is also expressed by the plural number of the objects paint colors and doors. It can be shown, however, that the motion verb constructions by themselves can bring about the sense ofrepetition. To see this, compare (51a) to (51b). (51) a. They released him and he had to go around tηlÏng to sign up with somebody. (about a professional athlete, SWB) b. They released him and he had to tη to sign up with somebody. In (51 a) go around implies that the person had to try again and again, making a number of separate attempts. If the motion verb is absent, as in (51 时， the sentence can be understood as involving only a single try, i.e. not as a repeated event. Example (51 a) also illustrates another semantic prope句r of these motion verb constructions. The motion component is often considerably backgrounded. In (51 时， the spatial trajectory connecting the individual subevents is less relevant than the fact that the same action is performed repeatedly. The motion meaning of go around appears to be bleached out. What remains is the aspectual meaning of repeatedness or iterativity. It is interesting to note that the original sense of motion appears most subdued
Emergent aspect constructions in Present-Day English 149 and the emergent repetitive meaning is most prominent in one particular use of these constructions, namely when speakers talk about an activity as repeated an unnecessary amount of times. (52)
a. You have ω have trust 的 people some place along the way. You can't go 缸ound testing everybody 卢r everything. (about drug screening, SWB) b. 1find it very silly the way politicians are running around creating more and more capital crimes instead of dealing with the real problems. (SWB)
In this function , when the verbs are used to imply the notion of excessive repetition, it is sometimes unc1 ear how motion is involved at al l. For example, the activity referred to in (53) is an intemal activity and not connected to any specific location. (53)
1 don 't think most people walk around worrying about how politicians are going to get money ω pay for their campaigns. (about campaign finance reform, CSPAE)
In summary, the English non-directed motion verb constructions show a c1ear emergent aspectual meaning of repetitiveness. The basic aspec阳al value seems to be repetitive , or 仕equentative， aspect. 13 This emergent meaning resembles progressive aspect in that the repeated events also have no c1 ear beginning or end point. The first and the last in the series of events they express are not focused on. Thus , like progressive events , they instantiate a subtype of imperfective aspect. In general , however, the repetitive meaning of the non-directed motion verb constructions is so c1 early different from progressive aspect that the diagnostic tests for progressive meaning need not be applied here.
13.If the repeated activity is characteristic of an entire time period, a habitual reading results. See Stefanowitsch (2000) for an analysis of the construction go around and Vas habitual.
150 Christian Koops
5. Conclusion The empirical investigation has shown that incipient grammatical constructions conveying aspectual meanings can be detected in Present-Day English. An emergent aspectual sense is noticeable for all three construction types , which warrants their analysis as incipient grammar. However，矶-e have also seen that only the locative constructions exhibit emergent progressive meaning. In the other two cases , pos阳re and motion verb constructions , the emergent aspectual meanings are best characterized as durative or repetitive , respectively. This demonstrates that in grammaticalization the meaning of a lexical item can be extended in different ways by speakers of different languages resulting in constructions that look ve可 similar formally. Apparently , the semantic potential inherent in posture and motion verbs can give rise to slightly different aspectual notions. Nevertheless , the aspectual meanings that were found (progressive , durative , and repetitive aspect) are all subtypes of imperfective aspect. Thus , on a more general level , the emergent constructions exhibit significant similarities to developments attested in other languages not only in form but also in meaning. This suggests that they ultimately instantiate the same general phenomenon. We can therefore conclude that processes echoing well-known cross-linguistic grammaticalization phenomena are indeed at work in English. Of course , this does not mean that the constructions noted here will necessarily grammaticalize further in the future. But although they are not predictable in their outcome or in their existence at all , phenomena like those noted here make perfect sense in the light of what is known independently about the grammaticalization processes feeding on the same lexical sources. This leaves the question of how the aspectual meanings come about in these constructions. It appears that different mechanisms of semantic extension are responsible for their emergence. In the case of the construction be in the middle 01 V-ing , the notion of containment, and , in particular, the notion of the middle of a contained region , are taken from the domain of space and mapped onto temporal events , especially actions , resulting in a highly restricted progressive meaning. Hence , the underlying mechanism is that of metaphorical mapping of SPACE onto TIME. For the other two construction types , an account in terms of conv
Emergenta，再pect
constructions in Present-Day English 151
verb when the two are used together in constructions like sit and V. Similarly, the notion ofr，叩etition can be abs位acted 企om the meaning of moving from place to place, which is inherent in verbs like go around, and projected onto the meaning of another verb in constructions like go around Ving. In the course of repeated usage , what starts out as a contextually induced extended meaning may come to be more or less routinely associated with the construction and thereby partially conventionalized, as it seems to have occurred in the case of the English pos阳re and non-directed motion verb cons住uctions. The analysis has also noted two features 也at characterize incipient grammatical constructions in general. First, such constructions are highly restricted in their applicability as markers of grammatical meanings. For example, the be in the middle 01 V-ing progressive additionally implies deep involvement in an activity. The posture and non-directed motion verb constructions apply basically only to activites performed in a certain posture or in motion. These cons位aints follow directly from 由e original meanings of the source elements which these constructions retain to a high degree. The second finding conceming incipient grammatical constructions was that speakers use the same constructions in which the grammatical meanings arise to convey other meanings , which are not always clearly sep缸'able from 由e aspec阳al notions. This can be seen, for example, in the case of sit-constructions, which are often used to give an evaluation of someone as lazy or idle. At this early stage of grammaticalization，也e constructions seem to be associated with various kinds of meanings , including 部pect， suggesting multiple motivations for their emergence. The observations on incipient grammar made in 也is paper complement those made by John Newman in his con位ibution to the present volume. Newman analyzes the ways in which experientially basic verbs, including some of the verbs discussed here such as posture or motion verbs , can appear in highly grammaticalized functions like noun classifiers, tense-aspect markers, or definiteness markers. Ultimately, the two papers are concemed with the same phenomenon, only viewing it from opposite ends of the grammaticalization path. While Newman's analysis concems the stage at which constructions involving certain basic verbs have already come to permeate the grammatical system of a language, this paper
，、 d
吨，缸
-E·E·
cIn.ms''-mn k ocpeu
structures as given "begs the question of how these linguistic structures came to be there in the first place." (1 93-218) This is the question that was pursued in this paper.
References Anderson, John M. 1973 An Essay Concerning A再pect: Some Considerations of a General Character Arising from the Abbé Darrigol 's Analysis of the Basque Verb. 白le HaguelParis: Mouton. Barlow, Michael 1998 Corpus of 年o胁， Professional American English. Houston: Athelstan. Bertinetto, Pier M. 2000 The progressive in Romance, as compared with English. In: Östen Dahl (ed.) , Tense and A再pect in the Languages of Europe, 189-225. Berlin: Mouton de Gruyter. Bertinetto, Pier M. and Denis Delfitto 2000 Aspect vs. Actionality: Why they should be kept apart. In: Östen Dahl (ed.) , Tense and Aspect in the Languages of Europe, 189225. Berlin: Mouton de Gruyter. Bertinetto, Pier M. , Karen H. Ebert and Casper de Groot 2000 The progressive in Europe. In: Östen Dahl (ed.) , 1旨nse and Aspect in the Languages of Europe, 517-558. Berlin: Mouton de Gruyter. Blansitt, Edward L., Jr. 1975 Progressive aspec t. Working Papers on Language Universals 18: 1-34. Bolinger, Dwight 1971 The norninal in the progressive. Linguistic Inqui.η2: 246-250. Bybee, Joan and Östen Dahl 1989 The creation of tense and aspect systems in the languages of the world. Studies in Language 13: 51一 103. Bybee, Joan, Revere Perkins and William Pagliuca 1994 1加 Evolution of Grammar: Tense, Aspect, and Modality 切 the Languages of the World. ChicagolLondon: University of Chicago Press. Comrie, Bemard 1976 Aspect: An Introduction to the Study of Verbal A司pect and Related Problems. Cambridge: Cambridge University Press.
Emergent aspect constructions 切 Present-Day English 153 Dahl , Östen 1985 Tense and A再pect Systems. Oxford: Basil Blackwell. Ebert, Karen H. 2000 Progressive markers in the Germanic languages. In: Östen Dahl (ed.) , Tense and A再pect in the Languages of Europe, 605 -6 53. Berlin: Mouton de Gruyter. Godfrey, John, Edward Holliman and Jane McDaniel 1992 SWITCHBOARD: Telephone speech co甲us for research and development. Proceedings of the IEEE International Conference on Acoustics, Speech , and Signal Processing 1992, vo l. 1, 517520. San Francisco: IEEE. Haugen, Einar 1982 Scandinavian Language Structures: A Comparative Historical Survey. Tübingen: Niemeyer. Heine , Bemd 1994 Grammaticalization as an explanato叩 parameter. In William Pagliuca (ed.) , Per.再pectives on Grammaticalization , 255-287. AmsterdamIPhiladelphia: Benjamins. Heine , Bemd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. ChicagolLondon: University of Chicago Press. Heine , Bemd and Tania A. Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Jespersen , Otto 1931 A Modern English Grammar on Historical Principles. Part 4: Syntax. Volume 3: Time and Tense. Heidelberg: Carl Winters. Koops , Christian 2001 Emergent Progressive A再pect Constructions: A Study of PresentDay English from a Cross-Linguistic Perspective. MA Thesis, University of Hamburg. Kuteva, Tania A. 1999 On ‘ sit' /‘ stand'/‘ lie' auxiliation. Linguistics 37: 191-213. Lakoff, George and Mark Johnson 1980 Metaphors W注 Live by. ChicagolLondon: University of Chicago Press. Langacker, Ronald 1991 Concept, Image， αnd 冉Imbol. BerlinINew York: Mouton de Gruyter. Li , Charles N. and Sandra A. Thompson 1981 Mandarin Chinese: A Functional Reference Grammar. Berkeley: University of Califomia Press.
154 Christian Koops Martin, Samuel E. 1992 AR价rence Grammar 01 Korean. Rutlan d/VermontIT okyo: Tuttle. Newman, John (ed.) 2002 The Linguistics 01 Sitting, Standing, and Lying. AmsterdamJ Philadelphia: Benjamins. Quirk, Randolph, Si也ey Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar 01 the English Language. London: Longman. Scheffer, Johannes 1975 The Progressive in English. Amsterdam: North Holland. Stefanowitsch, Anatol 2000 The English GO-(PRT)-AND-VERB construction. In: Lisa Conathan et al. (eds.) , Proceedings 01 the Twenty-sixth Annual Meeting 01 the Berkel可 Linguistic Socie纱， 259-270. Berkeley: BLS. Torres Cacoullos , Rena 1999 Variation and grammaticalization in progressives: Spanish -ndo constructions. Studies in Language 23: 25-59. Traugott, Elizabeth C. 1978 On the expression of spatio-temporal relations in language. In: Joseph Greenberg (ed.) , Universals 01 Human Language, vo l. 4 , 369-400. Stanford: Stanford University Press.
Section 111 Experiential motivation
Spatial experience, lexical structure and motivation: The case of in 巧Jvyan
Evans and Andrea
号ler
Abstract l This paper takes issue with the received view of lexical structure, which views the lexicon as being the repository of the arbitrary and the idiosyncratic. It is argued that the lexicon is systematically motivated. The traditional view of the lexicon is shown to be inadequate in three ways. These relate to the fact that a word can take on new meanings in novel contexts, that words tend to be polysemous, and that a single word can appear in a range of different lexical classes. The linguistic facts force us to conclude that, rather than being arbitrary and lacking in systematicity, lexical structure and organisation is in fact highly organised, achieving a particularly clear indication of a motivated system in the process of meaning extension. We illustrate the central claim that meaning extension is a highly motivated process, grounded in spatio-physical experience, with an analysis of the English lexeme in. Keywords: distributed semantics , meaning extension , polysemy, pragmatic strengthening, preposition, principled polysemy, proto-scene, radial category , sanctioning sense, semantic network, situated implicature, spatial scene
1. Authors' affiliation: Evans: Department of Linguistics and English Language,
University of Sussex, UK. Email: [email protected]; Tyler: Linguistics Department, Georgetown University, USA. Email: [email protected]
158
月lryan
Evans and Andrea Tyler
1. Introduction
In this paper we take issue with the received view of lexical structure, which views the lexicon as being the repository of the arbitrary and the idiosyncratic. A consequence of this position is that lexical items are arbitrarily related, resulting in a homonymy view of word meaning. Our purpose in this paper is to suggest that the lexicon, at least in the case of the closed class of prepositions, is systematically motivated. In Section 2 we argue that the traditional view of the lexicon is inadequate to account for the senses of prepositions in three ways. These relate to the fact that a word can take on new meanings in novel contexts , that words tend to be polysemous , and that a single word can appear in a range of different lexical classes. These linguistic facts force us to conclude that, rather than being arbitrary and lacking in systematicity, lexical structure and organisation is in fact highly organised, achieving its clearest indication of a motivated system in the process of meaning extension. In Section 3 we suggest that meaning extension, as it relates to prepositions, results from the complex interaction of spatio-physical experience and language use. We discuss this from the perspective of princ伊led polysemy (Evans 2004; Evans and Tyler 2004; Tyler and Evans 2001 , 200月， which constitutes a motivated account of word meaning and meaning extension. In Section 4 we illustrate one aspect of the motivated nature of meaning extension, i.e. the phenomenon of polysemy, by providing an analysis of part of the semantic network associated with the English preposition in. Accordingly , our goal in this paper is to demonstrate that the lexicon is highly motivated, and that due consideration reveals that lexical organisation and structure is highly intricate and systematic.
2. Lexical structure and motivation Traditionally, the lexicon has been viewed as the repository of the arbitrary and idiosyncratic, with all regularity and productivity associated with language taking place in the syntax (Bl oomfield 1933). This traditional view is still alive today , having been recently reasserted by Chomsky, who states , "1 understand the lexicon in a rather traditional sense: as a list of "exceptions" , whatever does not follow from general principles" (1995: 235). A consequence of this traditional position is to view the lexicon as "a static set of words and word senses, tagged with features for syntactic, morpho-
Spatial experiencιl臼ical structure and motivation 1S9 logical and semantic in岛rmation， ready to be inserted into syntactic frames with appropriately matching featur四" (Tyler and Evans 2001: 725). More recently, however, it has been realised by a range of lexical semanticists that the traditional view is simply incompatible with the linguistic facts (e.g. Brugman 1988; Brugman and Lakoff 1988; Evans 2004; Evans and Tyler 2004; Lakoff 1987; Pustejovsky 1995; Tyler and Evans 2001 , 2003). This 岛llows as there are (at least) three ways in which the traditional view of lexical organisation fails to match up with the behaviour ofwords. That is, the view ofthe lexicon as constituting a set ofstatic word senses is incompatible with the "dynamic" nature of word meaning as outlined below. First, the same word can be used in a diverse range of contexts, in which novel meanings are derived. This constitutes what we term the issue of online meaning cons住uction (Tyler and Evans 2001) or，岛llowing Sinha and Kuteva (1995) , distributed semantics (see Evans and Tyler 2003). That is , word meaning is context-sensitive drawing upon encyclopaedic knowledge as well as inferencing s位ategies which relate to different aspects of conceptual structure, organisation and packaging (see Croft and Cruse 2004; Evans and Tyler 2004; Langacker 1987; Sinha and Kuteva 1995; Sweetser 1999; Tyler and Evans 2003). For instance, consider the 岛llowing sentences (adapted from Herskovits 1986):
(1) (2)
Th ere 's someρ"Uit in the bowl. 1页ere 's a crack in the bowl.
These examples illustrate that a particular Trajector (TR) can, in conjunction with a particular Landmark (LM), affect the interpretation of a preposition. 2 On one hand, the inte甲retation ascribed to sentence (1) is that the LM, the bowl, contains or surrounds the TR, the ρ"Uit. On the 0由er hand, 世le conventional interpretation of (2) 扭曲at the TR, the crack, constitutes a flaw , which is p町t of the bowl. Additionally, the crack may either appear 2. The tenns trajector (TR) and landmark (LM) derive from Langacker's Cognitive Grammar framework. A TR is 也e focal participant in a profiled relationship, while the LM is the secondary participant. Moreover, the TR is likely to be the smaller, mobile entity, which is located with respect to the LM, which serves to locate it. In a spatial scene described by the sentence: Th e ant is in the box, the ant corresponds to the TR, while the box correspondsωtheLM.
160 Vyryan Evans and Andrea
ηler
as part of the interior or exterior of the bowl. Language users do not normally derive an interpretation for {匀， in which an entity identified as the crack is somehow located "within" the confines of the bowl , i.e. the interior space bounded by the bowl , in the same way that entities such as fruit can be. Clear1 y , our interpretation of the conceptual spatial relation denoted by a preposition, such as in , is in part constrained by sentential context, that is , by the characteristics of the entities which are designated. A crack is a dif二 ferent kind of entity from fruit. In conceptual terms , a crack is inherently relational , requiring a LM of which it constitutes a subpart, while 斤uit is a veridically distinct entity. The different status of these two items differentially affects the interpretation of the relationship designated by the preposition. In this way, the meaning assigned to the preposition is "distributed" across the sentence. The point then is that a theory of lexical structure requires a motivated account ofhow word meaning is integrated within the overall context ofthe sentence or even larger units. That is , a theory of distributed semantics is required (cf. Sinha and Kuteva 1995), a theory which recognises that utterance meaning is the result of the integration of word senses in a way which is coherent with, and contingent upon , real-wor1 d knowledge. As such utterance meaning does not reside in individual lexical items , but rather results from their interaction , which serves to prompt for integration at the conceptuallevel guided by a range of inferencing strategies (see Tyler and Evans 2001 , 2003). The second way in which the traditional view of the lexicon is inadequate relates to the phenomenon of polysemy. Under the traditional view , conventional word meanings associated with many lexical items sharing the same form are unrelated to one another (a homonymy position). Contrary to this position , we argue that lexical items do not act like static bundles of features , but rather evolve and change, such that new senses are derived from pre-existing senses. Hence , lexical items are modelled as constituting categories made up of distinct but related senses (Evans 2004; Lakoff 1987: Taylor 2003; Tyler and Evans 2001 , 2003). Influential studies such as Brugman (1 988) and Lakoff (1 987) , which examined the English preposition over, have demonstrated that senses as diverse as ‘ above' , e.g. The pic ‘
Spatial 臼;perience. 1，臼 ical structure and motivation
161
differential interaction with other sentential elements, so the inventory of word senses conventionally associated wi由 a lexical item can change, with new word senses evolving. Hence , a 也eory of lexical organisation requires a motivated account of the experiential and conceptual factors 也at facilitate the derivation of new conventional senses, and thus an account of the nature and origin ofpolysemy. The third way in which the traditional view of the lexicon is inadequate relates to the phenomenon of the p町t of speech of lexical items. Traditionally, lexical items have been viewed as being tagged with a range of lexical, semantic and syntactic features. Hence, a word such 臼 in， 岛r instance, is tagged as belonging to the lexical class of pr'叩ositions. However, in can also appear in a range of different syntactic configurations in which it is clearly not preposed with respect to a noun phrase. This is illustrated by its adverbial usage in (3时， with the copula, and its usage as p缸t ofa verb p町 ticle construction (or phrasal verb) in (3b):
(3)
a. Th e sun is in. b. Th e Uttle boy went in (for his supper).
Due to its view of word meaning as being static sets of features , the traditional approach is forced to conclude 由at a difference in lexical class is evidence for distinct lexical items. However, such an account fails to recognise that the meanings associated with in in a prepositional use such as (1) , for instance, and the usages in (3a) and (3b) appe缸 to be strongly related.
3. Meaning extension as a motivated phenomenon h 也is
section we address the issue ofpolysemy，由e phenomenon whereby a single linguistic form is associated with a number of related but distinct
senses.
3. 1.
Polysemy
In order to illustrate the phenomenon of polysemy consider the following examples:
162
(4)
月'ryan
a. b. c. d.
Evans and Andrea
吵ler
The puppy is in the box. She is in love. Ok, class, put your chairs in a circle. She cut the pie in half.
Each of these uses of in appears to convey a different meaning. In (4a) in appears to relate to the notion of containment by a three-dimensional LM. In (4b) in designates a particular state. In (4c) , in relates to the notion of a boundary as shape , while in (4d) in designates that an entity is divided. There are a number of reasons for concluding that these uses of in constitute related , and hence , polysemous word senses; that is , there are good reasons for assuming that the many distinct senses conventionally associated with a particular lexeme result from a motivated process. It is perhaps self-evident that an important function of language is communication. Moreover, communication is fundamentally purposeful (Gumperz 1982). This fact places certain non-trivial constraints on the use of lexical items. It seems obvious that a speaker intending to communicate , and hence achieve the desired purpose , would not use a lexical form with one established meaning to indicate something else , unless the speaker assumed the listener could readily work out the novel usage. In order for a novel use to be readily interpretable by the hearer, meaning extension must be somehow constrained and systematic. This strongly suggests that when a speaker uses a form with an established meaning to indicate something other than the conventional meaning , the choice of which lexical item to select is motivated. If this were otherwise , the speaker could not assume that the listener had a reasonable chance of interpreting the novel use. This line of reasoning suggests that there must be something about the conventional meaning associated with the lexical item that led the speaker to choose that lexical form rather than some other. Moreover, it is reasonable to assume that at an earlier stage in the language , a fonηsuch as in had fewer distinct, conventionalised meanings associated with ie; thus , many of the uses now conventionally associated with the form at one point represented novel uses. The homonymy approach begs the question of why it should be the case that a speaker would choose to use a particular established form in a novel way , rather than coining a new phonological string altogether.
3. This point is also made by Sweetser (1 990).
Spatial experience, lexical structure and motivation 163
Given the argument we have been developing , the synchronic semantic network associated with a lexical item is understood as a historical product. This contrasts with the traditional view, which assumes that the lexicon is the repository of the arbitrary and the idiosyncratic and hence that distinct meanings within a semantic network are arbitrarily related. As a result, the homonymy approach makes the implicit claim that the process of meaning extension itself is arbitrary, leading to the unsatisfactory conclusion that language change is ad hoc , lacking motivation. This contradicts the view that language evolution is a systematic process, as revealed by the voluminous grammaticalization literature (e.g. , Bybee et al. 1994; Hopper and Traugott 1993; Heine et al. 1991 for overviews, summaries and references). Finally, the homonymy approach fails to explain the ubiquity of the phenomenon of meaning extension. While we readily acknowledge that the "accidents" of history have resulted in instances of homonymy, the failure to account for the considerable systematicity that does exist misses important generalisations. As the homonymy approach fails to recognize that distinct meanings may be motivated and, hence , at some level systematically related, we are forced to conclude that it is inadequate.
3.2.
The actuation issue
In view of the proliferation of distinct senses associated with a lexical item such 邵阳， the actuation issue concerns accounting for the cause(s) of such a proliferation. We propose an experientialist view, suggesting that in our interaction with the socio-physical environment, certain spatial relations have non-创vial consequences, which in turn give rise to situated inferences. For instance , by virtue of a TR being within a container, the TR is thereby located with surety. For example, if you move a coffee cup, the coffee moves with the cup (unless , of course, you spill the contents). Hence , in this instance containment correlates with location. It has been observed by a number of scholars that si阳ated inferences deriving from experiential consequences, through continued usage, come to be conventionally associated with the lexical form identified with the situated inference (e.g. , Bybee et al. 1994; Evans To appear; Hopper and Traugott 1993; Fleischman 1999; Svorou 1994; Traugott 1989; Tyler and Evans 2001 , 2003). Following Traugott, we term this process pragmatic strengthening. This process results in the association of a new meaning component with a par-
164
月呵lan
Evans and Andrea 乃ler
ticular lexical fonn through the continued use of the fonn in particular contexts in which the situated inference results. That is , new senses derive 仕om the conventionalisation of inferences through usage 阴阳ms. By way of illustration, consider the following examples: (5)
a. She is in the prison. b. She is a prisoner. c. She is in prison.
The sentence in (5a) designates a scene in which the TR, she , is located in a bounded LM , prison. The pu甲ose of bounded LMs of this kind is to restrict the 仕'eedom of the inmates. Hence , the state of being a prisoner, described in (5b) , is tight1y correlated in experience with being 10cated within a p町ticular kind of bounded LM. Thus , it is the context itself (via inference and our knowledge of the real world) which provides the implicature of a particular state being associated with a particular bounded location. If an experientially-motivated inference is recurring, it can be reanalysed as distinct from the scene of which it is a part. Through continued use , this process may lead to the strengthening or conventionalisation of the inference, resulting in its development as a distinct meaning component associated with the lexical fonn with which it is related , i.e. in. As a consequence, in has, in addition to its "containment" meaning in (5时， a conventional State Sense associated with it, as illustrated by (5c). Indeed, this sentence could be a applied to a prisoner outside the prison on day release, while (5a) could no t. Once instantiated in semantic memory this additional sense can be employed in new contexts of use unrelated to the context that originally gave rise to it. Thus , in has developed a conventionalised State Sense where the original spatial configuration which initially gave rise to the situated implicature is no longer required, as evidenced by examples of the following kind which employ an ab由actLM:
p町ticular
(6)
a. We're in a state ofwarlemergencylholy matrimonylmartial lawlanarchy. b. She looked peac，矿ùl in death. c. They're always getting in trouble.
牛atial e.冗:perience， 1，臼 ical structure
3.3.
and motivation 165
The model1ing issue
The Modelling issue concerns how the (synchronic) polysemy exhibited by a particular lexeme should be model1ed. Fol1owing scholars such as Lakoff (1987) and Taylor (2003) and the results ofpsycholinguistic studies such as Rice et al. (1999) and Sandra and Rice (1995) , we will assume that lexical items constitute lexical categories, consisting of form-meaning pairings. The semantic pole of the form-meaning pairing we model in terms of a semantic network, organised with respect to a primary or sanctioning sense. Our sanctioning sense constitutes the synchronic sense which language users intuitively feel best constitutes the "most central" meaning associated with a particular lexical item (discussed further below). However, as the historical1y earliest attested meaning may still play an active p町t in the synchronic network associated with a preposition such as in , the historical1y earliest sense and the sanctioning sense may overlap.4
4. In terms of synchronic polysemy networks , the empirical work by Sandra and Ri ce (1995) suggests that it may not be the case that a particular lexical form has a single sanctioning sense by virtue of which language users categorise all other senses associated with a lexical item. Thus , their empirical work raises questions conceming the view that we can define polysemy as a strictly s泸1chronic phenomenon in which there is a relationship which speakers are consciously aware of holding between distinct senses of a particular lexical form. This is an empirical question for which we do not yet have sufficient evidence to address. If extensive experimental evidence shows that language users systematically and consistently fail to perceive some senses as being related, then we must call into question that what we are terming polysemy constitutes a phenomenon that is wholly synchronic in nature. While we believe all the senses in a particular semantic network are diachronically related, in terms of the adult lexicon, there may be differences in the perceived relatedness between distinct sets of senses , due to routinisation and entrenchment, obscuring the original motivation for the derivation of senses from pre-existing senses such as the primary sense for language users (see Rice et al. 1999, in particular). Hence, one of the reasons we term our approach principled polysemy is to reflect the view that due to processes of language change , not all senses associated with a particular phonological form may be recognised by a language user as being synchronically related. That is , while meaning extension is highly motivated, it may result in a semantic network, which may appear, to the language user (and perhaps also the linguist) , to be only partially motivated.
166
吵ryan
Evans and Andrea Tyler
The idea behind positing a primary or sanctioning sense is that language users appear to intuitively categorise senses with respect to some lexical "model" or lexical "theory".5 A word's semantic network, i.e. the range of conventional senses associated with it, can be modelled or organised with respect to the sanctioning sense. As we will diagram the semantic network for in as a radial-like structure (see Section 4) , we will follow the practice of referring to semantic networks organised with respect to a sanctioning sense as a radial category (Lakoff 1987). One advantage of modelling a lexical category in terms of a radial category is that this facilitates understanding degrees of relatedness between senses, and accounts for the appearance of chaining within categories (see Lakoff 1987). That is , while some senses will appear to be more closely related to the sanctioning sense , other senses may appear to be more closely related to other derived senses. This pa位ern of clustering suggests possible paths of derivation (see Tyler and Evans 2001), and provides predictions that can be assessed against what is known about the diachronic development of word senses from the historical record. In spite of the foregoing , we are not claiming that the sanctioning sense will necessarily be the same across a community of speakers. Patterns of entrenchment may vary from language user to language user. However, as prepositions appear to relate at some level to spatial relations , we hypothesise that the sanctioning sense associated with a particular preposition can be modelled in terms ofwhat we term aproto-scene. We will define a proto-scene (the unique sanctioning sense for a particular preposition) as a highly abstract representation of a recurring spatial configuration between two (or more) objects , a spatial scene (see Tyler and Evans 2003; see also Evans 2004; Evans and Green To appear). The protoscene for in will be presented in the next section.
4. The case of in In this part of the paper we turn to a consideration of the polysemy network associated with the English spatial marker in. Our pu甲ose here is quite limited, and for that reason we do not present an analysis of the entire semantic network associated with in. Such an analysis would necessitate a 5. An exemplification of this is lexicographic practice, which ranks word senses ,
often numerically.
Spatial 臼;perience，
lexical structure and motivation 167
paper many times longer than this. Moreover, we do not seek to adduce "decision principles" for determining distinct senses or the sanctioning sense, we have presented these elsewhere (see Tyler and Evans 2001 , 2003). Rather, in keeping with the theme of this volume, we present the following in order to support our contention that meaning extension in the mental lexicon, the phenomenon of polysemy, is highly motivated in nature. In specific terms we seek to show how the proto-scene gives rise through tight correlations in experience leading to si阳ated inferences 一 to derived meanings, which come to be conventionalised as distinct senses via pragmatic strengthening. That is , we are addressing what we termed the actuation issue. The semantic network for in has a large number of distinct senses associated with it; we have identified nearly thirty, which appear to be arranged into groupings or clusters derived from related experiences. In order to demonstrate the highly motivated nature of the process of meaning extension we will illustrate contrasting senses, which are nevertheless representative of each cluster. 6
4.1.
The proto-scene for in
The proto-scene for in constitutes a spatial relation in which a TR is located within a LM which has three salient structural elements, an interior, a boundary, and an exterior. In addition to the spatial relation designated, the proto-scene 岛r in is associated with the functional element of containment (see below) - this is a consequence of the structural elements of the LM , the fact that the TR is located within the interior portion of the LM , and the real-world force dynamics associated with containment. The proto-scene 岛r in is given in Figure 1. The LM is designated by the bold lines while the TR is designated by the dark circle.
6. In and its Indo-European cognates (particularly in French, Dutch and Gennan) have been extensively studied (e.g. , Dirven 1993; Hawkins 1988; Herskovits 1986, 1988; Ho忧.enroth 1993; Lindstromberg 1998; Miller and Johnson-Laird 1976; Quirk et al. 1985; Vandeloise 1991 , 1994). Despite having been so well studied, the present analysis represents , we suggest, the first that provides a me出odologically motivated account for the range of polysemy associated with
zn.
168
吵ηlan
Evans and Andrea
ηler
|牛| Figure 1. Proto-scene for in Linguistic examples ofthe proto-scene include the following: (7)
a. John is in the house. b. The kitten is in the box.
4.2.
The functional nature of containment
As intimated above , the semantics of in cannot be solely equated with a particular set of spatio-geometric properties. In addition , prepositions often designate a functional element (see Evans and Tyler 2004; Tyler and Evans 2003). This arises as a consequence of our continued and ubiquitous interaction with spatial scenes involving certain spatial configurations. Containment itself is a complex relation involving numerous functional consequences. In the guise of containers, bounded LMs constrain and delimit movement of their TRs , as in the everyday example of a coffee cup which constrains the coffee it contains to a specific location, namely the confines of the cup , or a prison cell , which restricts the movements of a convict. In certain circumstances , constraining movement can be understood as providing support, thus a cut flower can be held in an upright position as a result of being placed in a vase. If the boundaries of the container are opaque, they prevent us from seeing beyond them, or the interior area from being seen by entities outside, as in a walled garden or a windowless room. Containers can also provide protection, as with a jeweller's safe. For the elements within a container, the container surrounds and largely determines the environment in which those entities exis t. Different aspects of the experience of containment are profiled by the various uses of the prepositions in and out.
Spatial experience, lexical structure and motivation 169
A sec'Ond c'Onsequence 'O f 'Our interacti 'On with b 'Ounded LMs is that they can serve as g'Oals. F'Or instance, after leaving w 'O rk, f'Or many pe'Ople the g'Oal is t 'O arrive at h'O me in 'O rder t 'O interact with family , relax , etc. The salient space in which these anticipated activities take place is a b 'Ounded LM , the living quarters; thus , being in the salient space, the b'Ounded LM, is c1'O sely related t 'O achieving g'O als. A third way in which we interact with b 'Ounded LMs is that we emerge 企'Om them, as when we leave h 'O me each m 'O ming. Equally, we draw 'Other entities from b 'Ounded LMs , as when a jeweller withdraws a diam'O nd br'O ach from a safe. In this way , b 'Ounded LMs have s'Ource pr'Operties, as lexicalised by out of A 岛urth way in which we interact with b 'Ounded LMs is when we enter, pass fr 'Om 'O ne side t'O an'Other and exit, as when walking 仕'Om r 'O'O m t'O ro 'Om in a bui1ding. Thus , we can experience a b 'Ounded LM as a passage-way with a s'Ource l'Ocati'On and a highlighted end-p 'O int ('O r g'Oal). This relati 'On is capωred by through. Vandel 'Oise (1991 , 1994) has been 'One 'O f the m'O st f'O rthright adv'O cates f'Or the view that prep'O siti'Ons cann'Ot be equated s'Olely with spatial elements. C'O nsider the diagram pr'Ovided in Figure 2.
Figure 2. A bottle or a light-bulb (after Vandeloise 1994: 172)
Vande1'Oise 'Observed that an image such as that depicted in Figure 2 c'Ould be c'Onstrued as being either a b 'Ottle 'O r a light-bulb. H 'Owever, while we can felicit 'Ously describe the relati 'On between the light-bulb , the TR, and its LM , the socket, as in (8) , (8)
The bulb is in the socket.
we cann'Ot felicit 'O usly describe the relati 'On between a b 'O ttle and its putative LM , the cap , in terms 'O fthe spatial re1ati 'On designated by in: (9)
??The bottle is in the cap.
170
月Iryan
Evans and Andrea Tyler
Vandeloise points out that, as the spatial relation holding between the TR and LM in each of these sentences is identical , and yet (8) is fine while (9) is semantically anomalous , a spatial-geometric configuration alone cannot account for the relation described by in , or any other spatial particle. He argues that the relevant factor accounting for the contrast in acceptability between sentences (8) and (9) is functional , "[W]hile the socket exerts a force on the bulb and determines its position, the opposite occurs with the cap and the bottle." (Vandeloise 1994: 173). Put another way , not only is the position and hence successful functioning of the bulb contingent on being in (i.e. , contained by) the socket, the socket also prevents the bulb from falling to the ground and thus provides a constraining or supportive function. In contrast, the position and successful functioning of the bottle is not contingent on being in the cap.
4.3.
Non-canonical bounded LMs
Because of the flexibility of human conceptualisation in can be employed to designate spatial relations in spatial scenes that do not involve canonical three-dimensional LMs. For instance , in the following examples the LM is concep阳alised as being physically planar, and hence two-dimensional. By virtue of such LMs being construed as possessing an interior (and thereby a boundary and an exterior) , these LMs are conceptualised as bounded, licensing the use of the proto-scene 岛r in: (10)
a. The cow munched grass in thefield. b. The tiny oasis flourished in the desert.
In (1 0a) we conceptualise the cow as being "contained" by the field. Yet, the field is not a canonical three-dimensional LM. By virtue of a field having an interior (that part which constitutes the fiel哟， a boundary such as a track, road , fence or hedge which marks the perimeter, and an exterior (that part which is not the field), this particular spatial scene can be construed as involving a bounded LM and hence a containment relationship. It might be objected that fields containing cows and other livestock are often bounded with barriers such as gates , fences , or hedgerows which specifically constrain movement and thus this spatial scene does involve a canonical , threedimensional LM. Notice , however, that the LM appears to be conceptualised as bounded even in examples in which there is no physical impediment
Spatial experience, lexical structure and motivation 171 delimiting movement, as evidenced in (lO b). That is , by virtue of there being a LM which can be conceptualised as having an interior which contrasts with an exterior, a boundary is entailed and a concomitant designation of containment arises. Hence, in (10时 ， the oasis , the TR, is concep阳 alised as being "contained" by the desert , the LM, even though there are no physical barriers such as fences bounding the desert. 7 Analogously, in mediates the spatio-functional relation in spatial scenes involving continents, seas, countries, regions , provinces, and other geophysical divisions , such as cities, etc. (11) a. China is in Asia. b. London is the larg，臼 t city in the c. She lives in New York City.
United}乙ingdom.
In also denotes spatial scenes in which a prevailing atmospheric condition is conceptualised as enveloping the TR:
(12) a. Thejlagjlapped in the wind. b. The child shivered in the cold. c. The rabbit斤。'ze in the glare ofthe car's headlights. Other spatial scenes involving the proto-scene for in include examples such as the following, in which a collective of individuals is concep阳alised as a single bounded entity. Consider some examples:
(13) a. The child couldn 't be seen in the crowd. b. The old cottage was located in the wood. Langacker (1987) observes that collective entities, characterised by nominals such as crowd, team , etc. , profile the interconnections between the individuals that constitute the collective. Due to the perceptual phenomenon of closure, a collective of individuals may be conceived as constituting a single bounded entity. Moreover, in perceptual terms collectives can also be perceived as constituting single bounded entities. For instance, while up close to a crowd of people we perceive the individuals who make up the crowd. If we move away so that the collective is increasingly distant, a 7. Of course , as deserts are relatively inhospitable, they often constitute barriers to travel , expansion of settlements, etc.
172
月呵lan
Evans ù..阮
point is reached at which we perceive a bounded entity, a ‘ crowd' , without perceiving the individuals who make up the crowd. Due to phenomena of 也is kind, collective entities such as people can be perceived and hence conceptualised as a single bounded entity, a ‘ crowd\ Once this has occuπed， the LM , the crowd, can be construed as occupying a bounded space. Thus, such a single entity can be conceptualised as a bounded LM, possessing not only a boundary but also an interior and an exterior. Moreover, our experience of being part of a crowd often involves a number of the functional aspects of containment such as the sense of having our movements constrained, of being surrounded, of our view being obstructed, etc. Thus , the use of in is licensed in such situations.
4 .4.
Beyond the proto-scene
Due to the ubiquity of bounded LMs in our everyday experience and the range and differences in such LMs , it is hardly surprising that we interact with bounded LMs in many different ways. This is re f1 ected by the complexity of the polysemy network associated with in , and the range of senses associated with it. In this section we will provide a brief overview of a limited range of the senses associated with each of the five clusters of senses derived from the proto-scene. Each cluster re f1 ects the different configurational and functional elements associated with a bounded LM , whose functional element is containment. Figure 3 presents a summary of the senses and clusters of senses in the semantic network for in , which we will deal with.
4.4.1. The Location Cluster ρj
One aspect of the notion of containment relates to the movement or action of an entity being constrained by vi阳e of being enclosed by a bounded LM. Johnson (1987) has argued that a consequence of this imposition is that the location of a contained TR is determined by the location of the bounded LM. In other words , the bounded LM serves to locate the contained TR with surety. Take the example of an infant in a playpen - move the playpen and the infant thereby follows suit. Indeed, the contained TR is located with surety even ifthe TR is not perceptually accessible. For example, the location of the infant in the playpen is known , even if the infant is
Spatia/ experience, /，臼ica/ structure and motivation 173
not immediately within sight, as when the parent steps out of the room. Thus , there is a strong experiential correlation between a TR being contained and being located with surety. Activity (2 .3) Perceptual Accessibility (3.1)
State (2 .2)
•
-
In Situ (2.1)
ISEXTEivρSEGMEF丁FLEXIVγ 、、 y
、
..
v且
cnu
e
D-1,,‘,,mA4. a anpue ny--
Shape As Boundary (5.1)
Blockage (5.2)
Reflexive (6.1)
Figure 3. A Partial Semantic Network for in (Note: Clusters of senses are indi-
cated by un-shaded circles. Distinct senses are indicated by shaded circles.) Experimental work in child language acquisition (E. Clark 1973) also suggests a strong conceptual relationship between location and containment, and that young children perceive containers as default locations. In the first step of an experiment (E. Clark 1973), the experimenter asked 18 month olds to "Do this" as a block was placed in the child's cot. The children mimicked the experimenter's placement of the block perfectly. In the following steps, the experimenter again said "Do this" but placed the block beside the cot, then under the cot. In these conditions, the children still placed the block in the cot. In other words, given a moveable object, (a TR) , a container (a LM) , and the task of locating or placing the object, these young children seemed to construe containers as naturallocators. In the Location Cluster of senses, the notion that a bounded LM serves to pick out the salient space which contains the TR is privileged, and gives
174
吵吵'an
Evans and Andrea
ηler
rise to a range of c1 0sely related senses. We will briefly survey four of these: the In Si阳 Sense， the State Sense, the Activity Sense and the Means Sense. The In Situ Sense (2.1). An experiential correlate of being located with surety is that the TR crucially remains in a particular location. The conventional interpretation that the TR remains co-located with the salient space designated by the LM for an extended period provides additional meaning not apparent in the proto-scene. Moreover, this designation suggests an attendant pu甲ose for being so located. This sense is illustrated by the following examples: (1 4)
W加t are you infor? [asked in a hospital = "What's wrong with you?" or a prison = "What were you convicted o f?"] b. He stayed in for the evening. c. The workers staged a sit-in.
a.
In (1 4a) the TR is located at the LM (the hospital or the prison) for an extended period of time and for a particular pu甲ose. Notice that this question would not be felicitous if the addressee were c1 early at the hospital or prison for a brief visit , even though the addressee were physically located within the building, patent1 y a bounded LM. 8 In (14b), the TR，加， remains located at home 9, rather than, for instance. going to a nightc1ub or some other location. In (1 4c) the TR, the workers. remain at their place of work, refusing to leave in order to protest. In each of these examples , the TR remains located for an extended period, for a particular pu甲ose an d/or due to a volitional act or event. As such , the correlation between a TR being located with surety within the LM and the TR remaining at the location in question for an extended period , an In Situ Sense has become associated with in.
8. The appropriate question for the short term visitor is something like f同 atÆ悦。 have you come in for? placing emphasis on the act of coming , rather than state ofbeing in. 9. Lindner (1981) points out that in has developed a "special" sense of the TR being in the default location. For a person, we often think of the default loca. tion as the home.
Spatial experience, lexical structure and motivation 175
The State Sense (2.2). Grady (1997) and Lakoff and Johnson (1999) have argued that a primary metaphor, one which is based on a common correlation in experience, involves a particular location (LM) and the state experienced by the entity (TR) , or the particular situation that the TR happens to be experiencing. For instance, the infant (TR) sirting on the parent' s lap , enclosed in the p町ent's arms (LM) , will often experience a sense of security and love. Some young children experience a sense of isolation and fear when left alone in a dark room at night. Through recurring instances of a particular emotional state being experienced in a specific locale , the correlation between location and emotional and/or physical state becomes established. This correlation gives rise to conceptual associations such that we conceptualise and hence lexicalise states in terms of location. \0
pa此icles， such as on and at, can also denote State Senses. These State Senses may derive from the ubiquitous experiential correlation between states and location. However, there appear to be systematic, subtle differences in the properties of the states denoted by each preposition and thus, in many cases, we can identify the motivation for conventionalised uses of particular prep'O siti'Ons with particular states. For instance , an imp 'Ortant m'Otivati 'On f'Or c'Onventi 'Onal uses of in seems t'O relate t'O the constraint in extracting oneself fr'Om 'Or placing 'Oneself in the state, i.e. State Sense uses 'Of in 'Often draw 'On the n'O ti'On 'Of c'Ontainment. F'Or instance , English speakers seem t'O conceptualise an em'Oti 'Onal state such as l'Ove in such a way that 'Once 'One is "in" l'Ove , 'One cann'Ot easily 'Or v'Oluntarily leave that em'Oti'O nal state; similarly, 'Once 'One is "out 'O f' l'Ove with a particular pers 'On , it is difficult t'O get back "in" the state 'O f being in l'Ove. Similarly, if 'One is "in 仕'Ouble"， it is 'Often a state n'Ot easily escaped. In c'Ontrast, states such as being "'O n the take" 'Or "'On the pill" are 'O ften seen as being a ch'O ice, and hence p'Otentially m'Ore easily "escaped". This is c'Oherent with the prot'O-scene f'O r on , which d'Oes n'O t inv'Olve b'Oundaries or c'Onstraints 'On m'O ti 'On. M'Ore 'Over, the State Sense den'Oted by "'O n" 'Often inv'Olves s'Ome sense 'Of supp'O rt, which is als 'O c'Oherent with the prot'O-scene. Anal 'Og'Ously, the use 'Of "at" in a State use such as "at war" may be due t'O the 'Opp 'O siti 'Onal sense ass 'O ciated with "at", e.g. , He rushed at me. We suggest that the meanings ass 'Ociated with these examples reflect nuances from a number 'Of senses designated by each prep 'O siti'On, which is expected if a semantic netw'O rk is a semantic c'O ntinuum. Thus , although we hyp 'O thesise that all State Senses are m'O tivated by the c'O rrelati 'On between l'Ocati'On and state , we als 'O suggest that there are nuances, such as "b'Oundedness" 'Or "c'O nstraint", ass 'Ociated with uses , such as "in l'Ove", "in 衍'Ouble"， etc. , which relate t'O the semantic netw'O rk f'O r the particular prep'Ositi'O n in questi'On.
1O .0ther spatial
176
吵ryan
Evans and Andrea Tyler
As noted earlier (recall the examples in (5) and the related discussion) , in can be employed with certain states which are conceptualised as constraining the TR or posing difficulty in leaving , or with situations which can be conceptualised and hence lexicalised as states, e.g. We 're in a state of war/emergenc}咐 oly matrimony/martiallaw/anarchy, etc. The reason for this is that there is a tight correlation between being located in a bounded LM and a particular state which is conferred by virtue ofbeing so located. A consequence of the State Sense being instantiated in permanent memory is that in can denote relations between TRs and non-physical LMs. This is because in this sense in denotes relations between TRs and states, rather than bounded LMs per se. This follows as once a particular sense is instantiated in memo可 it can be employed in contexts of use absent those that originally motivated it. The Activity Sense (2.3). In addition to the tight correlation between bounded LMs and states , noted in the foregoing , there is a similarly tight and ubiquitous correlation between a particular activity and the bounded LM in which the activity occurs. For instance , a the position of a staff person who works for a govemment official might conceivably be described in the following way in response to a question as to the TR's profession (1 5)
A: Wh at's his line ofwork? B: He's in the governor's office. [= works for the govemor]
This example nicely illustrates the correlation between an activity and the bounded LM at which the activity takes place. Working for the govemor However, in some contexts more than one spatial preposition might conceivably constitute a ‘ best-fit\Accordingly, while two partic1 es may be motivated , a language or dialect may choose to conventionalise one while another may conventionalise another. An example of this is the distinction between British and American speakers. One of us is American and the other British. For the speaker of American English it feels "more natural" to describe the situation in which a female dog is fertile as: The neighbor's dog is in heat. For the speaker of British English it feels "more na阳ral" to describe the same situation as: The ne也hbour 's dog is on heat. Indeed, both partic1 es are motivated. A state of "heat" is relatively-speaking short, hence on; yet the animal cannot voluntarily escape this state , thus motivating the use of in. Hence, while both in and on are motivated, in this context which one is selected is a matter of conventionalisation.
Spatial experience, lexical structure and motivation 177 takes place in a bounded LM referred to as the govemor's office(s). Hence, the location can stand metonymically for the activity. A consequence of the correlation between activities and bounded LMs is that the notion of an activity can come , through pragmatic s仕engthening， to be reanalysed as a distinct meaning associated with in. Once instantiated in semantic memory , in can mediate a relation between a TR and an activity even when the activity designated is no longer overtly associated with a particular bounded LM. Consider some il1us位ative examples of the Activity Sense:
(1 6)
a. He works in stocks and shares. b. She 's in medicine. c. Th句) are in [the manufacture of] expensive baby clothes. d. She 's in graduate school. 11
11. At this point it is worth contrasting the principled polysemy approach to the concepωal metaphor approach. In their early study of conceptual metaphors , Lakoff and Johnson (1980: chapter 7) suggested that in an example such as: She 's in love , what they termed an ontological metaphor licensed the use of an "abstract" concept, such as love, being conceptualised as a container. The present approach suggests that it may be misleading to posit that a native speaker is somehow understanding a concept such as love as a bounded LM which physically contains the TR. Rather, due to a complex semantic network, in has associated with it a conventional State Sense, which licenses such uses. Th is sense, while conventionally associated with in , is derived uItimately from a tight correlation in experience between spatial regions, which are conceptualised as being bounded, and states (see Grady's 1997 ground-breaking work on experiential correlation). Hence , while states are not straightforwardly conceptualised as three-dimensional containers, as is sometimes implied in the conceptual metaphor literature, it is worth emphasizing that the present principled polysemy approach does maintain that the polysemy associated with in (and with other spatial partic1 es) is firmly grounded in spatio-physical experiential correlates. While we hypothesise that once distinct senses are conventionalised, they no longer straightforwardly draw on the experiential correlations that gave rise to them, these same experiential correlations, as well as coherent perceptual resemblances, are sti11 an active part of human experience and, hence, remain available for the creation of novel utterances. Moreover, we acknowledge Lakoff and Johnson's (1980) important insight that metaphorical pa忧ems tend to be consistent or coherent throughout the language. The fact that emotional states and atmospheric conditions are both associated with a bounded region
178 Vyryan Evans and Andrea
ηler
The Means Sense (2.4). As we have just noted in the foregoing discussion , locations and activities are strongly correlated in experience. One resu 1t of the conceptual association established between activities and locations is reflected in the lexical pa忧em of expressing aspects of activities with the parti c1 e in. Moreover, it is often the case that a particular activity utilises one (or more) particular means in order to be accomplished. Due to the tight correlation in experience between an activity and the means of accomplishing the activity , in has developed a distinct Means Sense. This has been possible precisely because in had an antecedent Activity Sense associated with it. Consider some examples of the Means Sense.
(1 7) a. She wrote in ink. b. He spoke in Italian. In (1 7a) in denotes the relation between a particular activity and the means of accomplishing the activity , ink versus pencil , for instance. Similarly, the activity in (17b) is accomplished through the medium of Italian rather than Japanese, say. The correlation between activities and their means of accomplishmem has led, then, through pragmatic strengthening to the development of the Means Sense. This provides an elegant i11 ustration of the way in which a preposition, through the development of conventionalised senses (e.g. the Activity Sense) , and recursive experiences correlating with these derived senses (e.g. activities correlate with means of accomplishment) , can give rise to further senses (e.g. the Means Sense). In addition , the means used to accomplish an activity can be conceptualised as crucially influencing or constraining the activity. This conceptualisation is coherent with the notion of containment present in the protoscene for in. Retuming to the example in (17时， the activity of writing is constrained by the means used to accomplish the writing , i.e. by the writer using ink. For instance, the writing is not easily erased so the writer must take certain additional care in the process.
and that emotional states are conceptualised (through perceptual resemblance) as weather conditions , as in They're in a stormy relation or She ￥切 a bright. sunny mood, form a coherent , reinforcing pa忧em of conceptualisation.
牛atial
experience, lexical structure and motivation 179
4.4.2. The Vantage Point is Interior Cluster (3)
Up t'O this p 'O int, the spatial scenes ass 'Ociated with in that we have expl'O red have all assumed an "'Off二stage" vantage p'Oint. H'Owever, there are spatial scenes in which the vantage p 'Oint can be l'Ocated "'On-stage" , i.e. within the spatial scene being c'Onceptualised. This has pr'Of'Ound implicati 'Ons f'Or h'Ow the scene is viewed and, c'O nsequently, f'Or senses derived fr'O m such a viewing arrangement (see Langacker 1987). In spatial scenes inv'Olving a b'Ounded LM , 'One 'Obvi'Ous vantage p'O int is interi'O r t'O the b'Ounded LM. The interi'O r vantage p'Oint gives rise t'O a number 'Of distinct senses, 'O ne 'O f which we detail bel'Ow: the Perceptual Accessibility Sense. F'O ll 'Owing w'Ork by disc 'Ourse analysts such as Schiffrin (1 992), we hyp'O thesise that taking the interi'O r perspective can c'O incide with a shift in deictic centre 'Of the scene (what we are terming the vantage p'O int). Taking the interi'O r regi 'On as the vantage p'O int c'Onc 'Omitantly highlights the perspective 'Of the TR. In this secti 'O n we als'O discuss tw'O further senses, the In Fav'Our Sense and the Arrival Sense, in which the TR within the interi'O r regi 'On is the vantage p'Oint 仕om which the scene is viewed. Hence , while in the Perceptual Accessibility Sense the vantage p 'Oint is within the b'Ounded LM but distinct fr 'Om the TR, in the In Fav'Our and Arrival Senses, the vantage p 'Oint and TR c'O incide. The Perceptual Accessibility Sense (3.1). A c'O nsequence 'O f the experiencer and vantage p'Oint being l'Ocated within a b 'Ounded LM is that TR(s) and interi'O r envir'O nment c'Ontained by the LM are available t'O the experiencer (and hence vantage p 'O int) by virtue 'O f his 'O r her sense-percept'Ory apparatus. Take visi 'On f'O r instance. Unless we p'O ssess X-ray visi 'O n à la Superman，岛r the c'Ontents 'O f a b'O unded LM (such as a cl 'O sed ro 'Om) t'O be visible t'O us , we must 'Ordinarily be l'Ocated within the b'Ounded LM. This is a significantly different 'Occurrence than that which takes place when the experiencer is l'Ocated 'Outside the b'Ounded LM. When the experiencer is l'Ocated exteri'Or t'O the b'Ounded LM , the interi'Or regi 'On and TR(s) tend n'O t t'O be accessible. At the same time , when the experiencer is l'Ocated interi 'O r t'O the b'Ounded LM , the limits 'Of the LM and the limits 'O f perceptual accessibility tend t'O c'O incide. F'O r instance, if we are in a r'O'Om , all we have visual access t'O is in that r'O'O m. Again, this reflects a substantially different c'Onsequence from being l'Ocated exteri'Or t'O the LM. Due t'O the tight c'Orrelati'On between the c'Onstruer being l'Ocated within the b 'Ounded LM , and the TR and interi'Or space delimited by the b'Ounded LM being available t'O the
180
吵ryan
Evans and Andrea Tyler
experiencer, in sense-percept。可 tenns such as visibility , in has derived a Perceptual Accessibility Sense. Consider some examples: (18)
a. b. c. d. e.
1 have it in view. 1 have him in sight. 1 stayed (with)in earshot 01 baby M饵 's cry. Thoreau always stayed in range 01 his mother 's dinner bell. Susan always tries to stay in touch.
In the sentences in (1 8) , in denotes a relation between a TR and senseperceptory availability with respect to a particular experiencer (= the vantage point). Notice that the sensory perception which is available is delimited by the LM , the bounding element. We suggest that this linguistic coding reflects the lived experience of the boundaries of the LM placing limits on perceptual accessibility when the experiencer is located interior to a bounded region. It might be objected at this point that some things are literally in the visual field , for instance , while others are no t. That is , the field of vision has certain natural limits , due to human physiology, and the nature of the physical properties of the environment. On this view , vision is naturally bounded , i.e. there is a delimited region of space we can see in any given moment, and hence the use of in to mediate the visual field represents the use of the primary sense for in , without requiring a distinct Percep阳al Accessibility Sense. From this "common-sense" perspective it might appear, on first inspection , that it is erroneous to posit a distinct Perceptual Accessibility Sense associated with in. However, while human sensory perception may be limited, a limit does not necessarily entail a three-dimensional bounded LM. For instance , when we look into the sky , while our visual apparatus allows us to see only so much, there is a lot beyond our vision. While we conceptualise that limit as a boundary , the boundary does not veridically exist. We suggest that the reason we conceptualise the limits of our physical perceptions as a threedimensional container (a bounded LM) is because of experiences of being located inside three-dimensional containers with physical walls (l ike rooms) , which obviously place additional limits on our ability to use our sensory perception organs to their fullest potential. Moreover, this tight experiential correlation is reflected in the linguistic system not only by the particular prepositions which are used to denote the limits of our perceptual access (in and out), but also by the nouns which are used, e.g. , jìeld (as in
Spatial experience, 1臼ical structure and motivation 181 field of vision , visual field , etc.) , which is often conceptualised as a bounded LM , and range (as in range ofvision , the visual range , He's out of range , etc.) , meaning here something like ‘ limits' , which is clearly associated with boundaries and bounded LMs. The In Favour Sense (3.2). The notion of being in has come to be associated with being valued or considered privileged. We hypothesise that the In Favour Sense derives largely 仕om the tight correlation between gaining access or en位Y to certain kinds of bounded LMs and the desirability of the event or activi守 within the confines of the bounded LM. For instance, we form queues to gain access to a whole host of venues on a regular basis. We must wait in a queue to gain access to the cinema to see the latest blockbuster on the first day of release , or for tickets to enter a theme park, a sports stadium, or when waiting for a 企ee table in a packed restaurant. En位Y to a whole host of such venues is often by no means guaranteed, precisely because the activities inside are desirable and thus sought after. Hence , to gain access is to be in a favourable or privileged position. Due to the tight correlation between being within the LM , and hence being in a favourable position，的 has developed an In Favour Sense. 12 To make this point consider the following example: (19)
He managed ω get in the stadium, even though places were limited.
In this example, which relates to the proto-scene, there is an implicature that being in , with respect to the bounded LM , the stadium, is to be in a
12.In addition, items that are valuable are often kept in containers to protect them. Hence, this sense while emphasising the inclusion status of the TR(s) may also draw upon our understanding that containers serve as a means of protecting items from potential extemally-originating harm, as when currency, jewels, and other valuable items are locked in safes , containers par excellence, which are meant to act as impenetrable barriers in order to protect their contents. Our understanding of human intentions leads us to infer that a primary motivation for placing items in a safe container is because those items are positively valued. Accordingly, in this sense, in designates the notion of an entity or an activity being valued or judged as positive in some way. Through pragmatic strengthening, the implicature of value associated with in has given rise to a conventionalised In Favour Sense.
182
月呵lan
Evans and Andrea Tyler
fav 'Ourable p 'Ositi 'O n. Thr'Ough pragmatic strengthening this implicature has c'O me t'O be reanalysed such that in designates the n 'O ti 'On 'Of an entity being valued 'Or judged as p 'Ositive in s'Ome way. C 'O nsider s'Ome examples 'O f this sense: (20) a. He's 切向i伪 the boss). b. She's part 01 the in-crowd. In these examples, in c1 early den'Otes relati'O ns with 'Other pe'Ople, which seem c1'O sely ass'Ociated with being In Fav'Our. H'Owever, this sense 'Of in is n 'Ot limited t'O human relati 'Ons as evidenced by the f'Oll'Owing: (21) a. Turbans are in (this season)! b. That's the in -joke this week. In these examples, in has the interpretati'On 'O f ‘ fav 'Ourable' 'Or ‘ privileged\ Clearly the use 'Of in t 'O designate a fav 'Ourable 'Or p 'Ositive meaning element is distinct 仕om the previ'Ous senses discussed. As evidenced in the examples in (21) , this sense is apparent in c 'O ntexts that n'O l'Onger relate t'O the 'O riginal c'O ntext 'O f use , i.e. with LMs that cann'O t be c 'Onstrued as b'Ounded. This str'Ongly supp'Orts the view that this is a distinct sense instantiated in semantic mem'O ry. The Arrival Sense (3.3). In many spatial scenes in which the experiencer is l'Ocated within a b'Ounded LM , a TR at 'O ne p 'O int l'O cated 'Outside the LM underg'O es l'Oc'O m'O ti 'On such that it c'Omes t 'O be l'Ocated within the LM , as when a train, f'Or instance , pulls int'O a train stati'O n. Fr'Om the perspective 'Of an experiencer l'Ocated interi'Or t'O the LM , the TR is perceived as m 'O ving c1'O ser t'O the interi 'O r vantage p 'O int. This gives rise t'O an implicature 'O f arrival. This n 'Oti 'O n 'Of arrival has bec'Ome instantiated in semantic mem'Ory as a distinct sense, as evidenced by the f'Oll'Owing: (22)
The train is finally in.
In this sentence, n 'O t 'O nly is the TR, the train , within the b 'Ounded LM , the train stati'On , but it has arrived, as attested by the f'O ll'Owing which represents a c1'Ose paraphrase: (23)
The train hasfinally arrived.
Spatial experience, lexical structure and motivation 183
Moreover, this sense is attested in verb-particle constructions such as the following: (24)
He reeled the fish in.
In this example, to reel the fish in reflects drawing the fish towards the angler such that it comes to be proximal with the fisherman. Wh ile the fisherman may then place the fish in a bucket, for instance, which is a bounded LM, the use of in in reel in relates to the notion of coming towards and thus arrival , rather than specifically designating a particular bounded LM. As such, we suggest that examples such as (22) and (24) relate to a distinct meaning component of arrival associated with in. This meaning component is also apparent in sentences such as: She clockedlpunched in at work.
4.4.3. The Vantage Point is Exterior Cluster (4)
An other common position from which bounded LMs are viewed is one in which the experiencer (and hence the vantage point) is located exterior to the LM. Spatial scenes of this kind give rise to a number of distinct senses subsumed under what we will term the Vantage Point is Exterior Cluster. Here we will briefly consider what we term the Disappearance Sense. The Disappearance Sense (4.1). The nature of many physical entities with an interior is that they are made of opaque substances and thus the bounda可 of the LM often obs田cts the observer's view of the interior and hence the contents. Certainly many of an infant' s earliest experiences with bounded LMs 一 LMs which possess interiors - would be of not being able to see the contents contained by such LMs. Food put into a care-giver's mouth cannot be seen, neither can toes and feet put into socks, toys put into cupboards, people moving to a different room, and so forth. Interestingly, nine month-old infants perform better on object-hiding tasks when the occluder consists of an upright and hence prototypical container (Freeman , Lloyd and Sinha 1980; Ll oyd, Sinha and Freeman 1981). Freeman et al. suggest that this superior performance with containers as occluders provides evidence that these infants have established a concept of containers as places where things disappear and reappe缸.
184ηryan
Evans and Andrea Tyler
We suggest that due to the tight correlation between LMs with interiors and occlusion, in has come to be reanalysed as having a distinct Disappearance Sense, as evidenced by the examples given below: (25)
a. The wine quickly soaked in. b. Angel，α rubbed in the lotion. c. The sun has gone in. / The sun is in.
The reasons for thinking that examples such as these evidence a distinct Disappearance Sense are as follows. First, a meaning of disappearance is not evident in any ofthe other senses associated with in. Second, the examples in (25) are context-independent. That is , they cannot be predicted based on any of the other senses for in considered. After all , in (25a) the wine is being absorbed by what appears to be a solid element with no discemable intemal spaces (at least to the naked eye). Similarly, when we say in (25b) that Angela rubbed in the lotion , the lotion is not entering the skin, only to be free to leave again. The skin is not being conceived as an entity with interior space. Rather, in all these examples, the correlation between containment and disappearance , from the perspective of a vantage point exterior to the container, is that in derives a Disappearance Sense which can come to be used in contexts unrelated to the original context which motivated this sense in the first place. Moreover, it is worth pointing out, at this point , that by acknowledging that spatial scenes can be viewed from different vantage points , we have a straightforward explanation for why in has senses which are near opposites , the Perceptual Accessibility Sense and the Disappearance Sense.
4.4.4. The Segmentation Cluster (5)
An important aspect ofbounded LMs is the notion of a boundary, which in part distinguishes interior from exterior. In this cluster of senses , the notion of segmentation or boundedness is privileged. A salient aspect of spatial scenes involving bounded LMs is that they serve to partition the environment , providing a physical means of separation and delimitation. Consider the following sentence: (26)
The farmer put the seed in a sealed box for next year.
Spatial experience, lexical structure and motivation 185 In this sentence, which is consistent with the proto-scene by virtue of being in a sealed box, the seed is protected from external forces and effectively separated from other seed. In a sentence such as (27), which is also consistent with the proto-scene,
(27)
The prisoner was locked in his cell for 23 hours α
day.
the LM acts as a container serving to restrict the prisoner's , the TR's , movement and, hence , separates the TR from other members of society. Thus , an important and frequent inference associated with the proto-scene of in involves the interpretation that bounded LMs effectively partition and segment that which is inside from that which is outside. We will su凹eytwo senses subsumed under this cluster: the Shape As Boundary Sense and the Blockage Sense. The Shape As Boundary Sense (5.1). Langacker (1987) noted that the shape of an object is almost certainly part of our mental representation of that object. Since the shape of a bounded LM is necessarily closely related to its boundaries, it is not surprising that in has come to designate a relation in which the shape of the LM constitutes both the entity and the boundary. In the Shape As Boundary Sense, the TR constitutes part of a delimited configuration forming a shape. Consider the following example:
(28)
Ok, class, put your chairs in α
circle.
This utterance is not typically interpreted as a command to place the chairs inside a circle drawn on the floor. Rather, its interpretation has to do with arranging the chairs such that they form a circle shape, in other words , the arrangement forms a boundary that delimits a circle. Other examples that evidence this sense include the following:
(29)
a. Iffire breaks out get in single file b. Can you get in line.
问(ore
leaving.
We suggest that since a salient aspect ofa bounded LM is its boundary , the use of in to relate a TR and a bounded LM thereby highlights a salient aspect of bounded LMs. Through pragmatic strengthening in has derived a Shape As Boundary Sense.
186ηryan
Evans and Andrea Tyler
The Blockage Sense (5.2). One consequence of being located within a bounded LM is that the boundary can serve to prevent the TR 企om moving beyond the LM. This situation is evident in the following sentences which appe盯 to derive from the proto-scene but which have the implication of blocking the movement of the TR out of the LM. (30)
a. Oxygen must be held in a sealed container (to keep it from escaping 的归 the air). b. In some ancient cultures, /i ve slaves were sealed in tombs with their dead masters.
This notion of blockage or constraint on movement has become conventionally associated with in , as evidenced by the following: (31) a. Wh en 1 got back to my caκ someone had boxedlblocked me in; b. In the northern territories you can get snowed infor months. In both of these examples, rather than being contained by the LM , an entity's movement is obstructed or blocked in some way. This notion of blockage is denoted by 的. In (31 时， the speaker's car is the TR which is blocked from moving by the placement of another vehicle. In (31 b) , the TR is people who are blocked from moving because of the barrier created by the snow. In addition to the 岛regoing， there also appears to be a second 可pe of experiential correlation which may motivate the Blockage Sense. Here the bounded LM 可pically facilitates or is associated with passage. If the TR is sufficiently large it may fill or obstruct the LM hence preventing passage. Consider the following examples which are consistent with the proto-scene. (32)
W旨 couldn 't move the car because a fallen tree was in the driveway. b. Th e portly gentleman got afishbone lodged in his throat.
a.
As noted , in these examples the LM is concep阳alised as a passage, usually a relatively empty space, which is being blocked by the TR. The implicature of blockage apparent here appears to have also contributed to the Blockage Sense associated with in. For instance, this notion of blockage is illustrated in the following examples:
Spatial experience, 1臼ical structure and motivation 187 (33)
a. There 's a bad accident in the roadway with traffic backed up ω the Wilson Bridge. b. The rock is in my way.
In these sentences there is a Blockage meaning associated with in. That is , in does not relate to containment but rather to a constraint on movement. This is strongly suggestive that there is a distinct Blockage Sense conventionally associated with in.
4.4.5.
R听exivity
(6)
The Reflexive Sense (6.1). In her study of verb particle constructions (VPCs) , Li ndner (1981) noticed that some spatial particles have a re f1 exive meaning element associated with them. In is such a spatial p町ticle. After all , in many everyday interactions with bounded LMs , such as crushing a box or a plastic coffee cup , when the sides of the container move inwards they eventually come to occupy what would have originally been the interior space and the position canonically occupied by the TR. Crucially then, in mediates a spatial relation between the same entity at two temporally discontinuous points. Clearly, the boundary of a particular LM cannot simultaneously occupy two different locations at once. Yet, in the Re f1 exive Sense the same entity is conceptualised as constituting the TR and the covert LM (i.e. the LM is contex阳ally understood rather than being linguistically encoded). As with other spatial Re f1 exive Senses (e.g. , see the discussion of the Re f1 exive Sense for over in Tyler and Evans 2001 , e.g. , The fence fell over) , this sense represents a sequence of events rather than a single event. Two consequences of the boundary of the LM moving inward are that i) the LM loses its original shape and ii) the original interior space no longer exists as interior space. Not surprisingly, the Re f1 exive Sense is often associated with collapsing and destruction of the LM and the contents. Consider some illustrative examples of the Re f1 exive Sense associated with in: (34)
a. The walls ofthe sandcastlefell in. b. The house caved in.
「
188
吵ryan
Evans and Andrea Tyler
5. Conclusion In this paper we have argued against the traditional view which treats the lexicon as being the repository of the arbitrary and the idiosyncratic. A consequence of this received position is that lexical items are arbitrarily related, resulting in a homonymy view of word meaning. Our pu甲ose m this paper has been to suggest that the lexicon is systematically motivated. In particular, we have suggested that there are three ways in which the traditional view fails to account for the linguistic behaviour of natural language , and which a motivated theo巧r of word-meaning must account fo r. These constitute 1) the fact that a word can take on new meanings in novel contexts , 2) the fact that words appear to be polysemous , and 3) the fact that a single word can appear in a range of different lexical classes. The linguistic facts force us to conclude that , rather than being arbitrary and lacking in systematicity, lexical structure and organisation is in fact highly organised, achieving a particularly clear indication of a motivated system in the process of meaning extension. We illustrated the central claim that meaning extension is a highly motivated process , grounded in spatiophysical experience, with an analysis of the English lexeme in. This study therefore provides insight into i) the non-arbitrary quality of the mental lexicon , ii) the highly creative nature of the human concep阳al system , and ii i) the fact that the way we experience renders spatio-physical interactions meaningful , which in turn gives rise to emergent conceptual structure.
References Beitel , Dinara , Raymond Gibbs and Paul Sanders 1997 The embodied approach to the polysemy of the spatial preposition on. In: H. Cuyckens and B. Zwanda (eds.) , Polysemy in Cognitive Linguistics: Selected Papers from the Fifth International Cognitive Linguistics Co叫声rence， 241-260 , Amsterdam: John Benjamins. Bloomfield, Leonard 1933 Language. New York: Holt, Reinhart and Winston. Brugman, Claudia 1981 The story of over. MA Thesis. Linguistics dept., UC Berkeley. Published [1 988] as The Story of over: Polysemy, Semantics and the Structure ofthe Lexicon. New York: Garland Press.
Spatial experience,
1臼ical structure
and motivation 189
Brugman, Claudia and George Lakoff 1988 Cognitive topology and lexical networks. In: S. Small, G. Cottrell and M. Tannenhaus (eds.) , Lexica/ Ambiguity Resolution, 477-507. San Mateo , CA.: Morgan Kau如lan. Bybee, Joan , Revere Perkins and William Pagliuca 1994 The Evo/ution of Grammar: 1旨nse， A再pect and Modality in the Languages ofthe World. Chicago: Chicago University Press. Chomsky, Noam 1995 The Minimalist Program. Cambridge, Mass.: MIT Press. Clark, Eve 1973 Nonlinguistic strategies in the acquisition of word meanings. Cognition 2: 161一 182. Croft, William and D. Alan Cruse 2004 Cognitive Linguistics. Cambridge: CUP. Cuyckens, Hubert, Dominiek Sandra and Sally Rice 1997 Towards an empirical lexical semantics. In: Birgit Smieja and Meike Tasch (eds.), Human Contact through Language and Linguistics， 35一54. Berlin: Frankfurt am Main. Dirven, René 1993 Dividing up physical and mental space into conceptual categories by means of English prepositions. In: C. Zelinsky-Wibbelt (ed.) , Th e Semantics of Prepositions:From Menta/ Processing ω Natu ra/ Language, 73-97. Berlin: Mouton de Gruyter. Evans , Vyvyan 2004 The Structure of Time: Language, Meaning and Tempora/ Cognition. Amsterdam!Philadelphia: John Benjamins. Evans, Vyvyan and Melanie Green To appear Cognitive Linguistics: An Introduction. Edinburgh: Edinburgh University Press. Evans, Vyvyan and An drea Tyler 2004 Rethinking English "prepositions of movement": The case of to and through. Be/gian Journa/ ofLinguistics 16. Fauconnnier, Gilles and Mark Tumer 2002 Th e Way We Th ink: Conceptual Blending and the Mind's Hidden Compl臼ities. New York: Basic Books. Fleischman , Suzanne 1999 Discourse markers across languages: Implications of a case study for historico-comparative and sociolinguistics. Paper presented at Georgetown University Linguistics dept., Nov. 1999. Freeman, N., S. Lloyd and C. Sinha 1980 Infant search tasks reveal early concepts of containment and canonicall刷ge of objects. Cognition 8: 243-262.
190
吵ryan
Evans and Andrea Tyler
Gibbs , Ra严nond and Teenie Matlock 1997 Psycholinguistic perspectives on polysemy. In: H. Cuyckens and B. Zawada (eds.) , Pol)忍emy in Cognitive Linguistics: Selected papers from the Fifth International Cognitive Linguistics Conference , 213-239. Amsterdam: John Benjamins. Grady, Jos叩h 1997 Foundations of Meaning: Primary Metaphors and Primary Scenes. Dept. of Linguistics , U.C. Berkeley: Ph.D Dissertation. Gumperz, John 1982 Discourse Strategies. Cambridge: Cambridge University Press. Hawkins , Bruce 1988 The category MEDIUM. In: Rudzka-Ostyn, B. (ed.) , Topics 切 Cognitive Grammar, 231-270. Amsterdam: John Beniamins. Heine, Bemd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Herskovits , Annette 1986 Language and 年atial Cognition: An Interdisciplinary Study of the Prepositions in English. Cambridge: Cambridge University Press. 1988 Spatial expressions and the plasticity of meaning. In: B. RudzkaOstyn (ed.) , Topics in Cognitive Grammar, 271-298. Amsterdam: John Benjamins. Hill, Clifford Alden 1978 Linguistic representation of spatial and temporal orientation. Proceedings of the Fourth Annual Meeting of the Berkeley Linguistics Society, 524-538. Berkeley: UC Berkeley Press. Hopper, Paul and Elizabeth Closs Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Hottenroth, Priska-Monika 1993 Prepositions and object concepts: A contribution to cognitive semantics. In: C. Zelinsky-Wibbelt (ed.) , The Semantics of Prepositions , 179-220. Berlin: Mouton de Gruyter. Johnson, Mark 1羽e Body in the Mind. Chicago: Chicago University Press. 1987 Lakoff, George 1987 Women , Fire and Dangerous Things: Wh at Categories Reveal about the Mind. Chicago: Chicago University Press. Lakoff, George and Mark Johnson 1980 Metaphors We Live by. Chicago: Chicago University Press. 1999 Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western Thought. New York: Basic Books.
年atial 臼:periencι lexical structure
and motivation 191
Langacker, Ronald 1987 Foundations 01 Cognitive Grammar. Volume 1. Stanford, CA.: Stanford University Press. 1991 Foundations 01 Cognitive Grammar. Volume 2. Stanford, CA.: Stanford University Press. 1999 Grammar and Conceptualization. Berlin: Mouton de Gruyter. Lindner, Susan 1981 A Lexico-semantic Analysis 01 English Verb Particle Constructions with out and up. Doctoral thesis. Linguistics Dept., UC San Diego. Distributed [1983] by the Indiana Linguistics Club. Lindstromberg, Seth 1998 English Prepositions Explained. Amsterdam: John Benjamins. Lloyd，乱， C. Sinha, C. and N. Freeman 1981 Spatial references systems and the canonicality effect in infant search. Journal 01 E苟perimental Child Psychology 32: 1-10. Miller, George and Philip Johnson-Laird 1976 Language and Perception. Cambridge: Cambridge University Press. Pustejovsky, James 1995 The Generative Lexicon. Cambridge , MA.: MIT Press. Quirk, Randolph, Sidney Greenbaum, Geof企ey Leech and Jan Svartvik 1985 A Comprehensive Grammar 01 the English Language. Harlow: Longman. Ri ce , Sally, Dominiek Sandra and Mia Vanrespaille 1999 Prepositional semantics and the 企agile link between space and time. In: M. Hiraga, C. Sinha and S. Wilcox (eds.) , Cultural, Psychological and Typological Issues 切 Cognitive Linguistics: Selected Papers 01 the Bi-annual ICLA Meeting in Albuquerque, July1995 , 108-127. Philadelphia: John Benjamins. Sandra, Dominiek 1998 What linguists can and can't tell us about the mind: A reply to Croft. Cognitive Linguistics 9: 361-378. Sandra, Dominiek and Sally Ri ce 1995 Network analyses of prepositional meaning: Mirroring whose mind - the linguist's or the language user's? Cognitive Linguistics 6: 89-130. Schiffrin, Deborah An aphoric then: Aspectual , textual and epistemic meaning. Lin1992 guistics 30: 753-92. Sinha, Chris and Tania Kuteva 1995 Distributed spatial semantics. Nordic Journal 01 Lin伊istics 18: 167-199.
192
月Iryan
Evans and Andrea Tyler
Svorou, Soteria 1994 1页eGrammarol年ace. Amsterdam: John Beniamins. Sweetser, Eve 1990 From Etymology ω Pragmatics: Metaphorical and Cultural Aspects 01 Semantic Structure. Cambridge: Cambridge University Press. 1999 Compositionality and blending: Semantic composition in a cognitively realistic 企amework. In: T. Janssen and G. Redeker (eds.) , Cognitive Linguistics: Foundations , Scope, and Methodology, 129一 162. Berlin: Mouton de Gruyter. Talmy, Leonard 1988 The relation of grammar to cognition. In: B. Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics, 165-205. Amsterdam: John Beniamins. 2000 Toward a Cognitive Semantics. Vol. 1 and 2. Cambridge, Mass.: MITPress. Taylor, John 2003 Linguistic Categorization. Third edition. Oxford: Oxford University Press. Traugo忧， Eliz冠beth Closs 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31-55. Tyler, An drea and Vyvyan Evans 2001 Reconsidering prepositional polysemy networks: The case of over. Language 77: 724-765. 2003 The Semantics 01 English Prepositions: 牛atial Scenes, Embodied Meaning and Cognition. Cambridge: CUP. Vandeloise, Claude 1990 Representation, prototypes and centrality: Meanings and prototypes. In: S. Tsohatzidis (ed.), Studies in Linguistic Categorization , 403-437. London: Routledge. 1991 Spatial Pr，写positions: A Case Study in French. Chicago: The University of Chicago Press. 1994 Methodology and analyses of the preposition in. Cognitive L切· guistics 5: 157-184.
Motivating the uses of basic verbs: Linguistic and extralinguistic considerations JohnNewman
Abstract Various grammatical facts associated with ordinary verbs are reviewed and shown to be at least partly motivated by some of the common experiential realities associated with the behaviours related to these verbs. The proliferation of idiomatic usages of an ‘ eat' verb in Hausa and the use of ‘ give' as a light verb in English (g ive the ball a kick, give the car a wash , etc.) reflect, arguably , a basicness about these concepts which make them especially available as sources for idiomatic and grammaticalized usages. Some experiences are naturally grouped together (e.g. , coming and going are converse experiences; sitting, standing, and lying , albeit with some cultural variations , are natural at-rest positions), and various grammatical facts reveal ways in which verbs encoding such concepts function as unique groups in languages. Finally, a gradation in degree of sensorimotor control, which is evident in the case of sitting, standing, and lying positions, can be seen as motivating the differential morphosyntax occurring with verbs encoding these concepts in some languages
Keywords: basic verbs , complementarity, experiential reality, extralinguistic motivation , grammaticalization, linguistic motivation, posture verbs , serial verb construction
1. Introduction The research which 1 report on here has developed out of ideas current in the cognitive linguistics movement (Lakoff 1987; Rudzka-Ostyn 1993) and
194 John Newman
Cognitive Grammar (Langacker 1987/1991 , 1994). In these approaches , and in the approach adopted here , language is seen as an integral part of human cognition, situated in the larger contexts in which human cognition plays some part. Cognition is taken to refer to the full range of 如nctions of the human mind , embracing the processing of sense-data, the conceptualization of events , the acquisition of beliefs and knowledge , social interaction , etc. Language takes its place alongside the myriad of other cognitive structures which underlie all human states and activities and is to be understood as situated in the realm of cognition, rather than as something entirely distinct from cognition (Langacker 1994: 27-28). Physical , social , and cultural contexts, and how speakers relate to them, are considered significant factors in the formation and adaptation of linguistic structures. The acknowledgement of important links between linguistic structure and human cognition (and, more generally, human experience) is a hallmark ofthe cognitive linguistic movement , acknowledged already in the introduction by Rudzka-Ostyn (1 993: 1-2) to a volume of papers based on the First International Cognitive Li nguistics Conference in 1989 (Geiger and Rudzka-Ostyn 1993). With a "situated" view oflanguage like this , it is only natural to incorporate some non-linguistic concepts into a discussion of language and its organization. Cognitive Grammar has made extensive use of some key principles drawn from our understanding ofhuman cognition which have applicability in perception and conceptualization. A striking example of this is the reliance upon the notions of figure and ground in the description of language structure. So , for example, the subjectiobject structuring of c1 ause content may be construed as a type of a pervasive figure/ground organization, thereby locating this aspect of syntactic structuring in the larger frame of human cognition. So , too , the meaning of a morpheme is understood as the selected, profiled sub-part (figure) of a larger frame of reference , or base (ground). Cognitive Grammar research has contributed greatly to our understanding of how such general cognitive principles can be quite naturally extended to accounts of language structure. At the same time , one needs to recognize that grammar, though grounded in human cognition, has its own distinguishing properti
Motivating the uses 01 basic verbs
1'J)
tive domains but is nevertheless distinctive in its structure and its particular type of complexity. Language exists as a sub-system within the larger cognitive context, and one can inquire into the connections between this subsystem and other aspects of cognition. In what fo l1ows 1 explore a number of cases where aspects of ordinary human experience which exist independently of language play an important part in shaping some grammatical structure. My intention is not to reduce the grammatical structure in question to something purely ex位alinguistic， but rather to argue that the combination of linguistic and extralinguistic factors motivates a particular structure. The extralinguistic influences 1 discuss relate to "experiential reali旷" i.e. , our ordinary experience of the world around us , and in particular some common and basic realities which humans in al1 societies experience on a day-to-day basis. Admittedly, an approach relying on such loose ideas as "ordinary experience" is open to a charge of unscientificness and sheer invention if it is not treated with some degree of discipline. How can there be, for example, any objective criteria to decide what counts as "ordinary" in human experience? While 1 am conscious of the 仕agility of an appeal to experiential reality, it would be quite wrong (from the cognitive-linguistic viewpoint) to completely exclude any reference to ordinary experience in accounting for linguistic phenomena. What is cal1ed for, rather, is a responsible approach al10wing reference to the most familiar aspects of ordinary experience, with the understanding that such references will always remain tentative to some extent. Despite qualms one might have about this methodology , the appeal to an experiential reality in explicating aspects of language structure and use is now familiar in cognitive linguistic research , especial1y research which tries to motivate metaphorical uses of language. In works such as Lakoff (1987: 260-268) , Lakoff and Johnson (1 980: 5ι60; 1999: 74-9月， Lakoff and Tumer (1989: 112-114), Johnson (1987) , and Sweetser (1 990: 28-32), the experientialist approach is viewed, in fact , as profoundly important with implications which reach far beyond discussion of metaphor or polysemy. In these works it is argued that our most basic modes of reasoning and understanding (not just some linguistic phenomena) are shaped by the earliest "primary" experie
196 John Newman
states/acts , natural links between certain states/acts , and our ordinary experience ofbody positions. 1 refer to these aspects as extralinguistic. The grammatical facts to be discussed are drawn from research which 1 have been carrying out as part of a larger, cross-linguistic study of selected "basic" verbs: sit, stand, lie; come, go; see, hear; eat , drink and give , take. This study is concemed with the lexical , morphological , syntactic and semantic properties of these verbs , along the lines of the analysis of ‘ give' verbs presented in Newman (1996). The focus on bodily states and acts in this research re f1 ects the experiential bias to language study described above , an approach aptly summed up in the phrase "the embodied mind" (as opposed to "the disembodied mind" , Lakoff and Johnson 1999: 74117). As with other bodily concepts (e.g. , body parts) these particular verbal concepts tend to play a significant role in language. 1 understand motivation in the present context as referring to one or more linguistic or extralinguistic properties or principles which are reasonably thought of as in f1uencing some outcome in the structure of language. Certainly there is no expectation when speaking of motivation here that the motivating factors exhaustively determine a particular resul t. In the examples considered below, both linguistic and extralinguistic factors are viewed as partial motivations in accounting for a given structure. In particular, 1 do not attempt to reduce al1 the linguistic facts to extralinguistic causes or even general cognitive principles. Rather 1 accept, for the present puφoses ， some observable general linguistic patteming in a language as given a priori, regarding it as some linguistic motivation for the phenomenon to be explained. Extralinguistic considerations are brought to bear in order to further motivate the particular structure in question.
2. Experiential basicness of states/acts The basicness of some experiences is relevant to language structure in quite a number of ways. One obvious way , al1uded to above , is the fact that our common experiences of such things as eating, drinking , seeing, hearing , etc. can be the basis for conceptualizing other more complex events. The linguistic manifestation of this fact is that those domains relating to such basic acts tend to be the sources for talking about other, possibly more complex, domains. 1 wi11 leave open the question of just how the connection between the basic experience and its linguistic consequences comes about. An idea which has found favor in the cognitive linguistics movement
Motivating the uses ofbasic verbs 197
is that such connections are mediated by 由e formation of basic image schemata (motion along a path , balance, container, etc. , cf. Tumer 1996: 16一25 岛r a recent overview of 由is research). That is , our earliest basic experip' 川 give rise to these body-based skeletal concepts and it is 由rou ler manipulation of these conc叩ts 由at we arrive at the full ..taphorical uses of language. It is not necess缸y here 协 accept range . all that has been claimed of image schemata and their role in linguistics. What is relevant here is to note that the verbs which encode meanings relating to the basic experiences tend to figure in many figurative expressions in a language. In cognitive linguistics, figurative language and metaphor fall comfortably within the realm of study and are not considered as "abnormal" uses of language. Consequent1y, a cognitive linguistic study of basic verbs will most likely include a substantial discussion of such extensions. All the verbs which 1 have been studying are productive in terms of their participation in figurative usage and many s阳dies have already been carried out documenting this aspect. Basic verbs which have been studied from a cognitive linguistic viewpoint include: sense-perception verbs (Sweetser 1990: 32-48); ‘ come' and ‘ go' (Radden 1996; Shen 1996); ‘ stand' and ‘ lie' (Serra Bometo 1996),‘ see' (Alm-A凹ius 1993); ‘ eat' and ‘ drink' (Newman 1997); ‘ give' and ‘ take' (Newman 1996, 1998), 'take' (Norvig and Lakoff 1987),‘have' (Wierzbicka 1988), and miscellaneous verbs referring to bodily acts (Pauwels and Simon-Vandenbergen 1995). In all these works , there is a p缸ticular interest in the metaphorical uses to which these verbs are put, consistent with the cognitive linguistic orientation underlying these discussions. 1 will not attempt here to summarize all the figurative extensions of the basic verbs and their grounding in ordinary experience. 1 refer 也e reader to the works cited above for fuller discussion of such uses. Sometimes, a basic verb can be used in such a productive and extensive way that it seems worthy of recognition even when one is attending mainly to grammatical structure. That is , the verb may be so prominent in terms of its productivity 由创 cognizance needs to be taken of it in discussing clause structure. A s创king example of such a verb is Hausa (West Chadic) ci ‘e剑' (based on Abraham 1962), illus往ated below with some of its uses: GATHE阳NG 剧， BRINGING TO ONESELF a. ci ‘ eat' + baabaa ‘ indigo dye' ‘ take the indigo dye well' b. ci ‘ eat' + audùgaa ‘ cotton' ‘ gather cotton' c. ci ‘ eat' + da l)Kò ‘ rubber' =‘ ga由errubber'
(1)
198 John Newman d. ci ‘ eat' +卢re ‘ tobacco blossom' (also hiiraa ωaci卢re e. ci ‘ eat' +加 díi 'money' f. ci ‘ eat'
=‘ gather tobacco' =‘ conversation became animated') =‘ embezzle money' =‘ accommodate (guests etc.)'
(2)EXPERIENCING SOMETHING PLEASANT BY SUBJECT 阻FERENT a. ci ‘ eat' + daadíi 'pleasantness' ‘ feel happy' b. ci ‘ eat' + dàacii ‘ bitterness' ‘ laugh jeeringly' c. ci ‘ eat' + gaadò ‘ property inherited' ‘ inherit property' d. ci ‘ eat' + karì ‘ profit ' ‘benefit, gain money' e. ci ‘ eat' + mòoriiyaa ‘ usefulness' =‘ find something useful' f. ci ‘ eat' + riba ‘ profit' ‘ get profit, get advantage' g. ci ‘ eat' + kàràatuu ‘ studying' as in yaa ci kàràatuu tasà =‘ he devoted all his attention to
h. ci ‘ eat' + duuniiyà ‘ world'
z
study' ‘ enjoy life'
(3) EXPERIENCING SOMETHING UNPLEASANT BY SUBJECT REFERENT a. ci ‘ eat' + baashì ‘ debt' ‘ incur a debt' b. ci ‘ eat' + zamaa ‘ state of being, seated, remaining , swelling' =‘ wait a long time' c. ci ‘ eat' + dù招umii ‘ removing completely' ‘ be left without a leg to stand on' d. ci ‘ eat' + wàhalà ‘ trouble' ‘ suffer 位'Ouble' e. ci ‘ eat' + wùyaa ‘ difficulty' ‘ experience difficulty' f. ci ‘ eat' + Kasa ‘ earth, soil' ‘ do humble obeisance' g. ci ‘ eat' + kunya ‘ feeling ashamed' ‘ feel ashamed' h. ci ‘ eat' + gam ‘ malodorous, green locust said to be poisonous to eat' =‘ bring trouble on oneself i. ci ‘ eat' + yaajì ‘ pungency' ‘ have a bad time' j. ci ‘ eat' + verbal noun =‘ undergo severely' , as in yaa ci duukà ‘ he was severely beaten'
(4) DESTROYING, INJURlNG, or DEFEATING OBJECT REFERENT a. ci ‘ eat' + sù ‘ them' ‘ conquer them' ‘ win the war' b. ci ‘ eat' + yaafà ‘ war' c. ci ‘ eat' +卢ska ‘鱼饵‘public1y humiliate' d. ci ‘ eat' + gìndi-nsà ‘bo忧om， base ‘ catch him by the ofthing-his' waist; backbite someone'
Motivating the uses ofbasic verbs 199
e. ci ‘ eat' + hanci ‘ nose' in sunàa cîg hanci f. ci ‘ eat' + irlì-nsà ‘ honour-his' g. ci ‘ eat' +卢ska tasù ‘ face their' h. ci ‘ eat' + kàasuuwa tasà ‘ market his' i. ci ‘ eat' + mutuncì-nsù ‘ treating others with due respect -their' j. ci ‘ eat' + ràboo ‘ luck'
=‘ they are quarreling ,
abusing one another' =‘ put him to shame' =‘ publicly humiliate them' =‘ malign him' =‘ treat them disrespectfully =‘ ruin luck' ，部 in yaa ci ràbansà ‘ his luck was short-lived'
(5) MISCELLANEOUS a. ci ‘ eat' + ado ‘ adomment' b. ci ‘ eat' + dàmarà ‘ belt' c. ci ‘ eat' + sàrautà ‘ having sovereinty' (also ci 'eat' without object d. ci ‘ eat' + laya ‘ written charm, Koran' e. ci ‘ eat' + dà 句Jau ‘ good， well' E 元àasuuwa ‘ market' + ci ‘ eat' in kàasuuwa tanàa cî g. ci ‘ eat' + kàasuuwa ‘ market'
=‘ to dress up' =‘ get ready for boxing' =‘ become ruler' =‘ be the ruler') =‘ swear on the Koran' =‘ work well , be efficacious' =‘ market's going on' =‘ buy goods at market; sell
goods at good profit' (also sunàa cî1J 始asuuwa
‘ market'
=‘ there's a hubbub')
Some, but not all , of these uses can be motivated by appealing to metaphorical mapping based on the source domain of eating and evident in many languages. Agent-oriented extensions, in which the extensions are based on the experience of the consumer, are common. They relate to images of intemalization，企om the relatively concrete (e.g. , to drink the air) to the more abstract (e.g. , to eat up adoration). This kind of ‘ taking in' sense is what we see in examples (1) , (2) and (3). The patient-oriented extensions of ‘ eat' and ‘ drink' in (4) , in which the effect of consumption on the food or 世ink is the basis of the extension, are also common in English, as in the rust ate inω the metal, the pain ate into her, etc. The extensions in (匀， however, are not quite so transparent in the way they relate to literal ‘ eat' and do not have parallels in many other languages. An ‘ eat' verb occurring in a wide range of idiomatic expressions can be found in some other African languages, such as Akan (Niger-Congo, cf. Sætherø 1995: 139-141). The prominence and importance of ‘ eat' in the
200 John Newman clause structure of these languages is acknowledged by W e1 mers , as part of the conc1uding paragraph to a very substantial work on African languages: 1 have attempted to sum up its [the Akan ‘ eat' verb's] uses under the general word for ‘四t' ， but with other objects it refers among other things to using up or wasting money, taking a day off, having sexual relations with someone, accepting a bribe , inheriting goods , winning a victory, defeating an opponent, playing a game , holding an office , enduring suffering , making a bargain , living in some specified way, and so on at considerable length. A language has not been well studied until the nature 01 such semantic ranges , if not an exhaustive list of all recordable collocations , has been noted. (Welmers 1973: 477 , my italics)
An other way in which ordinary experience bears on language structure concerns the use of lexical items referring to common, ordinary experiences as the basis for grammaticalizations. Many of the verbs 1 have been studying are well known as sources of grammatical elements and a considerable number of typical evolutionary paths from lexical verb to grammatical mo甲heme have now been either documented , or at least hypothesized (c f. Heine et al. 1991; Heine et a l. 1993; Svorou 1993; Bybee et al. 1994). This cross-linguistic research has revealed the diversity of sources for grammaticalizations, though lexical items referring to the basic entities of ordinary experience are the main lexical sources. Despite this overall trend, it is c1 ear that a simple, naïve notion of basicness is insufficient as a way of identifying exactly a mo甲heme' s potential for grammaticalization (cf. Heine et al. 1991: 32-36). So , for example , verbs of giving and taking seem to give rise to more instances of grammaticalization than do , say , eating and drinking. A category akin to grammaticalization, but one less studied in the cognitive-linguistic framework , is the use of basic verbs as semantically empty verbs in some languages , figuring in many combinations such as verb + object, verb + predicative adjective , etc. Although the verbs may still retain all or most of their characteristics as verbs , their usages may be so extensive and their meanings so schematic that they warrant a categorization of their own. Tenns such as dummy verb or light verb have been used to refer to this phenomenon. An example 仕om English is the use of give in expressions such as give
Motivating the us臼 01 basic verbs 201 cigarette a smoke, *give the beer a drink). The semantic effect of give in this construction is to provide a schematic sense of transfer from agent to patient, with the details of the transfer elaborated by the deverbal. The verbs which are liable to function as light verbs are basic verbs such as give, make, put, take, etc. There is some motivation for this construction 仕om within the English grammatical system, namely, the existence of a ditransitive construction (give NP 1 NP2) in which give is used in its central sense ‘ pass something by hand to a recipient'. This constitutes the linguistic motivation behind the light verb use. The extralinguistic motivation for employing give as light verb comes from the salience of the giving event in our ordinary experience. This salience in our lives makes give a prime candidate for developing a more schematic and grammatical function. Another way in which the basicness of verbs bears on grammar concems their zero realization in some languages, similar to the way in which some languages may lack a category of copula. An instance of this is the Amele (Papuan, Roberts 1987, 1998) ‘ give' construction. In this construction , affixes which would normally atlach to verb stems such as tense , aspect, mood, verb agreement, etc. , appear merely as a string of affixes without any verb root in the word. Examples of Amele ‘ give' are shown in (6). (6)
a. Naus Dege ho ut-en. Naus Dege pig 3SG.IO-3SG.SUBJ.PAST ‘ Naus gave Dege the pig.' (Roberts 1987: 34) b. /j a dana leis sab al-ig-a. 1 man two food 3DU.IO-1SG.SUBJ-PAST ‘ 1 gave the two men food.' (Robert忘 1987:316)
Compare these sentences with a normal three-place predicate like ihac ‘ show' in (7). PRED in (7) refers to a predicate marker i , which occurs between a verb stem and a following oblique-marking or indirect objectmarking affix. In (7) it appears after the verb stem ‘ show' and before the indirect object.
(7)
Jo eu ihac-i-ad-ig-en. house that show-PRED-2PL .IO-1SG.SUBJ-FUT ‘ 1 will show that house to you (plural).' (Roberts 1987: 69)
202 John Newman
Since the indirect object affixes appear in a position after the verb stem. one could analyze the verbs in (6) as beginning with a ‘ give' stem, realized as a zero morph , followed by the indirect object affixes. This is , in fact , the analysis implicit in Roberts (1987) , where ‘ give' is translated as the zero morph in the dictionary section of the grammar. One difference between the shape of ‘ give' words and other verbs in Amele concems the relative ordering of the direct object and indirect object prefixes when both are present. With verbs like he/ec ‘ to throw' , the order is DO …10, whereas in the case of ‘ give' the order is 10 - DO. This is illustrated in (8). (8)
a. he/-ad-ih-ec throw-3PL. DO-2SG.IO-infinitive ‘ to throw them to you.SG' b. ut-ad-ec 3SG .I O-3PL.DO-infinitive ‘ to give them to him!h町'
This difference in the order of the 10 and DO morphemes shows that it is not quite accurate to describe the ‘ give' forms simply as ‘ stem-less' variants of three-place predicate verbal forms (cf. Roberts 1998). Nevertheless , it remains true that in many cases, such as (6) above , the words translatable as ‘ give' appear on the surface to lack a ‘ give' verb stem and this is the implicit analysis of such forms in Roberts (1987). Some motivation for this way of encoding a giving event can be found within Amele. Firstly, note that in Amele verb stems can be minimal in terms of phonemic form and may consist of just one phoneme. Examples of such forms are q-oc ‘ to hi t', m-ec ‘ to put', o-c ‘ to get', /-ec ‘ to go' , b-ec ‘ to come up' , and n-ec ‘ to come down' , where -ec and -oc are infinitive markers. As Roberts (1998: 29) suggests, one could think of a zero ‘ give' verb as an even more reduced form than these other common verbs. Secondly, one should take into account the fact that there are affixal morphemes present in an Amele ‘ give' clause which indicate something about the type of event being referred to. The indirect object category, cross-referenced on the verb , refers mainly to a (human) recipient or possessor, for example, though it also covers the meanings of addressee , allative , ablative , locative, and partitive (Roberts 1998: 17). In (6) , the verbal affixes indicate a 3SG subject and a 3SG indirect object, and so there is already the sense of someone doing something to , or acting in some way towards or for , some other
Motivating the uses 01 basic verbs 203
person or thing. In other words , the richness of Amele verbal morphology provides a kind of skeletal clause structure in which the lexical expression of transferrlng something to someone through the use of verbal exponent could be dispensed with. Thirdly, there is ano由er construction type where one finds a string of verbal affixes without a verb stem, namely certain impersonal constructions describing human states. In 由is construction, the verb always shows 3SG concord, consistent with the noun in this construction being the subject. An example is given in (9). (9)
Ija
wen t-ei-a. lSG hunger ISG.DO-3SG.SU-TODP '1 was hungry.' (Roberts 1998: 14)
Through different choices ofthe noun (24listed in Roberts 1987: 316), the construction conveys meanings such as ‘ be tired' ,‘be prickled' ,‘ be afraid' , ‘ be confused' ,‘ feel pain' ,‘ feel cold' ,‘understand' ,‘ have desire' ,‘ feel hot' ,‘be itchy' , etc. A similar construction is used to express the optative (Roberts 1987: 317). While there is a basis in the existing structures of Amele 岛r the nonrealization of a ‘ give' verb stem, the question still remains as to why it should be the ‘ give' sense which attaches to the structure in (6) and not some other meaning. Granted 也就由.e presence of subject and indirect object affixes suggests a meaning consistent with the range of meanings of these affixes, this still allows 岛r quite a varlety of possible meanings. Why should it be ‘ give' , rather than, say，‘ sell' ，‘白row' ，‘push' ，‘prod' ，‘ lend' ， etc? It is here 由at 也e basicness of giving becomes relevant. The basicness of giving helps to explain why it is the sense of ‘ give' which is associated with such a minimal clause structure.
3. Natural groupings of states/acts In my study of basic verbs, 1 have chosen to consider verbs in groups such as eat and drink, come and go , give and take, etc. One reason for doing this is that such pairs reflect states/acts which occur in ways we experience as complementary. 80，岛r example , eating and drinking represent the two natural ways in which sustenance is taken into the body. In addition they complement each other by being done together, sequentially, in a natural way as part of 也e larger event, e.g. , a meal. The entrenched nature of collo-
204 John Newman
cations in English such as eat and drink, eating and drinki.晤， Jood and beverage, etc. is a reflection of the natural complementarity of the acts as part of ordinary human experience. 'Come' and ‘ go' are similarly complement町y. They represent opposites in terms of the direction of motion with respect to the speaker in their prototypical uses. At the same time the sequence of coming and going, i.e. , returning , is also a natural kind of human behavior. Their oppositeness is quite transp町ent in Yimas (Papuan). Yimas has a verb pura- ‘ come' , derived 企om pu- , related to 由e forms for ‘ go' in some neighboring languages , and the suffix -ra, indicating reversed direction (Foley 1986: 212). Pura- ‘ come' might be analyzed therefore as meaning literally ‘ go in the reverse direction'. Coming and going by humans is usually purposeful and so both ‘ come' and ‘ go' verbs collocate easily with other verb phrases indicating the subsequent or consequent action. Thus , we find these verbs used especially in serial verb constructions. Indeed , precisely these two verbs can be especially privileged when it comes to serial verb constructions. Foley and Olson (1 985: 41), in a discussion of serialization in various languages , identi马， the pair ‘ come' and ‘ go' as the most 企equent verbs entering into serial constructions. They draw attention to Kaititj (Arandic) , for example, which has only these verbs functioning as serial verbs. In Rama (Chibchan , Craig 1991: 467) , siik ‘ come' and taak- ‘ go' are singled out as the most common verbs involved in a lexica1ized pa忧em of combination with instrumental and comitative mo叩hemes， giving rise to meanings ‘ come with' =‘ bring'. Another language in which ‘ come' and ‘ go' play an important and complementary role in the serial verb construction is Kalam (Papuan). Pawley and Lane (1998: 218-222) single out the verbs ‘ come' and ‘ go' for their special role in what they call multi-scene serial verb constructions in this language. These are constructions referring to a series of acts which take place at different locations. They are characterized by a sequence of bare verb stems preceding a verb inflected for tense, aspect, mood, and subject reference. The general formula for representing the minimal way of reporting single deliberate actions is given below: Table 1. Kalam serial verb construction
12 Move / Stay Pivotal Act
3 Move (with Affected Object)
4 Outcome
Motivating the uses of basic verbs 205
An example ofthis construction is given below. Here, the morphemes ‘ go' , ‘ hit' ,‘ smash' ,‘ get', and ‘ come' are the bare verbs appearing before the final inflected verb ‘ put'. (10)
B
ak am mon p-uk d ap la-k Man that go wood hit-smash get come put-3SG-PST 'The man fetched some firewood.' (Kalam, Papuan, Pawley and Lane 1998: 204)
The special properties of ap ‘ come' and am ‘ go' in these multi-scene constructions are: (1) they are said to be the most common verbs occurring in slots 1 and 3 in Table 1; (2) these two verbs (and their compound forms) can combine with a verb ‘ get' or the adjunct ‘ carrying' to refer to the carrying of objects in slot 3 in Table 1; (3) ap and am can be followed by a wider range of verb stems than is the case with other verbs. Thus,‘ come' and ‘ go' 町e especially favored in the serial verb constructions as predicates which help "lead" to the next stage in the chain of actions which make up the whole. It is significant that it is the pair of verbs ‘ come' and ‘ go' which is special to this construction. Just as coming and going constitute a commonly occurring c1uster of converse experiences, so too the verbs referring to those experiences represent the two key verbs which figure in these serial verb cons位uctions. The human ‘ at rest' positions such as sitting, squatting, standing, leaning , lying face down , lying face up , etc. represent another kind of relatively natural grouping of experiences. Humans are all fam i1iar with some of these positions, though cultural practices vary considerably with respect to the positions people prefer to assume (e.g. , sitting upright versus squatting, sitting cross-legged versus sitting with feet dangling in 仕ont， etc.). Sitting, standing, and lying would appear to be the three main positions which humans assume , though admittedly with some cultural variations. We 可pl cally experience all three in any 24 hour cyc1 e, indeed it is uncomfortable and even painful for us to maintain any one of these positions for long periods. We aItemate between the three (and motion , of course) as part of our ordinary experience of living. They represent the three main choices we face in being at rest and so could be called ‘ complementary' opposites. Being so basic to our ordinary experience, they provide natural sources for a variety of images or image schema (cf. Gibbs et al. 1994), in particular images relating to rest, continuation, spatial extension and the shape of things. Of interest here is the way in which it is just the triplet of concepts
206 John Newman ‘ si t' ,‘ stand' , and ‘ lie' which can function as the basis for a three-way dis-
tinction which needs to be recognized in some part of clause structure. A common instance of this involves the use of these three predicates to describe the location or state of some thing, together with some categorization of its shape or appearance. In Mbay (N ilo-Saharan , Keegan 1997: 775 句， for example, existential constructions and predicative adjective constructions require one of the three verbs ndì ‘ sit', dà ‘ stand' and tò ‘ lie\A locative expression equivalent to 吁lere is/are, there is/are' also builds upon these three verbs by requiring one of the specifiers: ndìn for an object viewed as sitti吨， dàn for an object viewed as standing, and tèn for an object viewed as Iying. In the locative expression they are used in conjunction with a locative morpheme lã , as exemplified in (1 1).
(1 1) a. Mbëtë lò-í lã tèn. LOC Iying book poss-you ‘ Here is your book.' (Mbay , Keegan 1997: 76) b.
wé; lò-í lã ndìn. cup poss-you LOC sitting ‘ Here is your cup.' (Mbay , Keegan 1997: 76)
c. Bísé lò-í lã âàn. dog poss-you LOC standing ‘ There is your dog.' (Mbay , Keegan 1997: 76)
As with ‘ come' and ‘ go' in the Kalam serial verb construction, so here the verbs ‘ sit' ,‘ stand' , and ‘ lie' constitute the set of verbs which fiII a particular slot in a construction. The three basic human postures underlie the threeway distinction required by the constructions in question. In their use in these constructions, the verbs are extended weII beyond their use as verbs referring to human posture and have become like a set of classifier particles. The pos阳re verbs can also become grammaticalized to function as auxiI iary-like morphemes indicating progressive or continuous aspect (cf Heine et al. 1993; Bybee et al. 1994: 127一 133; Koops , in this volume). In some cases, the ‘ progressive' meaning co-exists with a positional sense , comparable to what we find in the Mbay existential and locative expressions. When aII three verbs develop in this way in a language , there arises a three-way distinction as part of the verbal complex. Indeed, the Mbay verbs
Motivating the uses 01 basic verbs 207
which play such a part in the locative and existential constructions also play a part in constructing the progressive aspect in that language (Keegan 1997: 69f.). As auxiliaries , they can occur with or without the subject pronoun prefixes, but they lack the subject number suffix characteristic of full verbs. In many instances the auxiliaries combine a ‘ progressive' meaning with an indication of posture. So, for example,‘ 1 am working' constructed with the progressive auxiliary ndì refers to work being done while seated, whereas with the progressive auxiliary dà it refers to work done while standing (see examples in Keegan 1997: 70). Of the three verbs , ndì ‘ sit' has the most general use as an auxiliary , this being the one which has generalized most, being used with verbs expressing movement, perception, and communication. So, for example,‘1 am going to the market' is constructed with the auxiliary ndì from the verb ‘ sit' , rather than , say , dà from the verb ‘ stand\ The Kxoé (Khoisan) pos阳re verbs function in a similar way, as described in Köhler (1962: 545; 1981: 530) and Heine et al. (1993). ‘ Sit' ,‘ stand' , and ‘ lie' all reportedly function as present tense markers especially referring to an action performed while sitting, standing, or lying respectively. The auxiliary morphemes , called variously suffixes and particles, are formed from the corresponding posture verbs , which are ‘ defective' , i.e. do not conjugate and are only used in the present. (ln the Kxoé examples, ;tis a nasalized alveolar click and jj is a lateral click.) (1 2)
‘present tense particle , in sitting position' b. ti:, ti ‘ stand, be' :-tè ‘ present tense, in standing position' c. JJoe ‘ lie , be lying' :- jjoè ‘ present tense marker' (in Heine et al. 1993: 139 called ‘ presenttense， habi阳al') d. ;tnùi ‘ sit' :-n ‘ present tense particle, in sitting position or in general'
a.
;tnùi 、it': ，向lUè
(13) a. tí máà, à 前luè ISG see PART sit ‘ 1 see (while sitting).' (Kxoé, Köhler 1962: 545; Heine et al. 1993: 139) b. tí mùù à tè 1SG see PART stand ‘ 1 see (while standing).' (Kxoé, Köhler 1962: 545; Heine et a l. 1993: 139)
208 John Newman
c. tí m i1i1 à - //oè ISG see PART lie ‘ 1 see (while lying).' (Kxoé , Köhler 1981: 530 , Köhler 1962: 545 , Heine et al. 1993: 139) In both Mbay and Kx oé , then , precisely the three main human posture5 familiar from everyday experience underlie a three-way distinction requireè in the verbal complexes of these grammars. Yuchi (Amerindian isolate , possibly Siouan) presents an interesti吨 variation on this kind of patteming in that the three-way distinction baseè on posture verbs is realized within a nominal complex rather than a verba: complex. In this language , the morphemes Ji ‘ sit , stay' , fa ‘ stand' , and ?c ‘ lie' form the basis of a three-way noun classification system (Wagn e:1933-1938: 321; Watkins 1976: 35 日. These three forms function as the articles occurring with singular inanimate nouns , as illustrated in Table .:: The three-way classification also shows up in the formation of relative pronouns. Animate nouns do not participate in this kind of classification. Instead they are sub-classified according to categories of 仕ibal affiliatioL kinship , and gender. Table 2.
Yuchi articles (based on Watkins 1976: 35f. and Wagner 320-322)
ji ‘ sit' class ti}i ‘ the rock' cet，αpaxJi 'my strength' dU呵Îi ‘ my heart'
1933-19三
h ‘ stand' c1ass
Pe ‘ lie' class
yafa ‘ the 位白'
ya?e ‘ the log' s?a?e ‘ the field' ce?eë?e ‘ the rain' hõwedene?e ‘ their language'
yl伪‘ the
house' cewon骂自‘ my spiri t' dicifa ‘ myeye'
Wagner (1 933一 1938: 321) differentiates the functions of the three classifying articles in terms of "objects with a prevailingly vertical dimension" (the fa class) , "objects with a prevailingly horizontal dimension" (the ?e class) , and "objects of a roundish shape or, generally speaking , of a dimension that is indifferent to the ideas of vertical and horizontal" (the Ji class). As is typical of classificatory systems , some members of the category are obviously related to the central meaning of the morpheme which is used to
Motivating the ωes 01 basic verbs 209
indicate the category, while other members are not. Thus, some of the words in the category based on fa ‘ stand' have an obvious vertical orientation such 部‘tree' and 吁louse'; some words in the 'le ‘ lie' category have an obvious horizontal orientation such as 'log' and ‘ field'; ‘ chair' in the Ji class is closely related to the ‘ sit' sense ofJi. However, a semantic basis for all the members of each categ。可 is not apparent. Despite the limited information about these three classes, it is still clear that ‘ sit' ,‘ stand' , and ‘ lie' are the sources 岛r the three mo甲hemes functioning as articles. It is the naturalness and salience of sitting, standing, and lying in ordinary human experience which underlies the choice of just these three concepts to 岛rm the basis of the classification system. In addition, other facts about the Yuchi grammatical system are relevant to motivating 也is aspect of the grammar. Yuchi makes extensive use of suffixes as part of its quite complex morphology. For example, suffixes function to form nominalizations, plurality of nouns , various locative relations such as 'at' ,‘ along' , etc. , verbalizations, verb tenses, modality, conjunction, disjunction, and subordination. A suffixal use of ‘ sit' ,‘ stand' and ‘ lie' 岛rms is clearly consistent with 由is grammatical pa忧em.
4. The experience of bodily positions In the previous section reference was made to the commonality between the 由ree main body positions , a commonality which results in the three verbs ‘ sit' ,‘ stand' , and 'lie' functioning in comparable ways in the grammar. Alongside the commonality, however, there are clear differences between these states in terms of the bodily control which one needs to exercise in order to maintain the position. In the case of standing, both upper torso and lower torso need to be sturdy and held vertical; with sitting it 扭曲e upper torso which needs to be held vertical while the lower torso can be quite relaxed, or even paralyzed; and with lying no p町t of the body needs to be exercising any muscular or bodily control at all. In terms of degree of control needed, then, there is a gradation from standing (requiring most control), through sitting, to lying (requiring least control). Notice that this gradation in degree of control required corresponds, in reverse order, to stages by which children develop, namely lying, then sitting, then standing. And of the three, the standing position, without any additional support, is 由e one which humans are least able to maintain for long periods of time. The control which needs to be exercised is not just a matter of force being ex-
210 John Newman erted upon any particular object, rather it is a combination of control over one's own body and the exercise of balance in a vertical position. Some morphosyntactic facts would appear to correlate with this gradation in degree of control associated with sitting, standing , and lying. As is well known , there are languages which differentiate intransitive verbal predicates according as to whether or not they are agent-like or patient-like (cf. Dixon 1979: 83ff.; 1994: 70-83; Merlan 1985; Mitht日1 1991; Palmer 1994: 65-87). There is considerable cross-linguistic variation in the details of these "splits" amongst intransitive verbs. Our attention here is focused on the pos阳re verbs and initial evidence suggests that when it comes to these verbs , the gradation from standing, through sitting, to lying is respected. That is to say, if there is a split it will occur at some point along this continuum. 1 will briefly discuss four cases illustrating this hypothesis. One possibility is provided by Manam (Oceanic) , where there is an interesting way in which tui ‘ stand' is distinguished from soa?i ‘ sit' and eno ‘ lie\The latter belong to a class of what Lichtenberk calls ‘ state' verbs in Manam , which have ‘ patient' subjects. Semantically, these verbs describe states or changes of state. The class also includes verbs with meanings such as ‘ be big , grow big' ,‘ be bad , become bad' , and ‘ be broken , break (intr.)\ With such verbs , according to Li chtenberk (1983: 219), the continuative aspect of Manam can mean that the state continues to change , that the state habitually , repetitively changes , or that the state itself is habitual , repetitive. The continuative aspect with the ‘ state' verbs cannot, however, mean that the state is in progress at the time of the speech act or at the time of another event. Nor can such verbs appear with the aspect associated specifically with this meaning , i.e. the progressive aspect. State verbs contrast with ‘ active' verbs which have agentive subjects. Examples of active verbs in Manam are the verbs meaning ‘ go' ,‘ work' ,‘jump' ,‘ speak' ,‘ hit' ,‘ give' and tui ‘ stand\Such verbs can be used in the continuative and progressive aspects and both aspects are capable of meaning that the event was in progress at the time of speaking or at the time of another even t. The examples in (1 4) illustrate this difference. Note also that tui in (14a) refers not to the more obviously "active" event of stand
(1 4)
a. áine éne i-tui-tui woman over there 3sG-stand-RDP ‘ The woman is standing over there.' (Lichtenberk 1983: 220)
Motivating the uses 01 basic verbs 211
b. péra ?aná-gu-lo i-en-éno house POSS-lsG-in 3SG-RDP-lie/sleep ‘ He habitually sleeps (lies) in my house. ' No t:‘ He is sleeping (lying) in my house. ' (Lichtenberk 1983: 220)
Thus , in Manam,‘ stand' is aligned with verbs of action involving agents , whereas ‘ sit' and ‘ lie' are aligned with verbs describing states without agents. Again , one may distinguish the linguistic motivation (an established grammatical patteming in Manam) and the extralinguistic motivation (the experiential reality about how we sit, stand, and lie). It is through a combination of these motivations that we can, in part, motivate the Manam grammar relating to pos阳re verbs. Another language relevant here is Tsova-Tush (also called Bats or Batsbi, Caucasian). According to Holisky (1987) , subjects of intransitives in Tsova-Tush can be marked as ergative or absolutive , with various classes of verbs behaving differently in this respect. The perfective verbs ot :ar/旨t:ar ‘ stand up , take standing position' , cu xa?ar/xe?ar ‘ sit down (SG)' , cu xabzar/xebzar ‘ sit down (PL)' , and cu dipsar ‘ lie down' occur with ergative marking, the case marking also associated with the subjects of intransitives such as ‘ come' , ‘ go' , ‘ run away' ，‘ fly' ，‘ a忧ack' ，‘ dance' , ‘jump' , and ‘ crawl'. The imperfective verbs la :tar ‘ be standing' and cuf edagar ‘ be si忧ing (SG)' , cU edaxk'ar ‘ be sitting (PL)' , and cu d伊sar 'be lying down' occur with the absolutive marking , as do verbs such as the copula,‘ go mindlessly , unconsciously' ,‘ fit , suit, go with' ,‘ belong to' ,‘ be hungry' ,‘ be cold' ,‘ be hot' ,‘ be thirsty' , and ‘ be afraid'. There is an interesting difference between ‘ stand' and ‘ sit' on the one hand and ‘ lie' on the other in the kinds of case marking they occur with. The former two can occur with ergative marking (with the same meaning) in addition to absolutive marking and in fact Holisky (1987: 126, fn 6) points out that the consultant preferred the ergative with these two verbs (in the imperfective meaning). The verb cu dipsar , on the other hand , can only be used with the absolutive marking to indicate the imperfective ‘ be lying down'; ergative marking is possible but with the change of meaning to the perfective ‘ lie down'. Here, then, we have (ergative) marking associated with more agentlike participants applying to ‘ stand' and ‘ sit' in a way it does not apply to ‘ lie' . An other possibility is that all three verbs in a language are treated alike with respect to any agent-like versus non-agent-like distinction in the
212 John Newman
grammar. The morphosyntax ofthe posture verbs in Alabama (Muskogean) is an example of this. Alabama makes a distinction between two sets of (non-third person) pronominal arguments on verbs. A1though there are difficu1ties in establishing one semantic generalization conceming the distinction , Davis and Hardy (1 993: 459-4 62) argue for a notion of control (by an actor) versus not in control as being the relevant semantic characterization. So , for example, the pronominal subject markers in the case of verbs like ‘ run' ,‘ listen to' ，吁mg' are drawn from the control set , whereas the pronominal subject markers in the case of ‘ be tall' , ‘ be sick' , and ‘ sneeze' are drawn from the not in control set. Similar1y , the pronominal subject markers in the case of ‘ fall (on purpose)' and ‘ cough (on purpose)' come from the control set, whereas in the case of ‘ fa l1 (accidentally)' and normal ‘ cough' come from the not in control set. The posture verbs ‘ sit' , ‘ stand' , and ‘ lie' interestingly require the contro/ set. Davis and Hardy (1993: 460 f.) appeal to the idea that the ‘ si t' verb chókkòo !i (in the sense of ‘ be sitting') is derived transparently from the process counterpart chokóo /i ‘ to sit down' (in the sense of changing from standing to sitting position). The consonantal gemination and the high tone - falling tone pa忧em in the stative verb , represent an aspectual mo甲heme indicating a continuing state. Davis and Hardy argue that because the basic (= non-derived) verb would be c1 early associated with the control set , SO , too , are the derived forms. However, the more agent-like marking can appear on the three posture verbs in other languages where there is no recourse to this kind of argumentation. For example, in Choctaw (Muskogean, Davies 1986: 37) intransitive subjects may trigger either nominative agreement on the verb (the unergative predicates) or accusative agreement (unaccusative). The posture verbs binili ‘ sit' , hikiya ‘ stand' , and ito/a ‘ lie' all belong to the c1 ass of unergative predicate requiring nominative-type agreement. Other unergative intransitives in Choctaw are ‘ eat',‘ play' ,‘ run' ,‘ dance' ,‘ go' ,‘ bathe' , ‘ swim' , as opposed to unaccusative predicates , which inc1 ude ‘ hungry' , ‘ tall' ,‘ sick' ,‘ cold' ,‘ sweat' ,‘ ashamed\ It seems that one must accept that the pos阳re verbs , as a group , may all be c1 assified as either more agent-like or more patient-like where a language makes such a dist
(‘ talk' , ‘ answer') , th'Ought (‘ guess' , ‘remember' , ‘think' , ‘ dream' , ‘understand' ,‘ imagine'), animate m'Oti'On ('g'O',‘ g'O up' , ‘enter' , etc.). It als 'O includes the p'Os阳re verbs dlδng ‘ s饲nd' ， duek ‘ sit', and cruep ‘ lie 'On st'Omach\Durie's examples with dδng ‘ sta时， and cruep ‘ lie 'On st'Omach' elsewhere in his b'O'Ok (Durie 1985: 174, ex. 7-31; 213 , ex. 8-195) make it c1ear that these verbs can be used to refer t'O states. (The dynamic verbs beud，δh ‘ get up (仕om si创ng or lying)' and êh ‘ lie d'Own to rest 'Or sleep' are als'O contro11ed verbs.) The uncontr'011ed verb c1 ass inc1udes ‘ expl'Ode' ,‘ bec'Ome' , ‘pass, reach' , ‘topple 'Over' , ‘ fa11' , ‘ spi11' ，‘happen，缸rive' ， pa曲。" logical b'Odi1y 'Or mental states such as ‘ itchy' , ‘ tick1 ish' ，‘由unk' ，‘ con fused' ,‘ sick, hurting' and vari 'Ous other categories. Table 3 sums up these facts in a way which suggests a kind 'Of hierarchy in the manner in which the posture verbs enter int'O agent-patient distincti'Ons in languages. It can be no m'Ore than suggestive since the table is based 'On no m'Ore than a handful of languages. Missing fr'Om 也is table at 也is p 'O int in time is evidence of a11 three verbs being c1 assified as patientlike , a c'Onditi'On which would seem as feasible as having a11 three verbs
being classified 臼 agent-like. Table 3 represents , then, no more than an initial, but (I believe) plausible hyp'Othesis. Table 3.
Agent-like morphosyntax (shown in bold) with posture verbs
Manam (‘ active' vs. ‘ state' verbs) Ts'Ova-Tush (ergative vs. absolutive case) Alabama (‘ c'Ontr'O l' vs. ‘ non-c'On位'01' pron'Ouns) Acehnese (‘ contr'011ed' vs. ‘ unc 'Ontro11ed'ageement)
s.
‘ stand' active
‘ sit' state
‘ lie' state
ergative absolutive control
ergative abs'Olutive control
absolutive
controlled
controlled
controlled
control
Conclusion
Experiential reality can be a risky c'Oncept to incorp'Orate int'O acc 'Ounts 'Of grammatical structure and yet ign'Oring it alt'Ogether d叩rives us of the op-
214 John Newman
portunity to fully understand why grammar is structured the way it is. To proceed as if grammar were a completely autonomous system, independent of the larger human context in which it has evolved and in which it serves such vital functions , is to return to an excessively myopic view of grammar. To resist this , it is necessary to venture , however tentatively , into the realm of everyday experience. Some scientific means are available to pursue this kind of study , e.g. , through psychological experiments, anthropological descriptions , etc. , though they are not easily available for every point one wishes to make about everyday experience. In the approach adopted here experiential reality has been understood in a pre-theoretical and relatively naïve way. Nevertheless , 1 have tried to proceed in a responsible way in dealing with the observations. By taking certain linguistic structures as given , 1 have simplified the task of explicating the phenomena in question. It begs the question of how these linguistic structures came to be there in the first place. Ultimately , the linguistic structures assumed in the discussion here need to be accounted for and also motivated by other linguistic and extralinguistic considerations. In this way , one is led to a kind of infinite regression into the very foundations of all language structure. Exactly to what extent all the features of language structure can be reduced to general cognitive principles is unclear, but it is certainly a question that goes well beyond the scope of the present discussion.
References Abraham, Roy C. 1962 Alm-Ar 凹 vius趴，
1993
Dictionary ofthe Hausa Language. 2nd edition. London: Hodder
and Stoughton. Christina The English Verb SEE: A
Study 的 Mult伊le
Meaning. Göteburg:
Acta Universitatis Gothoburgensis. Bybee, Joan , Revere Perkins and William Pagliuca The Evolution of Grammar: Tense, Aspect and Modality. Chicago: Chicago University Press. Craig, Collette G. 1991 Ways to go in Rama: A case study in polygramrnaticalization. In: Elizabeth C. Traugott and Bemd Heine (eds.) , Approaches to 1994
Grammaticalization: Volume II - Focus on Types of Grammatical Markers , 45 6-492. Amsterd am/P hiladelphia: Benjamins
Motivating the uses ofbasic verbs 215 Davies , William D.
Choctaw Verb Agreement and Universal Grammar. Dordrecht: Reidel. Davis, Philip W. and Heather K. Hardy 1993 The semantics of agreement in Alabama. lnternational Journal of American Linguistics 59: 453-472. Dixon, Robert M.W. 1979 Ergativity. Language 55: 59- 138. 1994 Ergativity. Cambridge: Cambridge University Press. Durie, Mark 1985 A Grammar of Acehnese. Dordrecht, Holland/Cinnaminson: Foris Publications. Foley, William A. 1986 Th e Papuan Languages of New Guinea. Cambridge: Cambridge University Press. Foley, William A. and Mike Olson 1985 Clausehood and verb serialization. In: Johanna Nichols and Anthony C. Woodbury (eds.), Grammar lnside and Outside the Claωe: Some Approach臼 ω Theory 卢'om the Field, 17-60. Cambridge: Cambridge University Press. Geiger, Richard A. and Brygida Rudzka-Os切1 (eds.) 1993 Conceptualizations and Mental Processing of Language. Berlinl New York: Mouton de Gruyter. Gibbs, Raymond W. , Din缸a A. Beitel, Michael Harrington and Paul E. Sanders 1994 Taking a stand on the meanings of stand: Bodily experience as motivation for polysemy. Journal ofSemantics 11: 231-251. Haas, Mary R. 1941 Tunica. Bureau ofAmerican Ethnology, B 40 .4. Heine, Bemd, Ulrike Claudi and Friederike HÜßßemeyer 1991 Grammaticalization: A Conceptual Framework. ChicagolLondon: University of Chicago Press. Heine, Bemd, Tom Güldemann, Christa K.i lian-Hatz, Donald A. Lessau, Heinz Roberg, Mathi部 Schladt and Thomas Stolz 1986
1993
Conceptual Shift: A Lexicon of Grammaticalization African Languagl臼. Köln: Ins剑tut fiir Afrikanistik. Holisky, Dee A. 1987
Proc臼'ses
in
The case of the intransitive subject in Tsova-Tush (Batsbi) , L切·
gua 71: 103-132. Johnson, Mark 1987 Th e
Body 切 the Mind: Th e Bodily Basis of Meaning, lmagination, and Reason. Chicago: University of Chicago Press.
216 John Newman Keegan, John M. 1997 ARφrence Grammar ofMb ay. MünchenINewcastle: Lincom. Köhler, Oswin 1962 Studien zum Genussystem und Verbalaufbau der zentralen Khoisan-Sprachen. Anthropos 57: 529-546. 1981 La langue kxoe. In: G. Manessy (ed.) , Les Langues de l 'Afrique Subsaharienne, 485-555. Paris: Editions du Centre National de la Recherche Scientifique. Koops , Christian This vol. Emergent aspect constructions in Present-Day English. Lakoff, George 1987 Women , Fire， αnd Dangerous Things. Chicago: University of Chicago Press. Lakoff, George and Mark Johnson 1980 Metaphors We Live By. Chicago: University ofChicago Press. 1999 Philosophy in the Flesh: Th e Embodied Mind and Its Challenge to W臼tern Thought. New York: Basic Books. Lakoff, George and Mark Tumer 1989 More Th an Cool Reason: A Field Guide to Poetic Metaphor. Chicago: University of Chicago Press. Langacker, Ronald W. 1987/1991 Foundations ofCognitive Grammar, Vols. 1 and 11. Stanford: Stanford University Press. 1994 Culture, cognition , and grammar. In: Martin Pütz (ed.) , Language Contact and Language Conjlict, 25-53. Amsterdaml Philadelphia: Benjamins. Lichtenberk, Frantisek 1983 A Grammar ofManam. Honolulu: University ofHawaii Press. Merlan, Francesca 1985 Split in位ansitivity: Functional oppositions in intransitive inflection. In: Johanna Nichols and Anthony C. Woodbury (eds.) , Grammar Inside and Ou tside the Clause: Some Approaches to Th eory from the Field , 324-362. Cambridge: Cambridge University Press. Mithun, Marianne 1991 Active/agentive case marking and its motivations. Language 67: 51 0-546. Newman, John 1996 Give: A Cognitive-linguistic Study. BerlinINew York: Mouton de Gruyter. 1997 Eating and drinking as sources of metaphor in English. Cuadernos de Filología Inglesa 6.2: 213-23 1.
Motivating the uses of basic verbs 217 Newman, John (ed.) 1998 1页e Linguistics of Giving. AmsterdamIPhiladelphia: Be叼 amins. Norv毡， Peter and George Lakoff 1987 Taking: A study in Lexical Network Theory. In: Jon Aske, Natasha Bee可， Laura Michaelis and Hana Filip (eds.) , Berkeley Linguistics Society: Proceedings of the 13th Annual Meeting Feb. 14-16, 1987: General Session and Parasession on Grammar and Cognition , 195-206. Berkeley: Berkeley Li nguistics Society. Palmer, Frank R. 1994 Grammatical Roles and Relations. Cambridge: Cambridge University Press. Pauwels , Paul and Anne-Marie Simon-Vandenbergen 1995 Body parts in linguistic action: Underlying schemata and value judgements. In: Louis Goossens , Paul Pauwels, Brygida RudzkaOstyn , Anne-Marie Simon-Vandenbergen and Johan Vanparys (eds.) , By Word of Mouth: Metaphor, Metonymy and Linguistic Action in a Cognitive Perspective , 35-69. AmsterdamIPhiladelphia: Benjamins. Pawley, Andrew and Jonathan Lane 1998 From event sequence to grammar: Serial verb constructions in Kalam. In: Anna Siewierska and Jae J. Song (ed时 ， Case, Typology and Grammar: In Honour of Barry J. Blake , 201-227. AmsterdamIPhiladelphia: Benjamins. Radden , Günter 1996 Motion metaphorized: The case of ‘ coming' and ‘ going'. In: Eugene H. Casad (ed.) , Cognitive Linguistics in the Redwoods: Th e Expansion of a New Paradigm 切 Linguistics， 423-4 58 , BerlinINew York: Mouton de Gruyter. Roberts, John R. 1987 Amele. BeckenhamIKent: Croom Helm. 1998 GIVE in Amele. In: John Newman (ed.) , The Linguistics ofGiving, 1-33. Amsterdam/P hiladelphia: Benjamins. Rudzka-Ostyn , Brygida 1993 Introduction. In: Ri chard A. Geiger and Brygida Rudzka-Ostyn (eds.) , Conceptualizations and Mental Processing of Language, 1-20. BerlinINew York: Mouton de Gruyter. Sætherø, Eli 1995 The semantic function of"minimal verbs" in the Kwa languages. Working Papers in Linguistics, University ofTrondheim 25: 6188.
218 John Newman Serra Bometo, Carlo 1996 Liegen and stehen in German: A study in horizontality and verticality. In: Eugene H. Casad (ed.) , Cognitive Linguistics in the Redwoods: The Expansion of a New Paradigm in Lin伊istics， 459-505. BerlinINew York: Mouton de Gruyter. Shen, Ya-Ming 1996 The semantics ofthe Chinese verb ‘ come\In: Eugene H. Casad (ed.) , Cognitive Linguistics in the Redwoods: The Expansion of a New Paradigm in Linguistics, 507-540. BerlinINew York: Mouton de Gruyter. Svorou, Soteria 1993 The Grammar of Space. AmsterdarnlPhiladelphia: Benjamins. Swadesh, Morris 1946 Chitimacha. In: H. Hoijer (ed.), Linguistic Structures of Native America, 312-336. New York: Viking Fund. Sweetser, Eve 1990 From Etymology to Pragmatics. Cambridge: Cambridge University Press. Tumer, Mark 1996 The Literary Mind: The Origins ofThought and Language. Oxford: Oxford University Press. Wagner, Günter 1933-38 Yuchi. Bureau ofAmerican Ethnology, B 40 .3. Watkins , Laurel J. 1976 Position in grammar: Sit, stand, lie. Kansas Working Papers 切 Linguistics 1: 16-41. Welmers, William E. 1973 African Language Structures. Berkeley: University of Califomia Press. Wierzbicka, Anna 1988 Why you can have a Drink when you can't have an Eat. Th e Semantics of Grammar, 293一357. AmsterdarnlPhiladelphia: BenJamms.
Section IV Cognitive motivation
The conceptual motivation of fictive motion* TeenieMα tlock
Abstract Beginning with Talmy's work in the late 1970's and earlyI980's, cognitive linguists have argued that fictive motion - roughly, mentally simulated motion along a path or linear configuration - motivates the use and structure of a class of figurative uses of motion verbs. On this view, a motion verb describes an inherently static scene , as in The road runs along the coast or A trail goes through the desert , but evokes the simulation of "movement" or "scanning" along a trajectory through imagined space. The imagery is believed to be subjective in that the conceptualizer enacts the movement or scanning, and its pu甲ose is believed to be functional - allegedly, simulating motion allows the language user to infer or convey information about the physical layout of a scene, especially the configuration and position of the path or 仕ajectory (e.g. , road). The goal of this chapter is to examine the linguistic behavior of fictive motion constructions (e.g. , The road runs along the coast) , and to discuss whether simulated motion and scanning motivates the use and comprehension of fictive motion constructions. Central to the discussion is relevant work from psychology, including
*
Acknowledgments: Many thanks to Herbert Clark, Raymond Gibbs Jr. , Dan Schwartz, and Leonard Talmy for many enlightening discussions , and to Ravid Aisenmann, Ben Bergen, Frank Brisa时， Maria Cristobol , and Paul Maglio , for insightful comments on early drafts. Thanks also goes to Michel Achard, Adrian Bangerter, Alex Cozzi, and Pemilla Quarfordt for non-English linguis啕 tic examples , and to my research assistants Nicole Albert, Jeremy Elman , Kat Firme, Sydney Gould, Krysta Hays , and John Nolan , for helping with the experiments discussed in this paper. A special thanks is extended to Günter Radden, co-editor of the volume, who provided many helpful suggestions.
222 Teenie Mat/ock psycholinguistic experiments that test fictive motion in English motion verbs.
Keywords: conceptual semantics , fictive motion , figurative language , imagery , linguistic motivation , mental simulation, motion verbs , psycholinguistics , spatial models.
1. Introduction Motion verbs such as go and run are pervasive. Ontologically primary and re f1 ective of one of the most fundamental human activities , motion verbs exist in all languages and show similar pattems of semantic extension crosslinguistically. When used literallyl , a motion verb describes how an agent changes position in physical space. Implicit in the change are the passage of time and continued movement from a starting point to an end point (Miller and Johnson-Laird 1976). For instance , in She drives to Palo Alto from Los Gatos , the driver starts her joumey in Los Gatos and ends her joumey in Palo Alto. En route , she passes through Campbell , Cupertino , Mountain View, and many other towns before reaching her destination. When used non-literally , a motion verb may have little or nothing to do with physical motion. For instance, in Christmas is coming and His mood goes from morose to ecstatic , no movement transpires. Instead , time is understood as "moving" toward the speaker, and one emotional state is understood as "moving" to another (see Boroditsky 2000; Clark 1973; Radden 1996). In some cases , a motion verb may even go so far as to grammaticalize , adopting a new grammatical function. This is the case with English go , which has become a future marker, as in I'm going to drive home and W丑。t are you going to do? (Bybee , Pagliuca and Perkins 1991; Heine , Claudi and Hünnemeyer 1991)? As is well documented, motion verb extensions like these are motivated by conceptual metaphors (e.g. , TIME IS SPACE, CHANGE IS MOTION) , which enable the understanding of a relatively ab-
1. The word "literally" is used somewhat loosely. 1 do not make a hard-and-fast distinction between literal and non-literal language , for such a distinction is problematic , as shown by Rumelhart (1 979) and Gibbs (1 994a, 1994b). 2. See Sweetser (1 990) and Emanatian (1 992) for similar developments in French and Chagga, respectively.
Th e conceptual motivation 01fictive motion 223
stract domain in terms of a relatively concrete one (e.g. , Gibbs 1994b; Lakoff 1987; Lakoffand Johnson 1980; Radden 1997; Sweetser 1990). Motivation is a popular theme in cognitive linguistics. Roughly speaking, it concems the extent to which a linguistic form or set of linguistic forms makes sense or is explainable within a language or across multiple . languages. Motivation encompasses a range of influences on the way language is used and the way it is structured. It includes extemal injluences , such as culture and social context，部 well 臼 intemal influences, such as perception and higher level cognitive processes (e.g. , categorization). Though motivation seeks to explain how language works and how it changes over time, it stands in sharp contrast to generative explanations for linguistic behavior. No rules 缸'e posited and no predictions are made (because there is no surface structure to be derived from a deep struc阳re (Langacker, 1987). As cognitive linguists have shown time and time again , motivation is a perfectly reasonable way to describe and account for linguistic 阴阳ms， inc1uding semantic extensions of verbs. (For in-d叩th discussion ofmotivation, see Chapter 1 in this volume and Lakoff 1987.) This paper concems the motivation of a p町ticul缸 type of motion verb extension, specifically, the type shown in the examples in (1). (1)
a. Th e road runs along the coast. b. A trail goes through the desert. c. The railroad tracks follow the river卢'om Briceberg ω El Portal.
Though common in everyday English，由is 叩pe of semantic extension dese凹es in-depth attention for a number of reasons. Especially remarkable is 由at despite the obvious presence of a motion verb and the potential 岛r motion, no explicit motion is expressed. For instance, (la) describes a stationary situation (nei也er the road nor the coast moves) even though it contains a motion verb (runs) and describes a scene that could easily serve as a good setting for movement - people, vehi c1 es , or other mobile entities could and do frequently travel on roads near coastlines. Tbe lack of motion in this kind of motion verb extension is especially app町ent when comparing sentences such 臼 (la)， (b), and (c)ωsentences such as (2a) , (b) , and (c). The latter set features the same motion verbs and the same spatial scenes, but explicitly expresses movement. (2)
a. Sarah runs along the coast. b. Th e coyote goes through the desert. c. A taxi follows the bus from Briceberg ω
224 Teenie Matlock In these cases , a mover changes position in physical space. For instance , in (2时， Sarah starts running at one location near a coas t1 ine , continues running along the coast1 ine , and eventually stops running at some other location along the coas t1 ine. Her movement occurs in a particular direction and unfolds in real time. lt forms a path with a starting point, an endpoint, and points connecting the two. Thus , the sentences in (1时， (b) , and (c) appear to be quite similar to those in (2时， (b) , and (c) because in both cases there is a path through a spatial scene. However, the difference lies in whether or not motion actually occurs. The semantic extensions of motion verbs in (1 时， (b) , and (c) are not limited to English. When examining motion verbs crosslinguistically, we see many languages with this same type of extension. To name a few , The road runs along the coast (or The road goes along the coast) is expressed in Swedish as Vägen går längs kusten , in ltalian as La strada corre lungo la costa , in French as La route parcourt le long de la côte , and in Spanish as El Camino va por la costa. 3 In such languages, just as in English, a set of motion verbs is frequen t1 y and systematically used by speakers to express stationary spatial scenes that "contain" path-like configurations (e.g. , roads). Despite the obvious absence of actual physical motion in sentences such as (1 a) , (b) , and (c) (as well as comparable motion verb uses in other languages) , cognitive linguists argue that the concep阳alization or construa14 of such sentences involves an implicit type of motion. This implicit type of motion has been called fictive motion by Talmy (1 983 , 1996, 2000) ， αb stract motion by Langacker (1 986) , and su句jective motion by Matsumoto (1 996).5 On the cognitive linguistic view , comprehending a sentence such
3. For comparable examples in Japanese , see Matsumoto (1 996). 4. My use of construal and conceptualization is in line with Talmy (1 975 , 1978 , 1983) and Langacker (1 987). Simply stated, conceptualization is equated with meaning and construal is the ability humans have to view a scene in altemate ways (for instance, from different perspectives, or with focus on different elements). 5. This phenomenon was original1 y studied by Leonard Talmy in the late 1970's and early 1980's. Early on , Talmy (1 983) also used the te口丑 virtual motion to refer to fictive motion. Recently , Langacker (2000) has begun using the term virtual motion for a wide rage of dynamic construal , inc1uding temporal scanning (for instance , "replaying" events in the historical present).
The conceptual motivation offictive motion 225
as (1a) requires the conceptualizer (language user, one who is conceptualizing) to mentally scan or "move" along the path associated with the subject noun phrase, for instance, road in (1 a). In doing so , the concep阳alizer experiences something akin to ac阳al motion , for it involves "going" from a starting point to an end point, and it takes time to complete. The "move- . ment" is believed to be subjective because it is the conceptualizer who induces or experiences the motion in the absence of an explicit agent. The "movement" is also purported1y necessarily enacted for it allows the conceptualizer to zero in on and infer information about the position and orientation ofthe path (see Langacker 1986; Matsumoto 1996).6 The goal of this paper is to investigate whether fictive motion (more generally mentally simulated motion) motivates the use and behavior of motion verbs in sentences like (1 a-c), what 1 will hereafter refer to as FMconstructions (fictive motion constructions).7 1 begin by discussing some of the grammatical and semantic behavior of FM-constructions (Section 2). Then 1 阳m to the dynamics of physical motion, perceived motion, and simulated motion (Section 3). After that, 1 discuss results from experiments that investigated whether fictive motion is actually involved in the comprehension of FM-constructions (Section 4). Last, 1 discuss what type of research needs to be done to achieve a better understanding of mentally simulated motion in language use (Section 5).
6. Fictive motion is not restricted to sentences such as (la) , (b) , and (c). According to Talmy (200时， for instance, fictive motion fig町es into the construal of a broad range of spatial sentences, including There is a bench every now and then in the park. It is also evoked with sentences such as His girlfriends seem to get taller eve.η year， as discussed by Langacker (2002) , and Sweetser (1997). 7. The term FM-construction is used for convenience in this paper. It is not intended as a basic-level construction like the constructions discussed by Goldberg (1995) and other constructional grammarians, although it may be appropriate to treat it as a subset of a more basic construction.
226 Teenie Mat/ock
2. When motion verbs describe stationary scenes8
An FM-cons位uction has the fo l1owing constituents: subject noun phrase (NP), motion verb, and either a prepositional phrase (PP) (e.g. , along the coast in Th e road runs along the coast) or a direct object (e.g. , the creek in Th e road crosses the creek). The subject NP represents the trajector (TR) , which in this case is a path or linear entity such as a road. The motion verb is a verb that specifies (in its literal uses) a change in location. The prepositional object or direct object corresponds to a landmark (LM) (e.g. , the coast as in The road runs along the coast) , or a set of landmarks (e.g. Capitola and Aptos in The road runs from Capitola ω Aptos)， which specifies the location of the TR (ne缸 to and paral1el with a coast).9 The TR is critical in FM-constructions because its construal shapes the overal1 meaning and structure of the construction, including what is general1y considered semantical1y and grammatical1y acceptable to English speakers. In formulating or making sense of an FM-construction, the concep阳al izer expresses or infers information about where the TR is located. In particular, the TR is positioned relative to a LM or set of LM's in the scene being described (see Matsumoto 1996). In (la) , The road runs along the coast, for instance, the road is proximal to and paral1el to the coast1 ine. In (1 b) , A trail goes through the desert , the trail extends from one end of the desert to the other (from the perspective of both the speaker and the listener). In (1 c) , The railroad tracks follow the river 卢om Briceberg ω El Portal, the railroad tracks are p缸al1el to the river. Each of these sentences makes sense because the TR has a concep阳al1y viable or plausible relationship with a LM or set of LM's. For instance, railroad tracks in the real world are often located along rivers , as linguistically expressed in (lc). This is not the case without an explicitly coded LM , as shown in (3a) and (b). (3)
a. ?The road runs. b. ?1页e railroad tracks follow.
8. Some observations (2000).
in 也is
section draw from Langacker (2002) and Talmy
9. In discussing the conceptual structure of FM-constructions, 1 used the terminology of Langacker (1986 , 1987). Talmy's (1975 , 2000) termρ'gure would
also have been fine.
The conceptua/ motivation offictive motion 227
The TR, the road, is not positioned relative to a LM in (3 时， and neither is the TR, the railroad tracks , in (3b). Thus , the oddity of (3a) and (b) arises with the absence of an LM. Unless there is substantial background or implicit context (e.g. , speaker provides iconic gestures for LM and TR or the speaker and listener are looking at a pic阳re)， the conceptualizer is unable to infer information about the configuration, position, or shape of the TR because there is nothing to relate it to. The TR in an FM-construction is usually an inherently rectilinear or spatially extended entity, for instance, road, underground cable, or garden hose , as shown in (4a) 一 (4c). (4)
a. The road goes斤'om Capitola ω Aptos. b. The underground cable follows the property line. c. A garden hose runs along the backfence.
However, when the TR is NOT inherently long, it should have the option of spatial extension, as shown in (5). (5)
a. 1页e table goes from the kitchen wall ω the sliding glass door. b. An a /p ine lakefollows the property line.
c. Thefish pond runs along the backfence. In reading (5 时， for instance, we automatically visualize a table that is long and narrow, perhaps a long oval table or a rectangular table. We do not imagine a small round kitchen table or a perfectly square coffee table. This "restriction" emerges 仕om the inherent concep阳al properties of the construction, including motion information contributed by the motion verb. The "res位iction" is not present with comparable constructions such as The tαble is between the kitchen wall and the sliding glass door or The alpine lake is next to the property line. Similar inferences and restrictions occur with alpine lake in (5b) andfish pond in (5c). The lake and the pond are visualized as being long and narrow. (For more examples, see Matlock, in press.) Examples (6a) and (6b) provide further evidence to show how critical spatial extension is in the concep阳alization of the TR in FM-constructions. Infelicitous forms often arise when the TR is not construed as long , as shown in (6a) with hoola hoop , or (6b) with hot tub. (6)
a. ??The hoola hoop runsfrom the door to the couch. b. ??The small, pe功ctly round hot tub goes along the back fence.
228 Teenie Matlock
Of course there is no problem with (6a) ifthe hoola hoop is bent into a long oval shape , in which case it would actually be a long oval. Similarly, there is no problem with (6b) ifthe concep阳alizer imagines viewing the hot tub from a low , side angle , for instance , while sitting on the ground 30 feet away (as opposed to viewing the hot tub from a bird's eye perspective). 1O Even when the TR in an FM-construction has a long shape, it must be relatively large , as shown in (7a) 一 (d). (7)
a. ??The cell phone goes from the cup to the book. (p hone on desk) b. The cell phone goes from the cup ω the book. (p hone in ad on billboard)
Imagine that the cell phone in (7a) is relatively long, say 2 inches by 4 inches (twice as long as wide). When construed with a phone of that size , most English speakers would agree the sentence sounds odd , but why would this happen when the shape is long? A plausible explanation (l ong shape, but sentence questionable) involves what Langacker (1 986) has called sequential scanning. The basic idea is that to obtain a coherent understanding of a scene , the conceptualizer performs a continuous series of transformations such that one configuration in a construal is transformed into another, and another to another, and so on. This allows the conceptualizer to simulate or build up a representation of the TR in real time or something not unlike real time. Without sequential scanning , the conceptualizer could only activate various locations or points in a construal as discrete and not contiguous. No coherent whole would be achieved. (This sort of gestalt is of course not limited to fictive motion.) Thus , on this view , sequential scanning is minimal or completely absent when conceptualizing a small object like a cell phone even though it is actually rectilinear. Simply stated, there is simply no reason to scan a cell phone because a coherent whole can be obtained with just one glance (real or imagined). Of course the situation for the cell phone changes when the scale is sufficiently large , as shown in (7 b). The sentence is fine when the cell phone is displayed on a large billboard advertisement because there is sufficient "space" in which to
1O .0bviously, (6b) is not problematic if the speaker intends to convey the belonging meaning of go , as in The socks go in that drawer , or Your pα'{Jer goes
in Journal 01 Memory and Language. But then such uses do not involve fictive motlOn.
1页e
conceptual motivation offictive motion 229
mentally scan the TR. A similar explanation applies to paths with motion, as shown in (8a) and (b). (8)
ac阳al
a. ?The sidewalk runs from here to there. (five feet long) b. The sidewalk runs斤。m here to there. (500 feet long)
A sidewalk must be sufficiently long to traverse or dynamically construe over time , that is , sequentially scan from one end ofthe TR to the other. Sequential scanning also explains the inconsistency we see in (9a) and (b) with temporal expressions , such asfor two hours, orfor two seconds.
(9)
a. The road runs along the coast for 2 hours. b. ?The road runs along the coastfor 2 seconds.
The FM-construction in (9a) is fine because it reflects a reasonable amount of time in which to scan along a coastline. In contrast, (9b) does not. Nor does it reflect a reasonable amount of time to actually scan along a section of coastline if a person were running. Of course, one could say (9b) if the intention were to contrast the section of road along the coast with some other section of road (for instance, through a city), as in The road runs through the city for over an ho帆 through the suburbs for 30 minutes, and then along the coastfor only two seconds! The TR in an FM-construction is also (virtually) always stationary, as shown in (1 0a). (10) a. ??Carol runsfrom the door to the couch. b. In the painting, Carol runs from the door ω
the
couch.
The sentence in (lOa) is an odd way to describe Carol if she is , say, imagined to be lying on the floor between a door and a couch (e.g. , head near door, feet near couch) , even ifCarol remains perfectly still. However, (10b) is more reasonable because Carol is a depiction in a painting. One explanation for why the TR must be stationary relates to limitations on processing in visual imagery. In construing (10a) (with a fictive motion reading , not an ac阳al motion reading) , the concep阳alizer must attend to two types of motion - Carol's motion and the conceptualizer's "motion" (i.e. , scanning). An other explanation, and not an incompatible one , is that by default the language user automatically assigns the actual motion
230
1旨enie
Ma t/ock
meaning to the TR, giving it "mover" status. A more interesting example to show this point is provided in (11). (11)
The river runs斤。m the ocean ω
the
mountain top.
When people imagine a river, as in (1 1), it is likely that scanning proceeds in a downward direction , for instance , from the top of the mountain to the ocean. This sort of scanning is consistent with what is generally known about the direction of the flow of water, in particular, that it flows in a downward direction , not an upward direction. Yet, the sentence structure in (11) (with the river as the subject) has us scanning in the opposite direction. What results is a situation in which water motion proceeds in one direction (downhill) , and fictive motion scanning proceeds in the other direction (uphill). Even though the sentence may be difficult to understand at first , it does not present a big problem because the reader or listener eventually se忧les on scanning in the "right" direction (from the coast to the mountain top). Notice that the construal of (11) is not problematic if prior to reading or hearing the sentence, we are informed that the river is totally dry , as in The dry river bed runs斤om the ocean to the mountain top. An other pair of examples shows that the TR must be stationary or at least nearly stationary, as shown in (1 2a) and (b). (12) a. The traffic runs all the way ω Carmel Valley Road. (cars traveling 2 miles per hour) b. ??The traffic runs all the way to Carmel Valley Road. (cars going 40 miles per hour) (1 2a) is felicitous only if the vehicles are not moving or if they are just barely moving. As shown in (1 2b) , the same sentence is no longer felicitous if the traffic is moving , for instance , if the cars are driving 40 miles per hour. (An exception would be if the speaker were imagining viewing the line of traffic from a plane or helicopter hundreds of feet above ground , in which case motion would be less noticeable.) A final observation about the TR in FM-constructions relates to the way the concep阳alizer anticipates motion along a path or what sorts of inferences the conceptualizer makes about paths and various types of motion. Each sentence in (1 a) 一份， for instance , conveys information about a path that is traversable , that is , one that people or other movers travel on or could potentially travel on (e.g. , roa d). Each sentence also features a spatial
Th e conceptual motivation 01fictive motion 231
region that could potentially accornmodate a traversable path (e.g. , coast). For example, in (1 时， The road runs along the coast, the road and the coastline create a good scenario that is conducive to travel. Compare (1 a) to Th e cable runs along the coast or Th e road runs across the glacier. These would not be good travel scenarios , for a mover cannot travel on a cable , . and a road is not ordinarily located on a glacier for obvious reasons. Let us use the term Type 1 fictive motion to refer to FM-constructions such as The road runs along the coast. Type 1 FM-constructions, i. e. , those with paths ordinarily associated with motion, tolerate manner verbs ll , such as crawl and race , as shown in (13a) 一 (13c).
(13) a. The highway crawls through the city. b. The highway races through the city. c. Thefoo伊ath staggers from the bar ω
the
outhouse.
In such motion verb uses , the manner verb does not describe an actual motion event or a particular way of moving. Nothing in (13a) is ac阳ally crawling (unlike in The injured soldier crawled through the city). Rather, the manner verb conveys information about how motion could occur or is known to occur along a given path, for instance, slowly in (13时， and quickly in (13b). This involves a part-岛r-p盯t metonymy whereby certain properties about movement give rise to certain properties about the path. For instance, in (13时， what is known about crawling (slow motion) yields the construal of a path ordinarily associated with slow motion (e.g. , Highway 101 in Northem Califomia has bad traffic at rush hour) , and in (13b) , what is known about racing brings about the construal of a path that is ordinarily associated with fast motion (e.g. , Highway 280 in the Northem California does not have much 仕affic). Specifically, this appears to involve a metonymy ofthe form MOTION ALONG A PATH FOR THE PATH. Finally, the manner verb in an FM-construction goes far beyond conveying information associated with speed (slow 例如t). A manner verb might also highlight information about the emotional or physical state of movers known to travel along a certain path, as shown in (13c). The verb stagger is often associated with inebriation and a particular way of walking. 11. Manner verbs express motion from one point to another but they also convey information about the way the mover moved, for instance, quickly, slowly, erratically, effortfully, whimsically, and so on (see Slobin 1996b; Talmy 2000).
232 Teenie Matlock Namely, a drunk person will walk erratically, zigzagging back and forth , and maybe even stumbling or falling down. The result of this type of motion is a crooked path (for instance, ifthe mover walks across a muddy field and one looks at the footprints left behind). This points to a more specific metonymy: MANNER OF MOTION ALONG A PATH FOR CONFIGURATION OF THE PATH. That is , the way a person walks (e.g. , when in a particular state) shapes the construal of the configuration of a path, even when no ac阳al motlon occurs. Incidentally, FM-constructions such as those in (13a) 一 (c) have a somewhat poetic flavor and are less conventional than many of the other FM-constructions discussed in this paper; however, they are perfectly acceptable in the right context. For instance, (13c) is fine because we know that bars are associated with drinking and an outhouse might be associated with something after drinking. Thus , bar and outhouse make for a favorable reading of the manner verb stagger. This would probably not be the case with Law School and School of Medicine , as in ??The foo伊ath staggers from the Law School to the School of Medicine. In any event, that manner verbs can be used in constructions that describe no 邵阳al motion , and that 由ey provide information about the resultant state of a path, including its shape or how things move along it suggests the presence of meton严ny.12
At this point, it should be noted that, unlike Type 1 FM-constructions, Type 2 FM-constructions 一i. e. ， those with a TR with no association with motion such as (5a--c) (e.g. , The table goes 卢'om the kitchen to the sliding glass door) 一 do not allow manner verbs. Examples of inappropriate Type 2 uses are shown in (l 4a) and (b). (14) a. ??Theflower bed races along the backfence. b. ??The underground cable crawls from Capitola ω Aptos. The oddity of (14a) and (b) arises for the following reason. Manner cannot be construed because no motion is expected to occur or even known to occur along these TR's. And from this it follows that properties associated with motion cannot metonymically map to properties about the shape or configuration of the path, as was the case with Type 1 fictive motion, as in (13a) , (b) , and (c). 12. For similar metonymies along with insightful discussion of the theoretical implications ofmetonymy, see Radden and Kövecses (1 999).
1现e
conceptual motivation offictive motion 233
section, 1 discussed the behavior and use of a number of examples ofFM-constructions. From these observations, it appe町s that the TR is critical to how an FM-construction is construed. In particular, the TR must be relatively long or capable of spatial extension. It is also usually inanimate and bears a spatial relationship to some landmark (for instance, is . P町allel wi曲， is close to). The verb in a FM-construction also plays a key role by contributing in岛rmation about the configuration of the TR. This is especially evident in manner verbs, which metonymically derive information about the properties about the configurations of the TR based on world knowledge about how motion normally occurs along a path or within a particul町 spatial region. Based on these observations as well as many other observations made by Talmy and 0由er cognitive linguists , it seems reasonable to at least entertain the idea that our ability to simulate fictive motion motivates 也e use and behavior of FM-constructions, including what is generally seen as being linguistica11y acceptable. For instance, it is clear that the interpretation of FM-constructions and the way they are used relates to the spatial layout of the scene (e.g. ， τ1豆's position relative to a LM) and information about the configuration of the path (e.g. , influence of manner). However, the problem is that most of the observations made here and many of the observations made by linguists elsewhere simply assume that fictive motion has an actual basis in cognitive processi吨， in particul缸，也at it is grounded in mental imagery. Before we can begin to understand how fictive motion motivates the use ofFM-cons位uctions， we must look at psychological work on mental imagery and mental simulation. Research in those areas may shed some light on fictive motion as a process and help us understand how it motivates sentences like (1 a-c ).13 In 也is
3. Motion: Actual, perceived, and simulated People constantly experience motion and this happens in a number of ways. They can move by their own volition through physica1 space, as in running across a field or walking to work, or 也ey can be moved by some extemal 13. This is not to say 伽.at it is the job of the linguist to do experiments or even cite experimental work. An important job of the linguist is to observe 位ends in a language and across languages and draw conclusions based on those observations (see Gibbs 1991).
234 Teenie Matlock
force , as in riding a raft down a river or being pushed against a wall by a strong wind. People can also move other objects through physical space, as in throwing a ball to third base, pushing a shopping cart off the side of a hill , or pulling a jammed piece of paper out of a printer. Each of these motion events involves actual motion , i.e. , real movement through physical space. People also experience motion through perception, especially vision (Arn heim 1969; Zeki 1993). Humans are constantly visually bombarded with objects coming into and going out ofthe visual field - a humming bird darts past the window or a squiηel scampers across a parking lo t. Humans are so "wired" to process motion that sometimes they see themselves moving when they are completely stationary, for instance, when sitting on a bus and seeing an a司jacent bus start to back up (for discussion of apparent motion , see Ramachandran and Antis 1986). People also readily detect motion through other sensory modalities , including 仕om auditory input, for instance, when a police car or ambulance races by with its siren on. These motion events constitute perceived motion. Both actual and perceived motion involve an agent that moves from one point in space and time to another point in space and time. They also involve direction of motion including front versus back, up versus down , north versus south , or left versus right. Actual and perceived motion also involves a path. When an entity moves through physical space, that entity moves along a pre-existing path (e.g. , paved trail) or creates a path if there is no pre-existing path (e.g. , visible or even invisible footprints in the sand) , and in both cases , a viewer perceives movement along a path (See Casati and Varzi 1999; Miller and Johnson-Laird 1976). People do not only experience motion while going from one place to another, or while watching some other entity go from one place to another. People often deliberately move objects around to help them plan actions , solve problems, and reason about the world. A good example is seen in cognitive science work on forming words from Scrabble tiles. When people are given a set of tiles and told to form as many words as they can, they do a dramatically better job (more words in given allotment of time) when they are allowed to use their hands than when they are not allowed to use their hands. When they manipulate the tiles , players are able to "offload" cognition and do some
The conceptual motivation offictive motion 235
1999). Similar results have been found in longitudinal studies on how people leam to play and improve at the video game Te位is. As Tetris players get better and better at the game , they curiously start doing more and more seemingly superf1 0us movements of game pieces on the screen, for instance, rapidly spinning pieces much more than is strictly necessary to help players access information that could lead to an appropriate placement (e.g. , angle or orientation ofpiece). Yet these movements have been shown to serve a useful function. They somehow help the player maintain state by freeing up working memory, which allows them to attend more closely to other aspects of the game (Kirsh and Maglio 1994). (See Kirsh 1995 ，岛r other ex缸nples of manipulation of spatial arrangements and simplification of perception and computation.) The benefits of physical movement are of course not limited to enacting physical actions in game-playing. They also show up in gesture. Although ges阳res were initially thought to be meaningless or to have no real communicative purpose, recent work in psychology has shown that simple movements , such as pointing at an object or making a downward sweeping movement (e.g. , to describe falling down) , play an important role in language. One set of experiments shows that people are better at solving and explaining math problems when they 缸e 仕ee to use their hands than when they are not. The same work also shows that ges阳res facilitate recall of lexical items (Goldin-Meadows , Nusbaum, Kelly and Wagner 2001). Another set of experiments shows that speakers are better able to access low frequency words or words with elusive meanings when they are allowed to gesture (compared to when they are not) (Krauss 1998). An d other research shows that a speaker and an addressee interact more effectively (e.g. , use fewer words) while working on ajoint project (constructing a Lego model) when they can see each other pointing and doing other gestures than when they cannot, providing further evidence for the idea that hand movements have a cognitive and communicative function (Clark and Krych 2004). Human experience with motion goes beyond ac阳al movement and perceived movement - it also includes mentally simulated or imagined motion. Numerous studies in psychology show that people "move" inside imagined spatial regions. For instance , in one classic study, participants were given a map of an island with various land
236 Teenie Mat/ock
that were c1 0se to each other on the map. Hence , the results suggested that scanning time of locations on the imagined map is proportional to actual distances between locations on the map (Kosslyn, Ball and Reiser 1978). Similar results have been obtained in studies that present participants purely verbal descriptions of spatial regions , suggesting that people can readily construct a spatial model from verbal input alone and mentally scan from location to location in their model (Denis and Cocude 1989). (See Barsalou 1999, 2002 and Glenberg 1999, for other examples of imagined movement through spatial models). Simulated motion is part of the general human ability to experience motion. Just as people manipulate objects in the world to help them plan , think about , and solve physical problems , they also mentally simulate manipulations of objects. The work of Shepard and Metzler (1 971) demonstrates this. In their c1 assic study, participants were told to look at threedimensional geometric shapes in two different drawings. The shapes were either identical or similar and appeared at different angles of rotation. Participants were instructed to specify whether the objects were the same or different. The results showed that the greater the difference in angle of rotation , the longer it took participants to determine whether the two objects were the same, indicating that participants relied on mental rotation to line up the objects so they could make their judgments, and pointing to a more general tendency in people to use dynamic mental imagery to solve problems An other good example of mental "movement" of objects in reasoning is seen in the recent work of Schwartz (1999). In his study, people were given two glasses. The glasses were the same height and had water filled to the same level , but one glass was narrow and the other was wide. The task required participants to judge how far the two glasses would have to be tilted before water would start to pour out. The results showed that people were generally bad at making corr四t judgments about when water would start to pour. For instance, they often mistakenly said that the water would pour from the two glasses at the same angle. However, when people c1 0sed their eyes and imagined tilting the glasses until the imagined water just reached the rim, they were much more likely to make correct judgments, namely , that the wide glass would pour the water first (for related
1页e
conceptual motivation offictive motion 237
4. Simulated motion and fictive motion constructions: Psycholinguistic evidence In the section above, 1 demonstrated that people naturally mentally simulate motion, and showed 也at in some ways , simulations are similar to actual motion and perceived motion. In this section, 1 discuss a series of empirical studies that direct1y bear on the issue of whether mentally simulated motion influences language understanding. Based on 由is work, 1 will argue th创 linguistic behavior of the FM-construction is motivated by the way that people naturally mentally simulate motion. In recent work (Mat1ock, 2001), 1 investigated the comprehension of fictive motion and whether it is influenced by the way people think about physical motion and space. In several reading studies, 1 measured how long it took people to process FM-sentences such as The road runs along the coast in a v缸iety of contexts. After reading about protagonists traveling through rugged and cluttered terrains , people took longer to read FMsentences than 由ey did after reading about protagonists traveling through smooth and flat terrains. People were also slower to read FM-sentences after reading about slow travel (versus fast) and long-distance travel (versus short-distance). The overall results suggest that in understanding an FMsentence, people re-activate and simulate 臼pects of the protagonist's motion, including speed, distance, and the terrain across which the movement occurred. In doing so, they construct a dynamic representation 由at mirrors the actual motion of the protagonist. That the same FM-construction which expresses no actual motion - is processed differently depending on the protagonist's movements provides evidence for fictive motion in the understanding of motion verbs. Similar results were obtained in off-line experiments with drawings designed to test whether imagery played a role in understanding figurative motion verb cons位uctions (Matlock, in press). In one experiment, p町tici pants drew pictures of sentences such as The lake runs between the go扩 course and the railroad tracks and The lake is between the go扩 course and the railroad tracks Gudged as describing the same scene in a norming study prior to the experiment).ηle pu甲ose was to get at people's conceptions of the scene and to obtain a non-linguistic measure of their conceptions. 14 Of 14. Drawings are extemal representations of people's conceptions of the world, and as such, they provide insights into how people concepωalize objects, 归.tes， and actions (Tversky 1999, 2001).
238 Teenie Matlock
special interest was whether the trajector (for instance, lake) had a more elongated or extended shape in the former case (with motion verb) than in the latter case (with a copula verb). A difference would suggest that mental scanning (or at least more mental scanning) would occur with FMconstructions than with non-FM-constructions. The results showed uniformly longer shapes in the former case , which provides indirect support for the idea that people mentally simulate motion in understanding FMconstructions. Another experiment in the same study had participants draw similar pairs of sentences, but in this case, the subject-NP's were entities that are inherently long , for instance, roads , mountain ranges , and so on. The same results were obtained. And yet another experiment in the same set of studies had participants simply draw an arrow to represent the roads (or other paths, such as trails) they visualized while reading sentences such as The road jets from one vista point to another or The road crawls from one vista point to another. The results showed reliable differences in the way people drew arrows to represent their understanding of the path in the sentence. For instance , longer arrows were produced for sentences with fast verbs than for those with slow verbs. Importantly, no sentences conveyed any motíon. In a later study (Matlock in progress) , people drew longer arrows for sentences about cluttered terrains than for ones about uncluttered terrains , for instance, The road goes through the crowded city (cluttered) versus The road goes through the desert (uncluttered). One explanation is that simply knowing that there was more clutter "slowed" down people's simulation, resulting in a slower hand and a longer line. An altemative explanation, however, is that the cluttered terrain involved a more complex 可pe of simulation, whereby people anticipated each of the obstacles they visualized. If this is the case , longer lines may reflect time taken to anticipate obstacles. In any event, these results , along with those from the two other drawing experiments, suggest that people activate motion information while reading and conceptualizing static scenes verbally depicted byfigurative uses of motion verbs. Further work, including on-line studies with pictures, need to be done to gain a better understanding of how motion activation proceeds in real time , and how it compares to other imager
1页e
conceptual motivation oj 卢'ctive motion 239
natural ability to simulate motion. As Langacker (2000) notes , it is intriguing that motion verbs such as go and spatial prepositions such as to and 卢om - typically associated with motion along a path - are regularly used in FM-constructions. 1 think this is no accident. Nor is it an accident that there are several restrictions and contingencies relating to the use of such conS位uctions， as outlined in an earlier section. For instance , sufficient length and scale are important because simulation, like real motion or perceived motion, takes time to complete. Therefore, one cannot simulate motion along an extremely short path such as a sidewalk that is only five feet long , but one can along a sidewalk that is , say, 500 yards long , as shown in (8时， The sidewalk runs from here to there. Finally, direction is important. In thinking about a long garden hose, as shown in (4c) , A garden hose runs along the back fence , 1 scan in one direction, going from one end (probably where the faucet is located) to the other (where the water comes out). 1 do not start at the middle of the hose and move out in both directions at the same time. Thus, in mental scanning, as experienced in understanding (many) FM-constructions, scanning usually proceeds in one direction. Further evidence to support the idea that people activate conceptual structure about motion when there is no explicit motion involved comes 企om other psychological work on figurative language about motion verbs and understanding abstract domains. Maglio and Matlock (1999) examined how people talk about the experience of using the World Wide Web. The studies were run in 1996, before the popularization of the web. Many had never used a Web browser and a few people had never even heard of the Web. When asked what their experience was like following a short session browsing the web , both novices and expert web users described the experience with language that was much like actual motion through physical space, for instance, 1 went ω a web site about tattoos , 1 stayed there for α while and then came back. In related work, Matlock and Maglio (1996) found that language canonically used to express movement along a horizontal plane was preferable to language about vertical movement (exception is vertical movement on same screen, but in the early days of web use, less information was presented on a single page, and consequently, there was less need to scroll) , suggesting that people use language that mi
240
Teenie 儿latlock
5. Conceptual motivation of fictive motion constructions In this paper, 1 have argued that simulated motion is grounded in and driven by cognition and perception. People run simulations because they can't help but do it. From this natural ability and people's on-going experiences with motion in the world , it follows that the use ofFM-constructions , such as The road runs along the coast , is motivated. However, we still have a long way to go before we can show exactly how fictive motion motivates the grammatical constructions. A good starting point might be to look at how simulation works across a variety of grammatical constructions, taking into account the sorts of schemas discussed by Bergen and Chang (in press). In addition , empirical work needs to be done on the more subtle underpinnings of fictive motion understanding , induding the extent to which it is subjective. Langacker (1990) argues that subjectivity is an integral part of construal of sentences such as The mountains run from Canada to Mexico , in that the language user, in the absence of an explicit mover, does the moving. The first question that comes to mind is what does subjectivity mean for on-line processing? Subjectivity is plausible and certainly compatible with the fictive motion verb experiments 1 have discussed here and elsewhere (e.g. , Matlock 2001) , but none of those tasks targeted subjectivity per se. Nonetheless , we can still see that the language user mentally scans a tr句 ectory， and that information about actual motion is activated in that simulation (e.g. , rate , distance , teηain). Further empirical work should also examine reconceptualization (Langacker 2002). Simply stated, reconceptualization is the idea that people have to mentally trace a path or some portion of a scene more than once , for instance, to scan in two directions. For instance , upon hearing The scar runs from her 切 ee to her ankle , my scanning starts at the knee and continues until 1 get to the ankle. But to understand The scαr runs to her knee from her ankle , 1 must first scan from the knee to the ankle , and then scan from the ankle to the knee. If Langacker (2002) is right about reconceptualization , we should see processing differences between the two sentences. Namely , the latter should take longer to process, even though on the surface the sentences appear to convey the same meaning. In any event, this may provide further evidence for or against the idea that people subjectively experience fictive motion For comparative pu甲oses ，
The conceptual motivation 01fictive motion 241
is whether languages that have traditionally been called "verb-仕amed"， such as Spanish - in which path information is expressed primarily or vir阳ally exclusively in the verb 一 will differ in terms of on-line processing from languages such as English and other "satellite-企'amed" languages 一 m which path is expressed primarily in adpositions and other verb "satellites" (see Slobin 1996a, 1996b; Talmy 2000).15 Langacker (2000) argues that understanding FM-sentences requires the concep阳alizer to use sequential scanning (roughly , build up a representation in steps by "moving" from point to point along the trajectory). In verb-企amed languages , such as Spanish, motion verbs tend to fall toward the inceptive or instantaneous end of the continuum. In this way , they are more like English exit and enter, which background the sense of on-going activity, and instead highlight the point at which a mover changes position, for instance, the point between being in one location and being in another location , as in He exited the room. It is possible that the binary flavor of these verbs requires less scanning time than verbs such as go , follow , meander, race , and so on. On this same note, it would also be useful to see how construal relating to boundary-crossing constraints (Slobin and Hoiting 1994) plays out in the on-line understanding of FM-sentences. For instance, if people were given a sentence such as The road crosses the property line , and primed with an explicit physical boundary, such as a concrete fence , it might make for slower processing times than if they had been primed with some sort of implicit non-physical boundary. Finally, it is important to think about motivation and fictive motion in light of language as a joint activity and people's need to communicate with others (Clark 1996). Think about when you would actually say (1 时， The road runs along the coast. You would probably not walk up to a total stranger in a library or say it out of the blue at the dinner table. But you might say it if somebody approached you on Pacific A venue in Santa Cruz, Califomia, and asked, Excuse me, where does this road go?, while pointing at a line representing Highway 1 on a map. The point is that FMconstructions are obviously contextually motivated. They occur in situations in which two or more people are talking about where something is or 15. This is oot to say that the satellite- versus verb-企amed laoguage dichotomy uniformly applies across languages. A good example of a "problematic" laoguage is Thai , which is traditiooally believed to be a verb-framed language. Kessakul (1999) cooclusively shows that it has maoy satellite-framed elemeots. Other examples are discussed by Croft (2002).
242 Teenie Matlock where they are going. For example , imagine you are driving in an unfamiliar area , and you look at a sign that says Highway 49. You ask your friend , who is sitting in the back seat looking at a map , W古 ere does that go? Your friend could respond with any ofthe utterances shown in (15a) 一(1 5c).
(1 5) a. On this map, it runs north from here. b. Looks like it goes to several gold rush towns. c. lt crosses the Merced River and eventually gets to Highway 120. Or here is another situation that shows that context is critical to how these sentences are used and how they are understood in the real world. In walking on campus with a colleague in an area neither of us knows well , we step on to a sidewalk that appears to continue around a comer. As we tum the comer, he says , Let's see where this goes. After tuming the comer we leam that the sidewalk dead-ends. 1 say, lt doesn 't go anywhere. What is important here is that FM-sentences like these constantly come up in everyday conversation. Their use is motivated because such utterances arise as people a忧empt to establish and maintain common ground (Clark 1996). When we hear them, we usually instantly know what the speaker is talking about. For instance , in hearing (1 5坊， 1 know that it refers to a road, specifically Road 49 , and that it does not refer to a bus or information (The bus goes to sever，α 1 gold rush towns , The information goes to several gold rush towns). An d when there is a problem , such as ambiguity or insufficient information , addressees can always ask their addressors for more information (e.g. , Oh , you mean that road over there? while pointing at the road). Thus , it is probably safe to assume that FM-constructions are useful to establishing common ground when people are conversing about spatial layouts and navigation, even though experimental results have not yet been published on the topic.
6. Conclusion That motion verbs are systematically used to describe stationary scenes is not at all trivial or random. Based on the systematic behavior of FMconstructions in addition to psychological work on mental imagery, simulated action , common ground, and more directly , on experimental work on fictive motion , we can conclude with a reasonable degree of confidence that FM-constructions are motivated by our cognitive ability to mentally
Th e conceptual motivation 01fictive motion 243
simulate motion along a path and by the natural urge to talk about where objects are located and where we are going in the wor1 d. In addition, FMconstructions are motivated by the need to move in the wor1 d , not merely for convenience or pleasure, but for surviva1. Not being able to move can be potentially dangerous (e.g. , not being able to run out of a burning building, being unable to obtain food) , and so is not being able to see motion (e.g. , getting hit in the face by a hard ball because you did not see it coming your way, driving offthe side ofthe road because you didn't notice the car pulling into yo町 lane). There岛re， the conc1usion that fictive motion is motivated is really not all that remarkable, especially given that our natural abi1ity to simulate motion ultimately lies in our direct experience with motion in the wor1d. Our abi1ity to simulate motion is merely a reflection of the need to be in motion and to perceive motion. Finally, our ability to use and make sense of language about motion in non-literal ways goes far beyond the types of sentences we looked at in this paper. Nonetheless, the same explanation applies: It simply reflects the primacy of motion in human experience and the embodiment of that experience in linguistic thought.
References Arnheim, Rudolf 1969 Visual Th inking. Berkeley, CA: University ofCalifornia Press.
Barsalou, Lawrence W. 1999 Language comprehension: Archival memory or preparation for situated action? Discourse Processes 28: 61-80. 2002 Being there conceptually: Simulating categories in pr叩aration for situated action. 10: Nancy L. Stein, Patricia J. Bauer and M. Rabiowitz (eds.) , R写presentation， Memory, and Development: Essays in Honor 01Jean Mandler , 1-15. Mahwah, NJ: Er1baum. Bergen, B.K. and N.C. Chang in press Embodied construction grammar in simulation-based language understanding. 10: Jan-Ola Östman and Mirjiam Fried (eds.), Construction Grammar(s): Cognitive and Cross-Language Dimensions. AmsterdamIPhiladelphia: Benjamins.
Boroditsky, Lera 2000 Metaphoric structuring: Understanding time through spatial metaphors. Cognition 75: 1-28.
244 Teenie Matlock Bybee, Joan L., William Pagliuca and Revere Perkins 1991 1加 Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago, IL: University of Chicago Press. Casati , Roberto and Achille C. Varzi 1999 Parts and Places: The Structures of Spatial Representations. Cambridge, MA: MIT Press. Clark, Herbert H. 1973 Space, time, semantics , and the child. In: Timothy E. Moore (ed.) , Cognitive Deve/，叩ment and the Acquisition of Language , 27-63. San Diego: Academic Press. 1996 Using Language. New York, NY: Cambridge Press. in press Pointing and placing. In: Kira Sotaro (ed.) , Pointing: Wh ere Language, Culture, and Cognition Meet. Hillsdale, NJ: Erlbaum. Clark, Herbert H. and Meredith Krych 2004 Speaking while monitoring addressees for understanding. Journal ofMemory and Language 50: 62-81. Denis , Michel and M. Cocude 1989 Scanning visual images generated from verbal description. European Journal ofCognitive Psychology 1: 293-307. Emanatian, Michele 1992 Chagga ‘ come' and ‘白': Metaphor and the development of tense-aspect. Studies in Language 16: 1-33. Gibbs , Raymond W. 1991 What' s cognitive about cognitive linguistics? In: Eugene Casad (ed.) , Cognitive Linguistics in the Redwoods , 27-53. The Hague: Mouton de Gruyter. 1994 a Figurative thought and figurative language. In: Morton A. Gemsbacher (ed.) , Handbòok ofPsycholinguistics , 411 -446. San Diego, CA: Academic Press. 1994 b The Poetics ofMind: Figurative Thought, Language, and Understanding. Cambridge: Cambridge University Press. Glenberg, Arthur M. 1999 Why mental models must be embodied. In: Gert Ri ckheit and Christopher Habel (eds.) , Mental Models in Discourse Processing and Reasoning, 70-90. New York, NY: North-Holland. Goldberg, Adele E. 1995 Constructions: A Construction Grammar Approach ω Argument Structure. Chicago: University of Chicago Press. Goldin-Meadow, Susan, H. Nusbaum, S.D. Kellyand S. Wagner 2001 Explaining math: Gesturing lightens the load. Psychological Sc ience 12: 51 6-522.
The conceptual motivation 01fictive motion 245 Heine , Bemd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammatica /ization: A Conceptual Framework. Chicago: Uni峭 versity of Chicago Press. K臼sakul， Ruetaivan 1999 Two faces of linguistic encoding in Thai motion events: Evidence from Thai spoken narrative discourse compared with Japanese. In: Graham Thurgood (ed.) , Papers Irom the Southeast Asian Linguistics Socie纱， 70-86. Arizona State University. Kir吐1， David 1995 The intelligent use ofspace. Artificial /ntelligence 73: 31 -6 8. Kirsh, David and Paul P. Maglio 1994 On distinguishing epistemic 企om pragmatic action. Cognitive Sc ience 18: 513-549. Kosslyn , Stephen, T. Ball and B. Reiser 1978 Visual images prese凹e metric spatial information: Evidence 台om studies of image scanning. Journal 01 Experimental Psychology: Human Perception and Pe价rmance 4: 47-60. Krauss , Robert 如t 1998 Why do we gesture when we speak? Current Directions in Psychological Science 7: 54-60. Lakoff, George 1987 Women , Fire, and Dangerous Th ings: What Categories Reveal about the Mind. Chicago, IL: University of Chicago Press. Lakoff, George and Mark Johnson 1980 Metaphors W泣 Live By. Chicago , IL: University of Chicago Press. Langacker, Ronald W. 1986 Abstract motion. Proceedings 01 the Twelfth Annual Meeting 01 the Berkeley Linguistics Society, 455-47 1. Berkeley, CA: Berkeley Linguistics Society. 1987 Foundations 01 Cognitive Grammar, Vo l. 1: Th eoretical Prerequisites. Stanford, CA: Stanford University Press. 2000 Virtual reality. Studies in the Linguistic Sciences 29: 77-103. D归amicity，负ctivity， and scanning: The imaginative basis of 2002 logic and linguistic meaning. Korean Linguistics Today and Tomorrow: Proceedings 01 the 2002 /nternational Co价rence on Korean Linguistics, 3-32. Seoul: Association for Korean Linguístícs. Maglio , Paul P. and Teenie Matlock 1999 The conceptual structure of information space. In: Alan J. Munro , K. Hook and D. Benyon (eds.) , Social Navigation ol/nlormation Space , 155-173. London: Springer-Verlag.
246 Teenie Matlock Maglio, Paul P., Teenie Matlock, Dorth Raphaely, Brian Chemicky and David Kirsh 1999 Interactive skill in scrabble. Proceedings 01 the Twenty -first 剧， nual Cognitive Science Society, 32ι330. Mahwah, NJ: Lawrence Erlbaum. Matlock, Teenie 2001 How real is fictive motion? Unpublished doctoral dissertation. University of Califomia, Santa Cruz. in press Depicting fictive motion in drawings. In: J. Luchenbroers , (ed.), Cognitive Linguistics: Investigations across Languages, Fields, and Philosophical Boundaries. AmsterdarnlPhiladelphia: BenJanuns. in prog. Drawing fictive motion. Matlock, Teenie and Paul P. Maglio 1996 Apparent motion on the World Wide Web. Proceedings 01 the Eighteenth Annual Cognitive Science Society. Mahwah, NJ: Lawrence Erlbaum. Matsumoto, Y0 1996 Subjective motion and English and Japanese verbs. Cognitive Linguistics 7: 183-226. Miller, George A. and Philip N. Johnson-Laird 1976 Language and Perception. Cambridge , MA: Harvard University Press. Radden, Günter 1996 Motion metaphorized: The case of ‘ coming' and ‘ going\In: Eugene Casad (ed.), Cognitive Linguistics in the Redwoods: The Expansion 01 a New Paradigm 切 Linguistics， 423-458. Berlinl New York: Mouton de Gruyter. 1997 Time is space. In: Birgit Smieja and Meike Tasch (eds.) , Human Contact through Language and Linguistics, 147-66. New York , NY: P. Lang. Radden, Günter and Zoltán Kövecses 1999 Towards a theory of metonymy. In: Klaus-Uwe Panther and Günter Radden (eds.), Meωnymy 切 Language and Thought, 1759. AmsterdarnlPhiladelphia: Benjamins. Ramachandran, V. S. and S.M. Antis 1986 The perception of 叩parent motion. Scientific American 254: 102-109. Rumelhart, David E. 1979 Some problems with the notion of literal meanings. In: Andrew Ortony (ed.), Metaphor and Thought, 78-90. New York, NY: Cambridge University Press.
The conceptual motivation offictive motion 247 Schwartz, Daniel L. 1999 Physical imagery: Kinematic versus dynamic models. Cognitive Psychology 38: 433 -4 64. Schwartz, Daniel L. and Tamara Black 1999 Inferences through imagined actions: Knowing by simulated doing. Journal of Experimental Psychology: Learning, Memory, and Cognition 25: 11 6- 136. Schwartz, Daniel L. and John B. Black 1996 Analog imagery in mental model reasoning: Depictive models. Cognitive Psychology 30: 15 4-219. Shepard, Roger N. and Jaqueline Metzler 1971 Mental rotation of three-dimensional objects. Science 171: 701703. Slobin, Dan 1. 1996 a From ‘ thought and language' to ‘ thinking for speaking'. In: John J. Gumperz and Stephen C. Levinson (eds.) , Rethinking Linguistic Relativity, 70-96. Cambridge , UK: Cambridge University Press. 1996 b Two ways to travel: Verbs of motion in English and Spanish. In: Masayoshi Shibatani and Sandra A. ηlOmpson (eds.) , Grammatical Constructions: Their Form and Meaning, 195-219. New York, NY: Oxford University Press. Slobin , Dan and Nini Hoiting 1994 Reference to movement in spoken and signed languages: Typological considerations. In Proceedingsof the 2 (jh Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Li nguistics Society. Sweetser, Eve E. 1990 From Etymology ω Pragmatics: The Mind-as-Body Metaphor 切 Semantic Structure and Semantic Change. Cambridge: Cambridge University Press. 1997 Role and individual readings of change predicates. In: Jan Nuyts and Eric Pederson (eds.) , Language and Conceptualization , 116136. Cambridge: Cambridge University Press. Talmy, Leonard 1975 Semantics and s泸ltaX ofmotion. In: John P. Ki mball (ed.) , Syntax and Semantics , Volume 4, 181-238. New York: Academic Press. 1978 The relation of grammar to cognition - a synopsis. In: David Waltz (ed.) , Proceedings ofTINLAP-2, 14一24. New York: Association for Computing Machinery.
248 Teenie Matlock How language structures space. In: Herbert L. Pick, Jr. and Linda P. Acredolo (eds.)，牛atial Orientation: Theory, Research, and Application , 225一282. New York: Plenum Press. 1996 Fictive motion in language and ‘ ception'. In: Paul Bloom, Mary A. Peterson, Lynn Nadel and Merrill F. Garrett (eds.) , Language and 牛ace， 211-276. Cambridge, MA: MIT Press. 2000 Toward a Cognitive Semantics, Volume 1: Conceptual Structuring Systems. Cambridge, MA: MIT Press. Tversky, Barbara 1999 What does drawing reveal about thinking? In: John S. Gero and Barbara Tversky (eds.)，的'sual and 年atial Reasoning in D臼ign， 93一 10 1. Sydney: Key Centre of Design Computing and Cognition. 2001 Spatial schemas in depictions. In: M. Gattis (ed.) , Spatial Schemas and Abstract Th ought, 79一 112. Cambridge, MA: MIT Press. Zeki, Semir 1993 A Vision 01 the Brain. London: Blackwell Scientific Publications. 1983
The goal bias in the encoding of motion events* Anatol Stefanowitsch and Ada Rohde
Abstract This paper investigates the claim 也at there is a goal bias in the encoding of motion events, i.e. that there are restrictions on the distribution of path-PPs expressing the source or traject。可 of a motion event that do not hold for path-PPs expressing the goal (岛r example，也at goal-PPs are more likely than other path-PPs to occur as the only locative in a clause). On the basis of co甲us-derived 仕'equency data, we investigate two explanations which have been put forth 岛r 由is bi脑: (i) that there is a psychological bias towards goals and pu甲oses of human actions, and (ii) that goal平Ps have a higher information value than other path-PPs and are thus more useful in arriving at a complete conceptualization of a motion event. We show 也at the second hypothesis makes more accurate predictions about the presence and the extent of the goal bias with different types of motion verbs. Keywords: construction, co甲us linguistics, frequency da饵， gapping, goal bias, locative, motion event, motion verb , path, windowing
*
The authors would also like to point out that since the submission of this paper in 2000, they have refined their use of quantitative co甲us data far beyond what is presented here; however, the basic points made in 也is paper still hold. The authors would like to thank G阳ter Radden, Klaus-Uwe Panther and Sally Rice for discussion and valuable suggestions.
250 Anatol Stφnowitsch and Ada Rohde
1. Introduction It has been c1 aimed that there is a goal bias in the encoding of motion events , i.e. that there is an asymmetry between spatial PPs encoding the goal and those encoding the trajectory or the source of the moving entity. Take the following examples (cf. Verspoor, Dirven and Radden 1999: 88): (1)
a. 1 climbed from my room up the ladder onto the roof b. 1 climbed onto the roof c. 1 climbed up the ladder. d. ??1 climbed斤。mmyroom.
Example (1 a) explicitly mentions all components of a path: the source the room) , the trajectory (up the ladder) , and the goal (onto the rooj). As (1 b) and (1 c) show, it is possible to express just the trajectory or just the goal , but it is semantically odd or even unacceptable to express just the source (Verspoor, Dirven and Radden's acceptability judgments). This phenomenon is an example of gapping, as developed in Talmy (1985 , 1996). We wi11 briefly summarize the main points of Talmy's framework here , and then rephrase the apparent constraint evident in examples (la-d) in terms ofthese notions. Talmy assumes that the conceptualization of motion events necessarily inc1 udes four components: (i) afigure (which we w i11 refer to as the theme) , i.e. an individuated object of some kind; (ii) the motion of this object; (iii) a path along which this motion takes place , consisting of an initial portion (the source) , a medial portion (which we call trajectoη) ， and a final portion (the goa l), and (iv) a ground, i.e. one or more landmarks with respect to which the motion is conceptualized, and which e1 aborate one of the three components of the path. In addition , a motion event may inc1 ude two optional components: manner of motion and cause of motion , which , however, are not relevant for the pu叩oses of this paper. As just mentioned , Talmy assumes that the first four components are necessarily present in the concep阳alization of any motion even t. This does not mean , however, that they are necessarily present in the linguistic material coding the conceptualization of the motion event; they may be highlighted (or windowed for attention) , in which case they w i11 be explicitly encoded, or they may be backgrounded (or gapped), in which case they w i11 not be encoded. (斤'om
The goal bias in the encoding 01 motion events 251
Thus , we can rephrase the constraint evident in examples (1 a-d) as follows: it is possible to window the tr离jectory or the goal by itself, gapping all other parts of the path, but it is not usually possible to window just the source, gapping traject。可 and goal. If there is such a cons位aint or distributional preference, the question arises as to its motivation (at least in a 仕amework like cognitive linguistics, which holds that grammar is to a large extent motivated). In the cognitive linguistic literature, two such motivations have been suggested. The first suggestion comes from Verspoor, Dirven and Radden (1999) , who , following Ikegami (1987) , assume that this distributional bias is motivated by the existence of a "goal-over-source principle", i.e. a "strong hierarchy in the everyday experience of the "source-path-goal" schema: for human actions the goal is usually more important than the source and the source and goal [together] are usually more important than the path" (Verspoor, Dirven and Radden 1999: 98). More generally , Verspoor, Dirven and Radden suggest that any human activity "involves human volition and we tend to be far more interested in the goal of the action than in the source of the action. Therefore, when human action is involved, goal is far more salient than source" (Verspoor, Dirven and Radden 1999: 98). In other words , they explain the distributional bias shown in examples (la-d) above by positing the existence of a natural psychological bias toward the goals (and purposes) ofhuman actions. They seem to suggest that this bias is an inherent part of the make-up of the human conceptual system, i.e. they do not provide further explanation for this psychological bias itself. We will refer to this analysis as the (psychological) salience hypotheSlS.
The second explanation comes from Ungerer and Schmidt (1 996). In a discussion of a particular kind of path windowing, which, following Talmy , they refer to as closed path windowing , they observe a similar bias towards the goal. They suggest that this dis创butional bias is motivated by the higher information value of goal-PPs: "[…] the fact of a departure taking place does not include enough information about the rest of the path to ensure that the whole motion event is realized" (Ungerer and Schmidt 1996: 225). Their account is based on the implicit assumption, shared by Talmy as well as Verspoor, Dirven and Radden, that the interpr l
252 Anatol St，φ nowitsch and Ada Rohde
whereas the medial and final portions of paths allow for an inferential conceptualization of the entire path, the information contained in the initial portions is not sufficient to establish the whole ensuing path" (Ungerer and Schmidt 1996: 226). In other words , Ungerer and Schmidt suggest that if we know the goal of a motion , we can infer enough about its traject。可 (which must lead to the goal) , and perhaps even its source , to arrive at a complete conceptualization of the motion event. 1 If their assumption that every motion event must be conceptualized in its entirety is correct, then the goal bias can be given an explicit psychological (cognitive) motivation: the encoding of the relatively informationpoor source raises the cognitive as well as the communicative costs; the cognitive cost because there is more inferencing to be done on the part of the hearer , and the communicative cost because the possibility that the hearer will make the right inferences is relatively low. 2 This is essentially the insight implicit in Grice's communicative maxims and made explicit in Sperber and Wilson's Relevance Theory (1 995). We will refer to this type of explanation as the complete-conceptualization hypothesis. In this paper, we will establish the nature and extent of this apparent bias on the basis of several co甲us analyses. We will examine whether the kind of acceptability statements given for the examples in (1 a-d) turn out to be descriptively adequate , i.e. whether they hold true for all kinds of motion verbs. We will then interpret the data in terms of the two hypotheses presented above.
1. This idea also seems to be the implicit assumption behind Ikegami's (1 987) account of the goal bias , which he characterizes as follows:
If we hear that something has started, we are still left with an expectation to be told that it has arrived at a certain point. Otherwise it will be felt incomplete as a description of a motion. On the other hand , if we hear that something has arrived at some place and ended its motion there , we feel quite satisfied with the description in spite of the fact that we are not told about the start ofthe motion (l kegami 1987: 135). 2. Maybe this is what Ikegami means when he says that "a clause representing the source is [...] less autonomous and more uncertain than a clause representing the goal" (1 987: 131).
Th e goal bias in the encoding 01 motion events 253
2. Corpus analysis 1 Aims and method. Our first aim was to establish whether the distributional bias discussed above could be empirically verified at all. ln order to do so we chose the word go as a relatively neutral (and thus presumably representative) motion verb. We extracted 1,000 instances of this verb 仕om the North American News co甲us and classified them according to 也e following p缸ameters: (i) literal motion (i.e. physical motion of a theme through space) vs. non-literal motion (i.e. subjective motion, metaphorical extensions, etc.); and (ii) presence or absence of spatial prepositional phrases, which were further subcategorized into source, trajectory , goal, as well as any combination of these. Prepositions classified as referring to the source were from , out ωand off, prepositions classified 臼 referring to 也e trajectory were, for example, along, through, and by, and prepositions classified as referring to the goal were to, into, toward，βJ， etc. PPs speci马ring a generallocation where the motion takes place were classified as referring to 也e trajectory, e.g. They swam in the lake. ηlÌs decision may be open to dispute, but since the main focus in 由is paper is on source and goal , a diι ferent way of classifying such PPs would not substantially affect our results. Where a preposition can have different readings, each token was categorized according to the meaning it had in context. For ex创nple， They cruised off the coast 01 Greece would be classified as trajectory, whereas They jumped offa clif.f would be classified as source. Results. Of the 1,000 tokens 226 were discarded because they were instances either of the going to-future or of the idiomatic constructions goverb and go and-verb. The results for the remaining tokens are shown in Table 1.
The following results emerge from this analysis: First, uses of go with more than one spatial PP are very rare; where they occur at all, they encode the source and the goal of the motion. Uses encoding the complete path, i.e. so町'ce-位离jectory-goal， do not occur at all. Second, as predicted by the two hypotheses discussed above, there is clear evidence for a preference of goal-PPs over trajecωry- and source-PPs. Goal-PPs make up the vast majority ofthe literal uses, and the largest single group ofthe non-literal uses.
254 Anatol St吃fanowitsch and Ada Rohde Table 1. Frequencies of the types of PPs occurring with literal and non-literal uses of go (n=774)
No path adverbial Source Trajectory Goal Source-source Source-traject。可
Source-goal Trajectory-位ajectory
Trajectory-goal Goal-goal Source-traiectorv- I!oal Total
Literal 12 .1 6% (36) 5 .4 0% (1 6) 4.05% (12) 77.36% (229) (0) (0) 1. 00% (3) (0) (0) (0) O 100.00%(296)
Non-literal 8.37% (40) 6.07% (29) 37.24% (1 78) 45.61% (218) (0) (0) 2.10% (10) (0) (0) 0 .42% (2) O 100.00% (478)
There are three kinds of exceptions to the general trend. First, cases where the source is explicit1 y mentioned in the preceding or subsequent discourse and is hence recoverable , as in examples like He knows that 扩 he goesfrom his home and hangs out at the nearby corner of Church and Flatbush in Brooklyn , N Y., he could get in trouble, or worse. In this example , the clause containing go only mentions the source of motion (斤。 m his home) , but the goal (t he nearby corner) is supplied in the next clause. The second type of exception are cases where the goal is recoverable from world knowledge , as in examples like The slight tremor of the plane as it went down the runway was caused by hot brakes. Again , the goal is not explicit1 y mentioned , but it can easily be recovered; since we know that , for a plane, a runway only has two possible endpoints (the terminal building if the plane has just landed, or the sky if it is just taking off), and since we also know that a plane is unlikely to brake during take-off, we know that the goal must be the terminal building. The third type of exception are cases where the goal could be said not to constitute an aspect of the motion event at all , or at least where the goal has such a low saliency that it does not actually need to be recovered as in 阳en his parents went out of town , he quickly cut loose or in One deputy glimpsed Lee going out of sight. Here , the goal is not mentioned at all because it is not necessary for a conceptualization of the motion events described. In the first case , it simply does not matter where the tr句 ectory of
1页e goal bias
in the encoding 01 motion events 255
his parents ends, and in the second case，也e point is knows where the trajectory of Lee ends.
precisely 由at
no one
Discussion. On the whole, the results of the first co甲us search can be seen 臼 evidence for a distributional bias toward goals. For the literal uses , sentences containing a goal-PP constitute the vast majority. For the non-literal uses, such sentences do not constitute the majority but still constitute the largest single group. However, this bias is far from absolute: sentences containing just a source-PP are rare but nevertheless they occur, and are in fact even more frequent than those containing a trajectory-PP. Let us therefore look at these exceptions to the goal bias in more detail. The frrst two types of exceptions mentioned above do not posit a problem 岛r either of the two theories discussed above: since the goal is recoverable at no great cognitive cost it makes sense that it should be possible to gap it in the encoding of the motion event. In contrast, the third type of exception posits a challenge. Note that for these uses the goal is recoverable neither from the context nor from world knowledge. In fact, it would be more accurate to say that the goal in these cases is completely irrelevant, 也就 it is simply not a (salient) p缸t of the event described. Go in these examples is used with the somewhat specialized meanings ‘ leave' and ‘ disappe缸'. It seems that these uses of go do not make reference to a goal at all , but 由at instead 也ey are inherentiy oriented towards the so町ce ofthe motion by virtue of the frame evoked by their lexical semantics. This suggests that there may be verb-inherent biases 也at go against the general trend. We decided to investigate this possibility in a second co甲us analysis focusing on directional motion verbs.
3. Corpus analysis 2 Aims and method. In order to test our hypothesis th创 the specific (lexical) frame semantics of a given motion verb may interact with the apparent goal bias discovered 如也e "neutral" motion verb go , we picked six verbs 由at intuitively have some inherent directionality associated wi曲曲em: climb andjlee (which seem to be goal-oriented), fall and 臼cape (which seem to be source-oriented), and cruise and stroll (which seem to refer to relatively aimless motion). We did not, at this point, speculate about the source of these intuitions (see Discussion below). In addition we picked four verbs
256 Anatol Stl伽owitsch and Ada Rohde 由at
make reference to the manner of motion rather than some direction: 舟，
jump , run , and march. Note that we did not pick verbs like enter and exit, i.e. ‘ path-conflating' verbs in Talmy's terminology, since these do not typi-
cally take spatial PPs but encode their respective path window as a direct object. For each of the verbs chosen, we ex位acted 250 tokens of literal uses with spatial PPs from the North-American News co甲us and classified them according to the type of PP (source , trajectory, goal , or any combination of these). Resu /ts. Of the combinations of two spatial PPs , only the combination
source-goal occurred with a frequency higher than one percent for any of the verbs, thus we collapsed all other combinations into a single category. Interestingly, the complete path (source-trajectory-goal) did not occur at al l. The 企'equencies of all other types of PP are shown in Table 2. Table 2.
Frequencies for the type of PPs occurring with the literal uses of each verb Source
Tr电jectory
Goal
Source-goal
Others
Total
100% (250)
c1 imb
15.6% (39)
17.6% (44)
64.0% (161)
2.0%(5)
0.4% (1)
flee
18.8% (47)
6.0% (15)
73.2% (183)
1.2% (3)
0.8% (2)
fall
36.0% (90)
15.2% (38)
47.6% (119)
1.2% (3)
escape
58.0% (145)
6.0% (15)
34.0% (85)
1.2% (3)
0.8% (2)
100% (250)
crulse
2.0% (5)
74.8%(184)
17.2%(43)
2.0%(5)
4.0% (1 0)
100%(250)
stroll
2.8% (7)
64.8% (162)
27.2% (68)
2.8% (7)
2.4%(6)
100% (250)
100% (250) 100% (250)
fly
8.0% (20)
28.4% (71)
54.0% (135)
9.6% (24)
jump
36.0%(90)
19.6%(49)
34.0% (85)
10.4% (26)
100% (250) 100%(250)
run
30.8%(77)
25.6% (64)
40.8% (102)
2.8%(7)
100% (250)
march
8.0% (20)
40.8%(102)
43.2% (108)
6.8% (1 7)
1.2% (3)
100% (250)
The results confirm our hypothesis that the lexical semantics of a given motion verb may strongly influence the distribution of source-, trajectory- , and goal-PPs. Roughly, our verbs fall into three groups. The first group consists of c/imb, fall, flee , fly, march and run. These roughly follow the pa忧em we already observed with go , in that tokens with goal-PPs constitute either the absolute majority or at least the largest group for each of these verbs , typically followed by source-PPs, with trajectory-PPs forming the smallest group in all cases except for fly and march. However, note th创 the
The goal bias in the encoding 01 motion events 257
bias is significantly weaker than for go in the cases of run and fall , where source-PPs are almost as frequent as goal-PPs , and in the case of march , where 位'ajectory-PPs are almost as frequent as goal-PPs. The second group is comprised by jump and escape, for both of which source-PPs form the largest single group , with goal-PPs forming the second-largest group. Fi- . nally , for the third group , consisting of stroll and cruise, trajectory-PPs form the absolute majority, with goal-PPs again making up most of the remainder. Discussion. The intuitions about item-specific directional biases of individual verbs are confirmed under the assumption that a general goal bias exists in addition. We assumed that fall and escape were inherently sourceoriented. For escape, this assumption is clearly bome out, and even though goal-PPs are more 企equent than source-PPs for fall , the difference between the two is very small. Thus , uses like He escaped from Alcatraz are more frequent than uses like He escaped to the United States , and uses like They jumped into the water are more frequent than uses like They jumped from the pickup truck. Likewise , the assumption that stroll and cruise both encode relatively undirected motion tums out to be correct. Uses like He strolled around his rivet:，斤ont property are more frequent than uses like He strolled to a nearby knoll , and uses like Th ey were cruising aimlessly around are more 企equent than uses like We cruised into Glacier Bay. The issues that need to be addressed are the source of these intuitions, which we assume have to do with the lexical semantics of the respective verb , and the reason why the lexical semantics interfere with the goal bias. This will be one of the central questions in the General Discussion below. Essentially, the answer depends crucially on an answer to the question whether it is in fact true , as is implicit in Talmy's and others' work, that every motion verb evokes a complete path, i.e. whether every motion event involves the conceptualization of a complete path from a source along a trajectory to a goal. Before we discuss this issue further , however, we must take up the issue of animate versus non-animate themes , which is implicitly argued to be relevant to the goal bias by Verspoor, Dirven and Radden.
258 Anatol St电fanowitsch and Ada Rohde 4. Corpus analysis 3
Aims and method. Recall that Versp 'O'Or, Dirven and Radden p 'O sit the existence 'O f a g'O al bias in p町ticular with respect t'O m 'O ti 'O n events inv'O lving human themes. The aim 'O f this analysis was t'O establish whether there is indeed a difference between human and n 'On-human themes with respect t'O the distributi'O n 'Of g'Oal-enc 'Oding PPs. In 'Order t 'O test this at a generallevel, we ch'Ose the m 'O ti 'On verb move (instead 'O f go , which 'Occurred extremely infrequent1y with inanimate themes). We extracted 50 literal examples each f'Or 由e f'Oll'Owing 叩pes 'O f themes: animate) (‘change 'O ne's p 'O siti 'On' as in She moved away from the window) , animate2 (‘change 'Of 'One's residence' as in He moved to Washington , D. C.), vehi c1 e as in The train moved 归ω the station , and inanimate as in Heavy rain clouds moved into the area. As bef'Ore, we ca1culated the relative prop'Orti 'Ons 'Of the different types 'Of pathPPs f'Or animate and inanimate themes separately. Results. The results 'O f the analysis are sh'Own in Table 3. C'Ombinati 'Ons 'O f m 'Ore than 'One PP were c 'O llapsed int'O a single categ'Ory. A pair-wise c 'Omparis'On 'O f all categ'Ories sh'Ows that the difference between each pair is significant at the 0.1 % level , except f'Or animate)/inanimate, which is significant at the 1% level. 3 Table 3. Frequencies of the types of PPs occurring with animate and inanimate themes in literal uses ofmove (n=774)
An imate) An imate2 Vehicles Inanimate
Source 14% (7) 12% (6) 12% (6) (0)
Trajectory 28% (14) (0) 14% (7) 38% (19)
Goal 42% 76% 6% 28%
Others (21) 16% (8) (38) 12% (6) (3) 68% (34) (l 4} 34% (17)
Total 100% 100% 100% 100%
(50) (50) (50) (50)
As predicted by the salience hyp 'Othesis , the g 'O al bias is s位onger f'Or animate themes than f'Or inanimate themes. The latter seem t 'O prefer 的lject'Ory PPs. N 'O te als'O由at with respect t'O animate themes , the meaning ‘ change 3. The exact statistics are: animate)/animate2 (χ2=19，26 (df=2) , p
1页e
goal bias 切
the
encoding ofmotion events 259
one's residence' has a much stronger preference for goal-PPs than the meaning ‘ change one's position\Vehicles have no strong preference for any single type ofPP. Discussion. Although the results seem to provide initial confirmation for the salience hypothesis, there is one finding that warrants further discussion: As mentioned, the two different senses of move with animate themes differ significantly with respect to the strength ofthe goal bias they display. This is not expected under the salience hypothesis since both senses of move are actions when the theme is animate. An explanation might arguably again be found in the specific frames evoked respectively by ‘ changing one's position' and ‘ changing one' s residence'. Intuitively, it seems that an event of ‘ changing one's residence' is harder to conceptualize in the absence of information about the goal: if we hear that somebody has moved we typically want to know where that person has moved. An event of ‘ changing one's position', however, seems less dependent on information about the goal; since this type of event is less specific we can easily imagine contexts \vhere the goal simply does not matter. If this intuition is correct and the different streng由 of the goal bias is indeed influenced by frame-semantic considerations, then this is clearly an argument in favor of the complete-conceptualization hypothesis: what counts as a ‘ complete concep阳alization' depends on the specific frame associated with the verb in question, and hence a goal bias may be present to varying degrees. In our final corpus analysis , we will now attempt to substantiate this intuition.
5. Corpus analysis 4 Aims and method. In order to tease apart the two hypotheses further, we must be more explicit about the predictions that each of them makes with respect to animate and inanimate themes , respectively. The salience hypothesis predicts that animate themes should show a S衍onger affinity to goal-PPs with all motion verbs that can be construed as actions. The complete-conceptualization hypothesis makes somewhat more complex predictions: first, there need not be a general goal bias: individual verbs may have a goal- or a source bias depending on the specific frame they evoke. Whether or not animate and inanimate themes behave differently with any particular verb depends on the type of motion event evoked by the combination of a given verb with either an animate or an inanimate
260 Anatol S.向自nowitsch and Ada Rohde
theme. In other words , a given verb may evoke different types of frames depending on the type oftheme. For example，卢y overwhelmingly evokes a JOURNEY frame with human themes , as in They decided to cut their trip short and jly back to New York, but it typically evokes a general MOTION THROUGH AIR frame with inanimate themes , as in Pieces of glass were 卢!y ing through the air. In order to test these predictions we chose five motion verbs which occur relatively frequently with both animate and inanimate themes: jly , roll , and escape , which intuitively encode actions with human themes and processes with inanimate themes , and slide and fall , which encode processes regardless of the type of theme. For each verb we extracted one hundred literal uses from the NAN , fifty with animate themes and fifì可 with inanimate ones. Results. The salience hypothesis predicts that the verbsjly , roll , and escape should show a clear goal bias for animate , but not for inanimate themes , while slide and fall should not show a goal bias for either. The completeconceptualization hypothesis predicts different types of biases based on the particular frames evoked. Consider Tables 4 through 6, which show the results for jly, roll and escape. Table 4.
F!y
Goa! Others I Tota! 74% (37) 18% (9) I 100% (50) 40悦。0) 28% (1 4) 12% (6) I 100% (50) 42% (21) 18% (9) 12% (6) I 100% (50) (χ2=48.19 (df=哟， p<ο.001 ， ***) - without vehicles: (χ2=42.94 (df=匀， p
Trajectory 2% (1)
***)
Table 5. Roll
Source Trajectory Goal Animate I 6% (3) 36% (18) 38% Vehicles I 10% (5) 42% (21) 30% Inanimate I 8% (4) 48% (24) 24% (χ2=4.03 (df=哟， p=0.8 , n.s.) - without vehicles:
Others I Total (1 9) 20% (10) I 100% (50) (1 5) 18% (9) I 100% (50) (1 2) 20% (1 0) I 100% (50) (χ2=2， 58 (df=匀， p=0 .4 6, n.s.)
The goal bias 切 Table 6.
the
encoding of motion events 261
E8cape
Source Trajectory Animate I 80% (40) 2% (1) Inanimate I 66% (33) 4% (2) (χ2=2 ，50 (df=匀， p=0.28， n.8.)
Goal 18% (9) 30% (15)
Others
Total 100% (50) 旦旦豆豆i
Clearly the data do not bear out the prediction of the salience hypothesis; although for fly the goal bias is indeed stronger for animates than for vehic1es or inanimates, for roll and for escape there are no significant differences at all (in the case of escape there is actually a trend in the opposite direction). Next consider the results for slide andfall (recall that these are not naturally construed as actions , and that there should thus be no differences between animates and inanimates according to the salience hypothesis). Table 7.
Slide
Trajecto Source 8% (4) 38% (19) Inanimate 18% (9) ~2% (16) (χ2=4.00 (df=匀， p=0.26 , n.8.) An imate
Table 8.
Goal 46% (2月 48% (24)
Other8 8% (4) 2%
Total 100% (50)
Goal 40% (20) 34% (17)
Other8 24% (12) 22% (11)
Total 100% (50) 100% (50)
100~i?O)
Fall
Trajectory Source Animate 32% (1 6) 4% (2) 38% (19) 6% (3) Inanimate (χ2=0.74 (df=匀， p=O. 邸， n.8.)
Again there are no significant differences between animate and inanimate themes. This is theoretically compatible with the salience hypothesis but since this hypothesis was not confirmed for the first three verbs , the results for fall and slide cannot be seen as very strong evidence. Discussion. A general goal bias for animates is not confirmed by these results: Process verbs (such asfall and slide) do not show a significant difference; this would be expected under both hypotheses. Verbs that can be either action-like or process-like depending on the animacy of their theme
262 Anatol St，笔fanowitsch and Ada Rohde
do not provide evidence for a general goal bias either. Although for 卢y there is a strong goal preference for animates , but not inanimates, there is no such significant difference for roll , even though the distinction between animate and inanimate themes here also correlates with the doing vs. happening distinction. Finally, for escape there is actually a trend in the opposite direction: if anything , the goal preference here is stronger for inanimate themes. Where there are clear differences , these are readily explained, however, by the complete-conceptualization hypothesis. Let us look at the results for Jly (the only verb for which there is a significant difference) and for escape and slide, where the differences fail to reach significance but where p
Th e goal bias 切
the
encoding 01 motion events 263
destination is not recoverable 仕om the other aspects of the path. This diι ference might account for the fact 也at inanimate themes tend to have a stronger goal affinity than animate themes wi也 escape. Fina11y, reca11 由e stronger tendency of slide wi也 inanimate themes to encode the source compared to the same verb with animate themes. We have no specific explanation to offer, but note 由at the completeconceptualization hypothesis at least offers a framework 岛r such an explanation whereas the salience hypothesis does not.
6. General discussion A complex picture has emerged with respect to 也e distribution of source-, trajectory- , and goal-PPs 臼 they occur with English motion verbs. The goal-over-source principle was basica11y confirmed but as a tendency rather than an absolute rule. In addition, the exceptions to the principle cannot simply be seenω ‘ noise' in 也e data; 也ey are themselves systematic and can thus shed some light on the possible motivations behind the goal-oversource principle 蹈 discussed in the In位oduction. In our opinion, Ungerer and Schmidt's suggestion, under the psychological interpretation that we have assumed above, is a plausible candidate when it comes to accounting for 也.e data. It natura11y accounts for the cases adhering ωthe goal-over-source principle, as it was designed ω. In addition, howev町， it can account for the exceptions as follows. First，也e cases where the goal is recoverable from 也e linguistic context 缸'e s位aightforward enough. In order ωconceptualize a motion event in its entirety the necessary path information does not have to take the form of a spatial PP. Second, the fact 也at in those cases where the goal can be inferred 企om world knowledge it does not have to be encoded is also accounted for. If we hear a sentence like He fell from the rooJ, we can conceptualize the complete path, including the trajecto叩 and the goal, without explicitly being told. Our knowledge about gravity will enable us to infer at no great cognitive cost 由at 由e 甜甜 ect。可 is down and the goal is the ground. Thesetwo 可pes of exceptions could be accounted 岛r by any motivation that we may suggest, since in both cases the goal is a salient p町t of the concep阳alization and thus these cases do not rea11y constitute exceptions to the goal-over-source principle at a11. However, there is a third type of case which clearly constitutes an exception: the case of verbs like cruise and
264 Anatol Stφnowitsch and Ada Rohde stroll, which occurred mainly with trajectory-PPs , and a case like escape, which occurred mainly with source-PPs. These cases are clearly different from the other two types of exceptions in that the goal is not just left implicit, but that it simply plays no role. Note that in sentences like They were cruising up and down Main Street or He was strolling through the park neither the goal nor the source can be recovered from linguistic context or world knowledge , and that in a sentence like He escaped from Alcatraz , neither the subsequent path nor the precise goal are recoverable. However, it is unproblematic to arrive at a complete conceptualization of the motion events described by these sentences. We can concep阳alize somebody strolling or cruising without conceptualizing a source or a goal because both verbs refer to kinds of motion that are executed for their own sake , simply because they are enjoyable. Likewise , we can conceptualize someone escaping from a prison without knowing exactly where they go after they have escaped (beyond the fact that they are now outside of the prison). The conceptualization of an act of escaping crucially involves certain obstacles that have to be overcome (s町， walls that have to be dug through , a stretch of water that must be crossed , etc.). Thus, there is an implicit trajectory involved, which is inferable from our knowledge about prisons , but the concep阳alization of the event is complete even if we do not know the ultimate endpoint of this tr句 ectory. With verbs like these , Ungerer and Schmidt's hypothesis would lead us to expect that a goal bias would not be found. If the conceptualization of the motion event does not include a goal , then we do not need information about the goal. Thus , the existence of such verbs has consequences for the question what constitutes a ‘ complete motion event'. It becomes clear that this question cannot be answered in the abstract; it depends to a large degree on the semantics of the specific motion verb used to encode the event. It is simply not the case that every motion event is conceptualized as having a source , a trajectory , and a goal. Assuming that all these components are necessarily present in the concep阳alization of a motion event even if they are not encoded linguistically is essentially a version of the objectivist fallacy that meaning is out there in the world. After all , there are no events in the ‘ real world'. Events are the result of a cogniti
Th e goal bias 切
the
encoding 01 motion events 265
Thus , it seems that the goal bias simply reflects the prototype case, in which the goal provides most of the information needed to arrive at a complete conceptualization of a motion event. Where this is not the case, i.e. where a motion event can be conceptualized in its entirety without reference to a goal , the goal bias is considerably weakened or disappears altogether. We can thus explain much of the goal-over-source principle by reference to a more general cognitive principle which essentially says that an utterance must contain enough linguistic clues to arrive at a complete conceptualization of the event encoded (this is essentially a re-formulation ofGrice's maxim of quantity). This does not mean that the goal bias is necessarily a mere epiphenomenon of this more general principle. Specifically, it does not mean that Ikegami's idea of a general interest in the goals and pu甲oses of human activity (as taken up by Verspoor, Dirven and Radden) is wrong. We have not investigated here the possibility that the goal bias is stronger in the case of the self-propelled motion of an animate agent than in the case of the externally caused motion of an inanimate object. Future research may well yield such findings , which would substantiate the salience hypothesis. Some evidence for the existence of a goal bias beyond the extent to which it can be derived 仕om a general cognitive principle may come 丘om a comparison of in位ansitive versus caused-motion uses of the types of verbs discussed here: Rohde (2001: 169), using the same co甲us and the same categorization criteria used here , found that the caused-motion use of move occurs with a goal-PP in 66.8% of all cases compared to our 42% (see Table 3 above). The caused-motion construction requires an intentionally acting causer acting on the theme (see Goldberg 1995: 174, Stefanowitsch 2001: 296 , Rohde 2001: 188f.) whereas intransitive motion does not include intentionality as an obligatory component at all. Thus the caused-motion construction is naturally concemed with human goals and pu甲oses and should have a stronger goal bias according to the salience hypothesis. However, Rohde (2001: 187) found 43.2% of goal-PPs with caused-motion uses of march, almost exactly the same proportion that we found for the intransitive use of march in Table 2 above. Thus further research clearly needs to be done on this issue. The general bias towards the goal even in those cases where
266 Anatol St，吃fanowitsch and Ada Rohde
eral schema would inherit the goal bias that can be found in most of the instances from which it was abstracted. Of co町se， this does not preclude the possibility of an additional motivation for the goal bias along the lines suggested by Ikegami and Verspoor, Dirven and Radden. The two explana蝴 tions are not mutually exclusive. The relation between conceptual structure and linguistic structure is complex and multi-dimensional , and multiple motivations ought to be expected.
References Goldberg, Adele 1995 Constructions: A Construction Grammar Account 01 Argument Structure. ChicagolLondon: The University of Chicago Press. Ikegami , Yoshihiko 1987 ‘ Source' vs. ‘ goal': A case of linguistic dissymrne位y. In: René Dirven and Günter Radden (eds.) , Concepts 01 Case, 122-146. Tübingen: Naη. Langacker, Ronald 1987 Foundations 01 Cognitive Grammar. Vol. 1. Th eoretical Prerequisites. Stanford: Stanford University Press. Rohde, Ada 2001 Analyzing path: Th e Interplay 01 Verbs , Pr，ξpositions and Constructional Semantics. Ph.D. dissertation, Department of Linguistics, Ri ce University, Houston, TX. Sperber, Dan and Deirdre Wilson 1995 Relevance: Communication and Cognition. Oxford/Cambridge , MA: Blackwell. Stefanowitsch, An atol 2001 Constructing Caωation: A Construction Grammar Account 01 Ana命tic Causativ，阻 Ph.D. disse阳.tion， D叩artment of Linguistics , Ri ce University, Houston, TX. Talmy, Leonard 1985 Lexicalization pa忧ems: Semantic structure in lexical forms. In: Timothy Shopen (ed.) , Language Typology and 砂ntactic Description, Volume 3, 57-149. Cambridge: Cambridge University Press. 1996 The windowing of attention in language. In: Masayoshi Shibatani and Sandra A. Thompson (eds.) , Grammatical Constructions , 235-287. Oxford: Clarendon Press.
The goal bias in the encoding 01 motion events 267 Ungerer, Friedrich and Hans-Jörg Schmidt 1996 An lntroduction to Cognitive Linguistics. London: Longman. Verspoor, Marjolijn, Re时 Dirven and Günter Radden 1999 Putting concepts together: Syntax. In: René Dirven and Marjolijn Verspoor (eds.) , Cognitive Exploration 01 Language and Linguistics , 87-115. AmsterdamIPhiladelphia: Benjamins.
Motivating the composition of Afrikaans reduplications: A cognitive grammar analysis Gerhard B. van Huyssteen
Abstract 1 This contribution is aimed at an explanation of the motivation for the composition of Afrikaans grammatical and onomatopoeic reduplications within a cognitive grammar 仕amework. It is illustrated that we can adequately describe the formation of reduplications in terms of four valence factors , viz. correspondence, profile determinacy , autonomy/dependency asymmetry , and constituency. By claiming that reduplications in Afrikaans are motivated largely by metonymies like MORE OF FORM FOR MORE OF CONTENT, WHOLE THING FOR A PART OF THE THING, TWO PROCESSES FOR AN INDEFINITE NUMBER OF PROCESSES, and PRODUCT FOR PRODUCER, an indication of the relation of metonymy to grammar is given , thereby offering a psychologically plausible and natural description and explanation of the motivation for Afrikaans reduplications.
1. This contribution is based on work that 1 have done for my PhD thesis. My promoters , Ari e Verhagen, Daan Wissing , and Adelia Carstens , have contributed immensely to shaping my thoughts. Parts of this contribution have also been presented at the 7th Intemational Cognitive Linguistics Conference (Santa Barbara, USA, in July 2001) , and 1 would like to thank colleagues for useful comments, especially David Tuggy, Ted Sanders, and Bertus van Rooy. Special thanks also to G缸lter Radden, who has commented extensively on earlier versions of this contribution, and provided me with new insights regarding the relationship between reduplication and metonymy. Needless to say that all errors and fallacies remain mine.
270 Gerhard B. van Huyssteen K句!Words:
Afrikaans, autonomy, cognitive grammar, composition, constituency, coηespondence， dependence, metonymy, profile determinacy, reduplication
1. Introduction In the debate on arbitrariness vs. motivation in language , examples of reduplication are often listed to illustrate iconicity in language (see for example Keller 1998; Kimenyi 1989 and Tai 1993 , among others). It is argued that "form and meaning resemble each other in a quantitative respect, which is to say that the form of reduplications is in a sense non-arbitrary or motivated" (Botha 1988: 3). This claim has been further supported from a cognitive-semantic perspective by suggesting that there is a general metonymic relationship between form and meaning in the sense that WORDS [stand] FOR THE CONCEPTS THEY EXPRESS (Lakoff and Tumer 1989: 108). On the basis of this metonymy , the implication MORE OF FORM FOR MORE OF CONTENT is justified, with reduplications providing an excellent case in point (Lakoff and Johnson 1980: 127-128). However, Kδvecses and Radden (1998: 74) point out that the relation ofmetonymy to grammar has not been worked out in enough detail and depth , and that it deserves further attention. With special reference to this relationship between metonymy and grammatical constructions , the motivation for the composition of A企1kaans reduplications is explored in this contribution. Two types of Afrikaans reduplications will be considered here , viz. socalled grammatical reduplications (examples 1, 2, and 匀， and onomatopoeic reduplications (examples 4 and 5):2 2. A third type of reduplication can also be identified , viz. echo-word reduplications, as in ietsie-bietsie ('very little/a dash of/a touch of), and woerts-warts (‘ very quickly; moving very swiftly'). Unlike grammatical reduplications and onomatopoeic reduplications , echo-word reduplications can be considered a more prototypical case of composition. In echo-word reduplications an autonomous constituent combines with a dependent constituent (where the dependent constituent exists on the grounds of its relation to the autonomous constituent), a profile determinant, as well as the order in which the constituents combine , can be identified (i.e. the first component structure dictates the mere existence of the second one , and determines the profile of the composite structure) , and only partial coηespondence is present on the phonological pole , and no correspondence on the semantic pole (i.e. in monosyllabic words coπespon-
Motivating the composition ofAfrikaans reduplications 271 (1)
Dit het net plek-plek gereën. It has only place-place rained ‘It has only rained in some places' (i.e. distributive)
(2)
均
(3)
Hulle het een-een deur die opening geklim. They have one-one through the opening climbed ‘ They climbed one-by-one through the opening' (i.e. serial ordering)
(4)
lek-lek oor sy droë lippe. He lick-lick over his dry lips ‘ He repeatedly licks his dry lips' (i.e. iterative)
hoep幽hoep
(onomatopoeic reduplication)
hoop-hoop 'Upupaq斤icana'
(5)
(name of a common South African garden bird)
kwê-kwêvoël (onomatopoeic reduplication) kwê-kwê bird ‘ Camaroptera brachyura' (name of a South African singing bird)
From a diachronic point of view there is no consensus on whether A丘i kaans reduplication should be considered a legacy of seventeenth-century Dutch dialects spoken in the Cape (e.g. Kempen 1962), whether it should be seen as a product of creolisation and languages in contact (e.g. Raidt 1981), or whether it is of spontaneous origin (e.g. Scholtz 1963). Nevertheless , synchronically speaking reduplication in A企ikaans is a fully productive means of word formation (Bouman 1939: 346; Kempen 1969; Raidt 1981; Scholtz 1963), found in both spoken and written language , irrespective of register or genre. dences between the two component structures seem to exist in the coda, and either the onset or nucleus of the component structures (e.g. woerts-warts , tjoef- tjaJ, leer-schmeer - variation indicated in bold) , or in the case of multisyllabic words , in the coda of the first syllable as well as all other subsequent syllables, with variation in either the onset or nucleus of the first syllable (e.g. ietsie-bietsie, kêrel-schmêrel)). Considering that echo-word reduplications are motivated mainly on phonological grounds, in the sense that correspondences between the two component structures are specifically created to express a certain meaning, it falls outside the aims and scope of this contribution.
272 Gerhard B. van Huyssteen In contrast to previous descriptions of the Afrikaans reduplication construction that have been done from a more structuralist or generative perspective (e.g. the seminal book of Botha (1988) , Form and Meaning in Word Formation: A Study 014卢ikaans Reduplication , which is situated in the Galilean style of inquiry) , this study is theoretically and methodologically conducted within a cognitive grammar framework (cf. Langacker 1987; 1991a; 1999). Being part of the cognitive-functional paradigm, it is assumed in this framework that language is an integral part of human cognition (Langacker 1987: 12-13). The general strategy is to study language structure not as an autonomous system, but rather as a reflection of our general cognitive capacities: autonomous language-specific structures are postulated as a last resort, only after the possibilities of finding functional or cognitive motivations for linguistic forms are exhausted (Geeraerts 1995: 111-112; Wilcox 2001). One of the implications of this assumption is that "language structure should be characterized relying only on mental abilities and phenomena that are either well known or easily demonstrated" (Langacker 1998: 2). In his paper "A dynamic usage-based model" , Langacker (1 999) recognises five basic and very general psychological phenomena that are essential to language , viz. entrenchment, composition, comparison (and as a special case of it categorisation) , abstraction (with schematisation a special case of it), and association (with symbolisation being a special case of it) (see also Sinha 2001). In our description of a certain usage event (i.e. a contextualised language u忧erance， characterised in full phonetic and conceptual detail) , we should therefore pay attention to at least these five psychological capabilities. As Langacker (1999: 106) puts it: A usage event has many 岛cets susceptible to categorization by conventional linguistic units , which in CG [cognitive grammar] are limited to semantic , phonological , and symbolic s仕uctures. The potential categorizing units are entrenched to varying degrees and form a vast, structured invent。可 through relationships of symbolization , composition, and categorizat lO n.
Against this background, 1 will discuss the motivation for the composition of Afrikaans reduplications by giving a detailed analysis of some representative examples of grammatical and onomatopoeic reduplications. 3
3. For the purposes of this contribution , 1 will only focus on the composition of Afrikaans reduplications. For a characterisation of Afrikaans reduplications in
Motivating the composition ofAfrikaans reduplications 273
2. Composition in cognitive grammar The ability to conjoin simple structures (or units or component s衍uctures) to 岛rm structures that are more complex (or composite structures) is one of the most basic cognitive capabilities of the human mind. Research from various fields has indicated that composition is central to problem-solving (Holyoak 1995), text production (Bereiter and 8cardamalia 1987), concept formation (Franks and Bransford 1971), memory (801so and McCarthy 1981), as well as other cognitive activities (e.g. An derson 1983; Mu叩hy 1988). Pinker (1997: 564) even holds the opinion that "the mind owes its power to its syntactic, compositional, combinatorial abilities" [my emphasis - GBVH]. From a cognitive grammar perspective, a high premium is therefore also put on the importance of composition in language usage. Langacker (1999: 94) defines composition in language as the integration of two or more component structures to form a composite structure. This integration of component structures in composite structures can be expressed as ([A] [B])C, where [A] and [B] are conventionalised units that have not combined before, and (C) is the novel composite structure that is produced when the two component structures integrate. 4 For example: in a compound like dompelkursus ‘ crash course' (Basson and Van Niekerk 2002: 17), the two component structures [DOMPEL] ‘ plunge' and [KURSUS] ‘ course' are combined to form the novel composite struc阳re (DOMPELKURSUS). It is important, though , to keep in mind that (DOMPELKURSUS) is not merely the combination of [DOMPEL] and [KURSUS]. When two or more structures combine in a complex structure, an integrated composite structure is formed , which is a unit in its own right and which can not be strict1 y reduced to the sum of its component structures. A view ofp町tial compositionality is therefore held within this framework (Langacker 1997: 247-248), implying that one should not only describe the component s位uctures or composite structures of grammatical constructions, but also their mode of integration and the relations between the different structures (Goldberg 1995: 16; Langacker 1987: 277).
terms of the other four identified psychological phenomena, see Van Huyssteen (2000). 4. Square brackets are used to indicate the unit status of a structure (i.e. conventionalised structures), while parentheses are used to depict novel utterances/s往uctures.
274 Gerhard B. van Huyssteen
The potential combination of component structures as well as the probability of combination is largely determined by the intemal structure of component structures (Taylor 1990: 525). The valence potential of a certain component structure (i. e. the potential of a s位ucture to combine with another structure) is govemed by its semantic structure, while the manner how the structure combines in valence relations is somewhat a function of linguistic convention (Langacker 1991b: 167). These valence relations are dictated by four primary valence factors , namely (1) correspondence, (2) profile determinacy, (3) conceptual and phonological autonomy and dependency, and (4) constituency. With regard to corre，早 ondence， we have to identify the corresponding substructures of the different component structures. The commonalities of the substructures serve as instructions for the superimposing of the specifications of the corresponding substructures to combine in a composite structure. Such correspondences might be sought on both the phonological and the semantic poles (Langacker 1987: 277-278; 1988c). With regard to the different kinds of reduplication in Afrikaans , 1 will illustrate that coη。 spondences on the semantic andlor phonological pole are a matter of degree: depending on the extent of correspondences between two component structures, different forms of reduplication emerge. When two component structures combine, the resu 1ting composite structure usually inherits the profile of one of the component structures (e.g. in a compound like gemakstoel ‘ comfort chair' , the composite struc阳re inherits the profile of stoel ‘ chair'). One should therefore as k: given the profiles of the component structures in a valence relation , what will be the profile of the composite structure? (Langacker 1987: 288). With reference to profile determinacy , Langacker (1 987: 291-292) remarks that there is no absolute way to predict which component structure will feature as profile determinant; in fact , he says , "it makes no real difference whether we say that all of the component structures are profile determinants or that none of them are." In the case of both lexical and onomatopoeic reduplications , 1 will demonstrate that neither of the component structures can be identified as the profile determinant, as both component structures are identical. Both component structures contribute equally to the profil
Motivating the composition 01 Afrikaans reduplications 275
make intemal reference to another structure, either semantically or phonologically. A s仕ucture that is characterised in terms of another structure and consequently makes salient intemal reference to that structuI毡， is called a dependent structure. An independent structure, on the other hand, is a phonological or semantic structure that exists in its own right, without presuming another structure. Although this AID asymme位y is common in grammatical structure (as in the case of onomatopoeic reduplications) , nothing precludes a relation of mutual dependence or guarantees that there will be a dependency relation between the two structures (Langacker 1987: 300) , with grammatical reduplications an excellent case in point. The last valence factor in terms of which we should describe the composition of Afrikaans reduplications is constituency, i. e. the order or sequence in which component structures combine to form more elaborate composite structures (Langacker 1987: 310). In the case of reduplications , one would suppose that constituency is rather non-essential , as the sequence of combination is irrelevant when two identical component structures combine. However, 1 will show that the level of constituency may have an effect on the meaning of reduplications. Let us consider these statements in more detail , firstly with regard to grammatical reduplications , and then pertaining to onomatopoeic reduplicatíons.
3. The composition of Afrikaans reduplications 3. 1.
Defining the Afrikaans reduplication construction
Based on the definitions of grammatical constructions in construction grammar (Goldberg 1995) and cognitive grammar, we can define the A仕i kaans reduplication construction as: a word-formation construction where a dependentlindependent morpheme [Xl x] combines with an identical morpheme [Xl x] to form a word [X- Xlxx] , of which the meaning [X-X] is not s位ictly predictable 企om the two component structures 仅仅] and [Xl x]. In simple terms this means that in a reduplication like 甲eel叩eelin 份， the meaning of ‘ easily' is not predictable from the meaning of 再peel.
276 Gerhard B. van Huyssteen (6)
Ons het 罚peel-speel die werk afgehandel. We have play-play the work finished ‘ We have finished the work easily'
Following from this working definition we can formulate the construction schema for the reduplication construction in Afrikaans as [[[X/x][ X/x]]XX/x-x] , instantiating example (6) as [[[SPEELlspeel][SPEELlspeel]] SPEEL-SPEELlspeel-speel] , where [SPEEL-SPEEL] means ‘ easily\
3.2.
Grammatical reduplications
With reference to grammatical reduplications in Afrikaans , the combination of two completely identical component structures can be considered to be a special (i. e. non-prototypical) case of composition. What 1 mean is this: there is one hundred percent correspondence between all the substructures of the two component structures, on both the phonological and semantic pole; the two component structures are in no way conceptually or phonologically dependent on each other; and the order in which the constituents combine can not be shown. However, the peculiar thing about grammatical reduplication is that, unlike in the case of prot。可pical compositions (l ike compounds) , the profile of the composite struc阳re (i.e. the reduplicated form) diverges slightly from the profiles of the two identical component structures (i.e. the non-reduplicated forms) , despite the fact that neither of the component s阳ctures can be identified as the profile determinant. This holds , in my opinion, the key to the motivation for the existence of grammatical reduplications: a speaker employs the reduplication construction when s/he wants to express a meaning that differs from the conventional meaning of the nor卜reduplicated form. Consider the following examples of nou ‘ now' and its reduplicated form nou-nou ‘ a moment ago/in a moment': (7)
均y
(8)
均F
nou hier. He is now here ‘ He is here now' lS
sal nou-nou hier wees. He will now-now here be ‘ He will be here in a moment'
Motivating the composition ofAfrikaans reduplications 277 (9)
岛
was
nou-nou
hier.
He was now-now here ‘ He was here a moment ago' To understand the reduplicated form better, let us first reflect on the semantic structure of the component struc阳re [NOU/nou] , as in example (7). As an adverb of time nou 、ow' is conceptualised as a stative relation: it specifies a relative position (the landmark), si阳ated somewhere between PAST (trajectorl) and FUTURE (trajector2)' In Figure 1, the specification of the relative position is indicated with a thick line around the profiled stative relation. For similar cases, Langacker (1987: 222) indicates that all 臼cets of the scene are co-activated and simultaneously available , and that its conception involves summary rather than sequential scanning.
T
' NOU
Figure 1. nou ‘ at this moment; now'
Given nou as a deictic expression, it involves that [NOU] is grounded and that there must be some reference to a ground element within its scope of predication (Langacker 1987: 126). [NOU] refers to the time of speaking, and is grounded in the processing time (aηow T in Figure 1), rather than in the conceived time (arrow t in Figure 1). The subjective grounding in the processing time is indicated as TG within the predicate scope, which, due to summa叩 scanning， can also refer to PAST and FUTURE , but rather identifies PRESENT as the primary figure , i.e. the prominent substructure in the scene around which the scene is organized (Langacker 1987: 120). Prototypically nou therefore has the sense ‘ at this moment; now\
278 Gerhard B. van Huyssteen
Although not very frequen t1 y, it might happen within certain contexts that [NOU] also profiles either FUTURE or PAST. Compare the following examples (taken 仕om a popular Afrikaans concise dictionary): (1 0)
Ek moet nou Woensdag die vergadering bywoon. 1 must now Wednesday the meeting attend ‘ 1 have to attend the meeting this coming Wednesday'
(11)
均
is nou net weg. He is now just left ‘ He has left just now'
T
NOU-NOU
T
T. NOU
NOU
Figure 2. nou-nou 'in a moment' 一 see example (8)
In example (10) the semantic/pragmatic context determines that nou , as a time indicator, profiles a position somewhere in the near future through [WEDNESDAY]'s elaboration of the elaboration site of [NOU] (i.e. moments in the conceived time , indicated by the hatched parts in Figure 1). In example (11) , on the other hand, [NOU] is elaborated by [N ET] , where PAST is part of the semantic structure of [NET] , rendering the meaning ‘ recent1 y , in the past\
Motivat切g the
composition 014.如kaans reduplications 279
Let us now look at the reduplicated fonn of nou， ωin examples (8) and where nou-nou profiles a stative relation either somewhere in FUTURE (see Figure 2) or somewhere in PAST (see Figure 3).
(则，
T
NOU-NOU
NOU
NOU
Figure 3. nou-nou ‘ a moment ago' - see example (9)
What is important is 也at part of the internal conceptual structure of [NOUNOU] can by no means profile PRESENT, contrary to the case of [NOU]. Compare the following sets of examples, where the second example in every set is impossible in Afrikaans: (12) a. Is jy nou al moeg? Are you now already tired? ‘Are you already tired now?' b. * Is jy nou-nou al moeg? Are you now-now already tired?
(13) a. Het hy nou reeds probleme met sy motor? Have he now already problems with his car? ‘ Does he now already have problems with his car?'
280 Gerhard B. van Huyssteen b.
* Het hy nou-nou reeds probleme Have he now-now already problems
met sy motor? with his car?
It should be c1ear at this stage that the profile of the composite structure of grarnmatical reduplications diverges slight1y from those of the component structures: when two component structures with exact1y the same profile combine in a complex composite structure, a different part in the same scene is pro fi1 ed as a result (in other words FUTURE or PAST in Figures 2 and 3). By means of the reduplication construction, the speaker can therefore express a meaning that differs from the original meaning. This difference in meaning is motivated, in my opinion, by the metonymy MORE OF FORM FOR MORE OF CONTENT, where the MORE OF CONTENT that the reduplication achieves in this case is that of distancing away 仕om PRESENT to either FUTURE or PAST (i.e. a pro fi1 e shift in the conceived time). Let us now look at another form of pro fi1 e shift, which involves the "recategorisation" of a perfective verb as an imperfective verb. Here , the effect of the metonymy MORE OF FORM FOR MORE OF CONTENT is that of extending a perfective process in time and thereby construing it as imperfective. Compare the following example: (14)
均
lek-lek oor sy droë lippe. He lick-lick over his dry lips ‘ He is licking his dry lips repeatedly'
[LEKllek] ‘ lick' usually pro fi1 es the perfective process of ‘ passing one's tongue over to taste/moisten/clean' by means of sequential scanning. The reduplicated form lek-lek, however, only pro fi1 es an imperfective process, thus changing the punctual verb into a progressive verb. In my opinion, this sense of ‘ iterativity' or ‘ continuation' is motivated iconically by reduplication plus metonymy, which may be stated as TWO PERFECTIVE PROCESSES FOR AN INDEFINlTE NUMBER OF PROCESSES. In languages like English or Dutch, this imperfective process is indicated by means of grammatical aspect (e.g. be licking in English , or aan het likken zijn in Dutch). Afrikaans , on the other hand, uses the reduplication construction to achieve this goal (i. e. to express grammatical aspect). Likewise, the verb [BRUL/brul] ‘ roar' normally pro fi1 es a perfective process but, due to the impact of reduplication plus the above-mentioned metonymy, the perfective process is imperfectivised by (BRULBRUL/brul-brul) , as can be seen in example (15):
Motivating the composition ofAfrikaans reduplications 281
(1 5)
Die leeu brul-brul
weg.
The lion roar-roar away ‘While roaring , the lion left' The perfective process of roaring described by brul may involve one or several completed acts of roaring; the imperfective process described by brul-brul involves an indefinite number of roars. Therefore, in Figure 4 , the temporal profiles in the component structures are represented as discrete line segments, indicating the boundedness of the process. In contrast, the temporal profile in the composite structure is shown with a series of dots at each end to indicate the expansibi1 ity of the imperfective process (see also Langacker 2000: 224 for a detailed description of the differences between perfective and imperfective processes, as well as an altemative way to depict these processes in quasi-pictorial diagrams).
t
t
1m
-
..
BRL儿.-BRUL
1m
BRUL
BRUL
Figure 4. bru/-bru/ ‘ roaring' (imperfective process)
We can also apply this analysis on vat-vat in example (1 6) (based on an example by Botha 1988):
282 Gerhard B. van Huyssteen (1 6)
Die dokter vat-vat aan die swelsel. The doctor touches-touches on the swelling ‘ The doctor touches the swelling repeatedly/again and again'
The verb [VAT/vat] ‘ touch' usually profiles a transitive process that is inherently unmarked for aspectual specification (i.e. it can profile either a perfective or an imperfective process). In the case of the reduplication [VAT-VAT/vat-vat] , we once again see an imperfectivising shift based on the metonymy TWO PROCESSES FOR AN INDEFINITE NUMBER OF PROCESSES, as [VAT-VAT] can only profile an imperfective process (without the possibili可 to profile a perfective process). The sense of ‘ repetition' can be 邸， cribed to the fact that the use of the progressive aspect turns a punctual verb into an activity, thereby giving [VAT-VAT/vat-vat] a ‘ repetitive' sense (which is also the communicative function ofthis reduplication). Additional to this ‘ repetitive' sense, Botha (1988: 94) correctly suggests that a sense of ‘ attenuation' could also be ascribed to example (16) , rendering the meaning ‘ The doctor touches the swelling hesitantly/tentatively/non-intensely' (岛r an elaborate discussion of this meaning extension, see Van Huyssteen 2000). Whichever sense is activated, will obviously be dictated by the semantic/pragmatic context. However, the level of constituency can have a narrowing effect on which meaning is activated. Consider the following examples: (1 7)
Die dokter het aan die swelsel gevat-vat. The doctor has on the swelling touched-touch ‘ The doctor touched the swelling hesitant1y/repeatedly'
(1 8)
Die dokter het aan die swelsel gevat-gevat. The doctor has on the swelling touched-touched ‘ The doctor touched the swelling repeatedly'
In example (17) , in which gevat-vat is the past tense form of [VATVAT/vat-vat] , either the sense ‘ repetition' or the sense ‘ attenuation' could be ascribed to the reduplication, depending on the usage context. This is not unexpected, as the p臼t tense form of the reduplication is sanctioned by the highly entrenched and productive schema for past tense formation in A仕1kaans , viz. [PST+Xlge+x] , thus motivating [PST+VAT-VAT/ge+vat-vat] , with both the senses of the infinitive form available for activation. [VAT/vat] and [VAT/vat] are combined on the lowest level of constituenc)
Motivating the composition 01Afrikaans red叩lications 283
(see Figure 匀， so that "word-status" is assigned to the composite structure , which is therefore then accessible for 且lfther morphological elaboration (i.e. by inflectional, derivational or syntactic structures).
vat
vat
Figure 5. Constituency of gevat-vat ‘ touched hesitantly/repeatedly'
In example (18) , on the other hand , gevat-gevat only has the meaning ‘ repeatedly' , and not 吁lesitantly'. This semantic narrowing can be ascribed to the level of constituency: [VAT/vat] and [PST/ge+] combine firstly , to form the past tense verb [PST+VAT/ge+vat] , which profiles a perfective process th创 has taken place somewhere in the past (i.e. a completed perfective process). When [PST+VAT/ge+vat] and [PST+VAT/ge+vat] combine on the next level of constituency (see Figure 份， the metonymy TWO PROCESSES FOR AN INDEFINITE NUMBER OF PROCESSES motivates the interpretation of only ‘ repeatedly' , since each of the processes have already been completed (i.e. TWO COMPLETED PROCESSES FOR AN INDEFINITE NUMBER OF COMPLETED PROCESSES). We can therefore establish that when a speaker unambiguously wants to express repetition, the constituency is motivated by the TWO PROCESSES FOR AN INDEFINITE NUMBER OF PROCESSES metonymy, and sanctioned by the schema ([[PST+X][PST+X]] PST+X-PST+X) , rather than by the ([[PST][X-X]]PST+X-X) schema, which can lead to ambiguous interpretations.
284 Gerhard B. van Huyssteen
ge+vat-ge+vat
/\
ae+vat
ae+vat
/\/\
ge+
F聊re
--+ vat
ge+
一...
vat
6. Constituency of gevat-gevat ‘ touched repeatedly'
Let us look at one last example where reduplication plus metonymy leads to some form of "recategorisation". Cases of nominalisation via metonymy are found when two identical numerals combine in the reduplication construction. Compare the following examples, taken from Botha (1988: 36): (19)
Drie-drie storm deur die hek. Three-three storm through the gate ‘ Groups ofthree storm through the gate'
(20)
Tien-tien verlaat die kamer. Ten-ten leave the room ‘ Groups often leave the room'
Our conceptualisation of numerals like [DRIE] and [TIEN] entails that they represent a highly specified atemporal. relation where the landmark is a specific point on the counting scale, and the trajector is either a continuous or a discrete entity. This entity also represents the elaboration site of the concept, which must be elaborated by some other structure. However, in the case of mutual elaboration where two identical component s位uc阳res combine, the entity (i怠 the trajector in this case) is extended by means of metonymy to an indefinite number of groups of three. In conclusion: it should be clear 企om the above analyses that the profile shift that occurs in grammatical reduplications can be ascribed to the effects of metonymisation, i.e. MORE OF FORM FOR MORE OF CONTENT. This view
Motivating the composition ofAfri归ans red.ψ/ications 285
of profile shift is fully compatible with Moravcsik's view that there are subtle, though definite, differences between the meanings of reduplicated and non-reduplicated forms: [...] 1 have found no clear example of an unreduplicated constructions' meaning properly including the meaning of the corresponding reduplicated one, or ofthe two overlapping [...] (Moravcsik 1978: 316) Langacker accounts for this kind of phenomenon by stating that "there is no c1 aim to the effect that the composite structure is fully compositional or derives in any mechanical way 企om the semantic specifications of its components" (Langacker 1987: 292). Instead, "it is more appropriate to say that the component structures motivate aspects of the composite structure, and that the degree ofmotivation is variable" (Langacker 1987: 292).
3.3.
Onomatopoeic reduplications
Onomatopoeic reduplications , just like grammatical reduplications , can also be considered a special case of composition, in the sense that there are also one-to-one corre.甲ondences between the component structures on the phonological pole (and therefore by implication also on the semantic pole) , and that the order in which the constituents combine can also not be shown. As is the case with grammatical reduplications , the profile of the composite structure differs from the profiles of the two identical component s位uc阳res， despite the fact that neither of the component structures can be identified as the profile determinant. However, unlike grammatical reduplications , the component structures in onomatopoeic reduplications show a high degree of conceptual and phonological dependence , in the sense that onomatopoeic interjections can be considered highly dependent mo甲hemes. The c1 aim here is that the motivation for the existence of onomatopoeic reduplications lies in the fact that speakers use the reduplication construction to refer metonymically to birds (e.g. hoep-hoep ‘ Upupa a仕icana')， mammals (e.g. dikdik ‘Rhynchotragus damarensis') , fish (e.g. gor-gor ‘ Pomadasys benne旷)， or plants (e.g. kêr-kêr ‘ Erica peltata') in terms of some onomatopoeic resemblance with the sound-producing referen t. To elucidate 也is c1 aim, let us firstly look into conceptual and phonological autonomy and dependence in more detai l. Two factors mainly determine the measure of dependency between two structures when they
286 Gerhard B. van Huyssteen
combine in a valence relation, namely salience and elaboration. Consider the following definition ofdependency by Langacker (1987: 300): One structure, D , is dependent on the other, A , to the extent that A constitutes an elaboration of a salient substructure within D.
It is therefore important for us to determine which substructure of [D] is salient, and whether this salient substructure is elaborated by [A]. The salience (i.e. the degree of prominence) of a substructure can be based on two factors , namely the degree 01 centraUψand the probability 01 activation. The centrality of a substructure is described in terms of the extent to which a specification is conventional, generic, intrinsic, and characteristic (Langacker 1987: 150). A substructure that is more probably activated with a given structure, also enjoys greater prominence, and must therefore be identified. Other factors that could determine the salience of a substructure (e.g. p町t vs. whole, inanimate vs. animate, etc.) should of course also be taken into account. When the salient structure/substructure has been identified, we should also determine whether this s位uc阳re is elaborated by another structure. Elaboration can be seen as the instantiation of a schema, where a special case of the schema is rendered more precisely and in more detail (Langacker 1987: 68). Prototypically, when two components [D] and [A] enter into a valence relation , a substructure of [D] necessarily corresponds with the profile of [A]. This substructure is called the elaboration site (or e-site) , because of its relation of schematicity with [A] and the fact that it is elaborated by [A]. In short, we must determine which schematic, dependent substruc阳re (the e-site) is elaborated by a more specified, autonomous structure. Let us consider some onomatopoeic reduplications in terms of these factors. In examples like hoep-hoep 'Upupa a仕icana' ， tinktinkie ‘ Cisticola rufilata' , and kwê-kwêvoël ‘ Camaroptera brachyura', the elements HOEP, TINKandKW直 can be seen as onomatopoeic interjections (i.e. the sounds of birds). AID asymmetry should therefore be an important valence factor in the combination of these elements in more complex structures, due to the high degree of dependence on the phonological pole. Such dependency becomes c1ear 仕om a unipolar analysis of these elements as component structures of the conventionalised expressions hoep-hoep , tinktinkie, and kwê-kwêvoël. Following the fact that these component structures only occur in peripheral instances in other expressions than the conventionalised ex-
Motivating the composit.归nq问fri，阳ans reduplications 287
pressions, we can conclude 也at these component structures are prototypically mutually dependent.
t
00:
•
hoep-hoep
/\
就; '
141.....
.. . . . ..... . . ..... .... .
h. ……· o. ……. . . e 嘈…….
p 嘈……·
就; • • ... h
-
• • • • .... 0
… .e • • • .... p
Figure 7. hoep-hoep ('hoop-hoop; Upupa a副cana')
If we then say that [HOEP/hoep] and [HOEP/hoep] are mutually dependent, it implies 由at both contain a salient substructure that is reciprocally elaborated by 也e other component structure. As an onomatopoeic interjection (e.g. when used to imitate the sound ofthe bird), [HOEP/hoep] is represented as a profiled moment of a perfective process, where the 位ajector is the producer of 也e sound (see Figure 7). If we measure the salience of this substruc阳re (i.e. the tr;苟 ector/由e producer/the bird) in terms of centrality and probability of activation, we can conclude with certainty that it has a high degree of salience: 一
Degree
01 centrality: In terms of the DEFINING PROPERTY FOR CATEGORY metonymy (Kövecses and Radden 1998: 56-57), the relation
between 也e
sound and the bird 伽.t produces the sound is strongly conventionalised; it is generic to the extent that speakers know that it refers
288 Gerhard B. van Huyssteen
一
to a kind of bird, without necessarily having any specialised knowledge about the species; it is intrinsic to the relati'On, in the sense that it is unlikely that another entity but this specific species will produce this sound; and it is characteristic of the produced sound, in the sense that there is a strong mutual association between the bird and the sound it produces. Probability 01 activation: In the meton严nic relation between the bird and the sound it produces, the PRODUCT is more salient than the PRODUCER is , i.e. we use the sound that the bird produces (the PRODUCT) to refer to the bird (the PRODUCER) , and not vice versa; thus rendering the metonymy PRODUCT FOR PRODUCER (SOUND FOR PRODUCER OF SOUND). Although Kövecses and Radden (1998: 5ι57) claim that the PRODUCT FOR PRODUCER metonymy does not seem to exist (based on the HUMAN OVER NON-HUMAN and ANIMATE OVER INANlMATE cognitive principles) , onomatopoeic reduplications are in my opinion clear examples to justify the existence ofthis metonymy.
Due t'O the strong metonymic association between the PRODUCER (the bird) and the PRODUCT (the sound) , the trajector of [HOEP] (i息 the thing in the process) will function as the e-site of the structure. In terms of our definition of a dependent structure, one therefore expects that [HOEP/hoep] should be elaborated by an autonomous structure. However, in the case of onomatopoeic reduplications , where two identical , dependent component structures combine, the e-site of the one component serves as elaboration of the e-site of the other component, and vice versa. We can theref'Ore speak of mutual elaboration when two identical dependent structures c'Ombine in a valence relation. The mutual identity is indicated in Figure 7 by means of the double-headed line arrow, while the double-headed elaboration arrow indicates that the two substructures elaborate each other mutually. To summarise: the distinction between autonomous and dependent elements is important for our description of the reduplication construction. It was illustrated that, in the case of dependent structures, the dependent substructures in the dependent components elaborate each other mutually , resulting in an autonomous composite structure. 1 have also shown that onomatopoeic reduplications are motivated by various metonymic relations , like DEFINING PROPERTY FOR CATEGORY, and PRODUCT FOR PRODUCER. With reference to Kövecses and Radden, we can then conclude that these metonymies "provide speakers with natural ‘ cognitive links' that enable them to move 仕om one entity ... to another ... without any effort or even
Motivating the composition ofAfrikaans reduplications 289
unconsciously. They are a part ofthe mutual knowledge that speakers share and r e1 y on in creating and understanding (reduplications) with ease" (Kövecses and Radden 1998: 61).
4. Summary and conclusion At the outset of this contribution, 1 aimed at an explanation of the motivation for the formation of Afrikaans reduplications by giving a detailed characterisation of composition in the reduplication construction, with special reference to the relationship between metonymy and grammatical constructions. It was illustrated that grammatical reduplications are motivated by the fact that the reduplicated 岛rm is used to express a meaning that differs from the conventional meaning of the non-reduplicated form , as well as in the absence of another means to convey this meaning (e.g. in A仕ikaans， grammatical aspect is often expressed by grammatical reduplications). The existence of onomatopoeic reduplications is motivated by the fact that speakers use the reduplication cons位uction to refer metonymically to referents , in terms of some onomatopoeic resemblance with the referent. 1 have indicated that we can adequately describe the formation of reduplications in terms of four valence factors , viz. correspondence, profile determinacy, autonomy/dependency-asymmetry, and constituency. On the grounds of these valence factors , we can conclude that grammatical reduplications can be seen as quite an uncharacteristic case of composition, while onomatopoeic reduplications are more prototypical , and echo-word reduplications even more so (see footnote 1). 1 have also illustrated that reduplications in Afrikaans are motivated largely by metonymies like more of form for more of content, whole thing for a part of the thing , two processes for an indefinite number of processes , defining proper句， for category, and product for producer. 1 hope to have given an indication of the relation of metonymy to grammar, in that 1 have shown how the grammatical properties of words and constructions are changed when they are used metonymically. 1 believe that such an analysis gives us a psychologically plausible (Langacker 1998: 2) and natural (Langacker 1987: 13-14) description and explanation of the motivation for Aι rikaans reduplications.
290 Gerhard B. van Huyssteen References
Anderson, John R. 1983 The Architecture of Cognition. Cambridge: Harvard University Press. Basson, Joey and Frans Van Niekerk (eds.) 2002 Wilde Woordeboek 2. [Wild Dictionary 2]. Pretoria: LAPA Uitgewers. Bereiter, Carl and Marlene Scardamalia 1987 Th e Psychology ofWritten Composition. Hillsdale: Erlbaum. Botha, Rudi P. 1988 Form and Meaning in Word Formation: A Stu咿 of Afrikaans Reduplication. Cambridge: Cambridge University Press. Bouman, A.C. 1939 Over reduplicatie en de woordsoorten. [About reduplication and part-of-speech]. De Nieuwe Taalgids 33: 337-353. Franks , J.J. and J.D. Bransford 1971 Abstraction of visual 阴阳ms. Journal of Experimental Psychology90: 65一74. Geeraerts, Dirk 1995 Cognitive Linguistics. In: Jef Verschueren, Jan- Ola Östman and Jan Blommaert (eds.) , Handbook of Pragmatics: Manual, 111一 116. Amsterdam: John Benjamins. Goldberg, Adele E. 1995 Constructions. Chicago: University ofChicago Press. Holyoak, Keith J. 1995 Problem solving. In: Daniel N. Osherson (gen. ed.) , An lnvitation to Cognitive Sc ience, Volume 3: Thinking , 267一296. Cambridge, Mass.: MIT Press. Keller, Rudi 1998 A Theory ofLinguistic Signs. Oxford: Oxford University Press. Kempen, Willem 1962 胁。rdvorming en 卢nksiewisseling in 拼ikaans. [Word Formation and Class-Changing in Afrikaans]. K础pstad: Nasionale Boekhandel. 1969 Samestelling, afleiding en woordsoortelike mee功nksionaliteit in Afrikaans. [Compounding, Derivation and Part-of-年eech MultiFunctionality in Afrikaans]. Cape Town: Nasou. Kimenyi, Alexandre 1989 Reduplication and binomial expressions in English: A case of syntagmatic iconicity. In: Irmengard Rauch and Gerald F. Carr
Motivating the composition ofAfrikaans reduplications 291
(eds.) , The Semiotic Bridge: Trends from Ca/.协rnia， 347-365. Berlin: Mouton de Gruyter. Kövecses , Zoltán and Günter Radden Meton归ny: Developing a cognitive linguistic view. Cognitive 1998 Linguistics 9: 37-77. Lakoff, George and Mark Johnson 1980 Metaphors We Live by. Chicago: University ofChicago Press. Lakoff, George and Mark Turner 1989 More than Coo/ Reason. Chicago: University of Chicago Press. Langacker, Ronald W. 1987 Foundations ofCognitive Grammar, Volume 1: Theoretica/ Prerequisites. Stanford: Stanford University Press. 1988 The nature of grammatical valence. In: Brygida Rudzka-Ostyn (ed.) , Topics in Cognitive Linguistics, 91-125. Amsterdam: John Benjamins. 1991 a Foundations ofCognitive Grammar, Volume 11: Descriptive Application. Stanford: Stanford University Press. 1991 b Concept, Image, and 冉Imbo/: The Cognitive Basis of Grammar. Berlin: Mouton de Gruyter. 1997 The contextual basis of cognitive semantics. In: Jan Nuyts and Eric Pederson (eds.) , Language and Conceptualisation, 229-252. Cambridge: Cambridge University Press. 1998 Conceptualization, symbolization and grammar. In: Michael Tomasello (ed.) , The New Psych%gy of Language: Cognitive and Functiona/ Approaches ω Language Structure, 1-39. Mahwah: Lawrence Erlbaum. 1999 Grammar and Conceptualization. Berlin: Mouton de Gruyter. Moravcsik, Edith 1978 Reduplicative constructions. In: Joseph H. Greenberg (ed.). Universa/s of Human Language, Volume 3: Word Structure, 297334. Stanford: Stanford University Press. Murphy, Gregory L. 1988 Comprehending complex concepts. Cognitive Sc ience 12: 529一 562. Pinker, Steven 1997 How the Mind Works. London: Penguin. Raidt, Edith H. 1981 Oor die herkoms van die A企ikaanse reduplikasie. [On the Origin ofthe Afrikaans Reduplication]. In: Ampie J. Coetzee (ed.) , Hu/se/s van krista/. [Enve/opes of 的Ista巧， 178-189. Cape Town: Tafelberg.
292 Gerhard B. van Huyssteen Scholtz, Johannes du P. 1963 Taalhistoriese opstelle: Voorstudies tot ‘ n geskiedenis van Afrikaans. [Essays on the History of Language: Preliminary Studies towards a History ofAfrikaans]. Pretoria: Van Schaik. Sinha, Chris Re: Chomsky. Contribution on Cogling ([email protected]). 5 2001 November at 22:13. E-mail: [email protected]. Solso, R.L. and J. E. McCarthy 1981 Prototype formation of faces: A case of pseudomemory. British Journal ofPsychology 72: 499-503. Tai , James H. Y. 1993 Iconicity: Motivations in Chinese grammar. In: Mushira Eid and Gregory Iverson (eds.) , Princ伊les and Prediction: The Analysis of Natural Language: Papers in Honor of Gerald Sanders , 155166. Amsterdam: Benjamins. Taylor, John R. 1990 Schemas, prototypes and models: In search of the unity of the sign. In: Savas L. Tsohatzidis (ed.) , Meanings and Prototypes , 521-534. London: Routledge. Van Huyssteen, Gerhard B. 2000 Die reduplikasiekonstruksie in Afrikaans: enkele aspekte van 、n kognitiewe gebruiksgebaseerde beskrywingsmodel vir Afrikaans. [The Reduplication Construction in A.卢ikaans: Some Aspects of a Cognitive Usage-Based Model for A.斤ikaans]. Unpublished Ph.D. thesis. Potchefstroom: PUCHE. Wilcox , Sherman 2001 Re: Chomsky. Contribution on Cogling ([email protected]). 2 November at 03:4 1. E-mail: [email protected].
Metonymic motivation in anaphoric reference Fr，α ncisco
José Ruiz de Mendoza Ibáñez and Olg，α Isabel Díez Velasco
Abstract 1 This paper addresses the issue of metonymic anaphora, i.e. the selection of anaphoric pronoun whenever the antecedent is metonymic. Although previous analyses have attempted to account for the principles that govem metonymic anaphora and for the existence of purported irregularities, a comprehensive and systematic solution to the problem has not yet been provided. Our starting point has been the basic assumption that the nature ofthe underlying mapping in a metonymic antecedent has an important role in determining the selection and behaviour of the adequate anaphoric dev Ïc e. With this in mind, we have provided a definition of metonymy which takes into account the nature of the inclusion relationship between source and target and have accounted for the cognitive principles which determine this relationship. We have 岛rther distin伊ished between single and double metonymic mappings. On the basis of this discussion, we have put forward a general constraint on metonymic anaphora and three principles , the Domain Availability Principle , the Domain Precedence Principle , and the Domain Combinability Princ伊le. The interaction and/or tension between these principles, which are graded in terms of their relative strength, govems the selection of the relevant anaphoric pronoun. Finally, we have distinguished between reference to a metonymic noun and implicative refer-
1. Financial support for this research has been provided by the DGI , Ministry of
Science and Technology, Spain, grant nO BFF2000-0934. We would like to th缸tk Klaus-Uwe Panther and Linda Thomburg for their valuable comments on a preliminary version of this paper.
294 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco
ence. This distinction has proved useful in dealing with some apparently problematic examples. Keywords: anaphor, antecedent, deferred indexical reference, domain ex-
pansion, domain highlighting, implicative reference, matrix domain, meton严ny， predicate transfer, source domain, target domain
1. Introduction The advent of cognitive linguistics in the ear1y 80s brought with it a new vision of language, which has had important consequences for the understanding of the role of semantics within linguistic the。可. Thus , semantic descriptions became sensitive for the first time to such issues as the encyc10paedic conception of meaning, basic-level categorization, and prototypicality effects. In 由is context it was argued that there is no c1 ear boundary between linguistic and extralinguistic knowledge and that semantic meaning is organized in terms of concep阳al structures (cf. Fi11more's 1985 frames; Fauconnier's 1985 mental spaces; Lakoffs 1987 idealized cognitive models; Langacker's 1987 domains). There was also a m碍。r change of perspective in the conception of metaphor and metonymy. These two phenomena were no longer seen as deviant uses of language but rather as two basic cognitive processes employed in reasoning and talking about our experiences of the wor1 d. In Lakoff and Johnson's (1980) seminal work, metaphor and metonymy were described as conceptual mappings or sets of coπespondences between different domains (in metaphor) or between different elements ofthe same domain (in metonymy). In consequence ofthis view, from the beginning cognitive linguists have invested a great deal of effort in studying the role of metaphor in cognition, as is evidenced by the large amount of literature on the topic (e.g. Lakoff and Tumer 1989; Kövecses 1990, 2000; Lakoff and Johnson 1980, 1999, among others; cf. also Taylor 1995). Initially metonymy received less attention, and was primarlly studied in connection to metaphor (cf. Goossens 1990; Dirven 1993). However, in more recent years there has been an upsurge of interest in this phenomenon. Thus, apart from analyzing the relationships between metaphor and metonymy (Radden 2000; Barcelona 2000; Dirven and Pörings 2002), research has also been concemed with setting up definitional and typological criteria (Croft 1993; Langacker 1993; Kövecses and Radden 1998; Radden and Kövecses 1999), and with
Metonymic motivation in anaphoric r，价rence 295
the study of the role of metonymy in inferencing (Gibbs 1994; Thornburg and Panther 1997; Panther and Thornburg 1998, 1999; Ruiz de Mendoza 1999; Pérez and Ruiz de Mendoza 2002). However, all these studies are only concerned with metonymy on a conceptual level, while the role of metonymy as a mechanism which lies at the basis of a number of grammatical phenomena has not been consistently addressed in the literature. In fact , only a few such phenomena have been investigated in some detai l. Thus , Ruiz de Mendoza and Pérez (2001) have discussed the way metonymy motivates and constrains various forms of categorial and subcategorial conversion, and modality shifts. 2 The strong link between metonymy and morphology has been pointed out by Panther and Thornburg (2002) , who have obse凹ed th创 the formation of -er nominals is ruled by metonymic and metaphoric principles, while Waltereit (1 999a) has studied the metonymic motivation of the semantic change undergone by some French re f1 exive verbs. Other areas of inquiry have only received passing attention. A case in point is metonymic anaphora, Í. e. the selection of anaphoric pronoun when the antecedent is metonymic. Some cognitive linguists have observed a number of regularities in this phenomenon and have offered preliminary explanations, but so far there has been no 硝empt to deal systematically with its intricacies. It will be our aim in this paper to make up for this lack of systematicity and determine what principles underlie the observed regul町ities.
In order to achieve our aim, a consistent definition of metonymy is needed. However, available accounts of metonymy only make use of working definitions. We shall attempt to overcome this problem by constructing a more refined definition of metonymy which takes into account relevant aspects of previous descriptions. The resulting account will provide us with the necessary tools to carry out a detailed analysis of anaphoric reference in the context of metonymic relations. In so doing, it will be observed that there is a ve叩 close connection between the principles underlying the choice of anaphoric referent and the nature of the metonymic pa忧ern in the antecedent. Ultimately, we are dealing with another grammatical phenomenon whose behaviour is motivated by metonymic actI vlty. 2. Other studies on the relationship that exists between meton严ny and modality have been carried out by Traugott and König (1 991), Bybe et al. (1994) , Heine (1995) , and Goossens (1999).
296 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco 2. What is a metonymy? When dealing with the notion of metonymy, its primarily referential function has often been highlighted (cf. Lakoffand Johnson 1980: 36; Waltereit 1999b: 234). However, this is not one of its defining features , but rather a fairly common characteristic which , as will be seen below , results from other aspects of its nature. First, we note that metonymy can be both referential and predicative , as illustrated by the following examples:
(1)
We need one more brain in our team.
(2)
He 's a real brain.
In (1) , the notion of ‘ brain' is used metonymically to refer to a person that has excellent intellectual abilities. In contrast, in (2) the same mapping has a predicative function. Interestingly enough , as pointed out in a number of recent studies (see Ruiz de Mendoza 2000; Ruiz de Mendoza and Pérez 2001; and Ruiz de Mendoza and Díez 2002), it is also possible to find cases ofreferential metaphor as illustrated in (3): (3)
The old cow wants to fire me.
In this expression , old cow serves to identify a person in a way similar to the way metonymy does. The comparison between these examples suggests that meton严nies and some metaphors such as PEOPLE ARE ANIMALS have some characteristics in common which allow them to be employed both predicatively and referentially.3 Compare now the following examples: (4)
John is a brilliant brain.
(5)
John is a stin炒 rat.
3. It must be noted that, although possible, instances of referential metaphor and of predicative metonymy are rather unusual.
Metonymic motivation in anaphoric reference 297
jentence (4) contains a predicative metonymy while (5) is a predicative netaphor. 4 80 far, we have observed that both coincide in selecting a sali~nt feature ofthe SOUTce domain ('intelligence' and ‘ unfaithfulness '), which s then applied to the target domain; however, they differ in the kind of :ognitive mechanism needed for their successful interpretation. Thus , in (5) t highlighting operation takes place whereby a salient feature typically tttributed to rats (i. e. their unfaithful behaviour) is given prominence and nade to correspond with an a时ibute of John's behavioUT. As a result, John s described as a person that cannot be trusted. 8imilarly, in (4) there is also a. highlighting process by means of which the intellectual capacity connected to the brain is singled out and given primary status; however, this operation must be followed by a domain expansion process in which ‘ brain' is mapped onto the broader notion of ‘ person who has great intellectual abilities'. Accordingly, a predicative metonymy involves two different processes, domain highlighting and domain expansion , while predicative metaphors only involve domain highlighting. However, this processing difference between the two phenomena is but the consequence of the different kinds of domain relationship that exist in metaphor and metonymy. We noted in the introduction that in metaphor two separate domains are involved whereas in me阳nymy there is only one domain. Here, we further observe that domain expansion, as in (份， is but the natural consequence of the source domain (i.e. ‘ brain') being a subdomain of the target domain (i.e. ‘ person'). This lends further support to our claim that the distinction between metaphor and metonymy is not based on its referential potential but merely on the domain-intemal/domain-extemal nature of the mapping. Domain expansion occurs whenever the source domain is a subdomain of the target. But it may also be possible for the target to be a subdomain of the source (see figures 1 and 2). In this case , a domain reduction operation takes place, as in Denmark won the match , where by Denmark we metonymically refer to the team that represents this country in a sports event. These two functions of metonymy correspond to the two basic types of metonymic mapping identified by Ruiz de Mendoza (2000): one in which the source is a subdomain of the target and another in which the target is a 4. The metonymic nature of (4) and the metaphoric nature of (5) are best seen from the following paraphrases: a brilliant brain is ‘ a person who has a brilliant mind' and a stin炒 rat is ‘ a person who behaves like a stinky rat is thought to behave'.
298 Francisco José Ruiz de Mendoza Ibáñez and Olglα Isabel Díez Velasco
subdomain of the source. The first type gives rise to source-in-target metonymies; in them access to a whole domain , called the matrix domain , is provided by one of its subdomains in a process of domain expansion (e.g. the sax in The sax has the flu maps onto ‘ the person who plays the sax'). The second type, in tum, gives rise to target-in-source metonymies. This metonymic type involves domain reduction and is frequently employed to deal with concepts which are difficult to pin down with accuracy or whose exact nature is not clearly delineated. Consider, in this respect, the following sentence: (6)
Chrysler has laid offa hundred workers.
Here , Chrysler is used to refer to the person or group of people in the car company who have the power to lay off workers , but whose exact identity it is unnecessa叩 to determine. An additional consequence of domain reduction is the highlighting of a relevant part of the domain. Thus , Chrysler brings into focus one relevant subdomain, i.e. the committee board, and gives it primary status, thereby providing the hearer with the implication that it is the whole board rather than just one of its components that is responsible for the action.
TARGET
Figure 1. Source-in-target metonymy
Furthermore, we suggest that the fact that the two basic functions of metonymy correspond to these tw。可pes of metonymic mapping argues against the existence ofpart-for-part metonymies. In this connection, it may be observed that examples ofwhat have been traditionally regarded as partfor-part metonymies can be altematively explained as instances of targetin-source metonymy. Consider, for example, the metonymy RULER FOR
Metonymic motivation in anaphoric r，φrence 299 ARMY , as in Napoleon lost at Waterloo. While it is 衍ue that a ruler and the anny under his cornmand may be regarded as different elements of a common conceptual 仕创ne (a war or a battle), it is also possible to conceive of Napoleon's army as part of our knowledge about the French general. That is to say, Napoleon and his army may be said to stand in a domainsubdomain relationship. Similarly, it is 位ue that press reporters and magazines (or newspapers) seem to be different elements within the domain of joumalism. However, in the interpretation of the meton严ny in Th e Times has not arrived yet (COMPANY FOR EMPLOYEES) , we can na阳rally think of the press reporters as part (and therefore a subdomain) ofthe company they work for. The consequences of disregarding part-for-part metonymies have been fully explored in Ruiz de Mendoza (1997 , 2000) and Ruiz de Mendoza and Pérez (2001). In the following section we shall enquire into the relevance of the source-in-targetltarget-in-source distinction to account for the principles which govem metonymic anaphora.
Figure 2. Target-in-source metonymy
From our discussion so far it follows that metonymy is a concep阳al mapping within a single domain where the source may be a subdomain of the target or the target a subdomain of the source; in the former case the mapping involves domain expansion, while in the latter it involves domain reduction and the consequent highlighting of a relevant part of the matrix domain. Only the former metonymic type may be used predicatively. Before proceeding one further remark should be made regarding the nature of the ma位ix domain of a metonymic mapping , and the conceptual criteria which allow us to distinguish it within a metonymy. By definition , the ma创x domain of a metonymy is the most encompassing of the two
300 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco
domains involved in a mapping. As a result, it usually acts as a domain of reference. For this reason , it must be a very well-defined domain whose exact nature can be easily identified and accessed; that is to say, it must be a very clearly delineated concept. In contrast to what is the case with the matrix domain, which cannot be vague , conceptual subdomains are not always easy to access an d/ or determine with precision. This poses no especial problem since any subdomain is to be understood in relation to the matrix domain whose precise na阳re is easily identifiable. Consider again example (哟， where Chrysler metonymically stands for management. Whereas 'Chrysler' as a company is a well-defined concept, the target domain is no t: we cannot determine with certainty the precise nature of the people involved (e.g. it can be the president in person, his secretary, a committee, etc.) , which naturally prevents this domain from being the matrix domain. In fact , it is the imprecise nature of the subdomain which motivates the use ofthe matrix domain to refer to it. It may be the case that the two domains involved fulfil the basic requirements to qualify as the matrix domain of a metonymy. In situations like these , the domain-subdomain relationship is probably a function of other general principles which govern cognitive saliency, and thus regulate the preference of a domain to be the matrix domain. Some of these principles , which have already been identified by Langacker (1 993) and Kδvec ses and Radden (1 998), are the following: human over non-human; container over content; controlling entity over controlled entity; whole over part. For example, in INSTRUMENT FOR PLAYER, two of the abovementioned principles, i.e. human over non-human and controlling over controlled entity, coincide in the selection of ‘ player' as the matrix domain; in SHOE FOR SHOELACES it is the whole-over-part principle that applies; and in VEHICLE FOR DRIVER it is again both human over non-human and controlling over controlled entity which govern the selection of ‘ driver' as the matrix domain. In general terms , there exists a preference for the controlling entity to be the matrix domain as long as this domain is sufficiently well-delineated to be eligible as such.
3. Anaphoric reference The principles that govern the choice of anaphoric pronouns whenever the antecedent is metonymic remain one of the least explored areas of enquiry in cognitive linguistics. Fauconnier (1985) and Panther and Radden (1 999)
Metonymic motivation in anaphoric r，价'rence 301 '-ve identified some of the crucial questions related to this issue but they '-ve provided only ve可 partial answers. Nunberg (1 995) and Stirling
(1996) are perhaps the most detailed a忧empts to disclose the nature of such principles. In order to account for the differences that exist in the selection of the anaphoric pronoun where the antecedent is metonymic , Nunberg (1995) has proposed the existence of two types of linguistic mechanisms: tleferred indexical r，价rence and predicate transfer. The former is defined s 也e process whereby an indexical is used to refer to an object that coηe 笔JOnds somehow to the contextual element chosen by a demonstrative , whereas the latter occurs whenever the name of a property that applies to something in one domain is used to refer to the name of a property that 哩)J>lies to things in another domain (Nunberg 1995: 111). Consider the fòUowing examples from Nunberg (1995): (η
This
(8)
1 am parked out back.
is parked out back.
Imagine (7) is uttered in a situation in which the speaker is holding a key in his hand. According to Nunberg (1995) , this sentence is an example of deferred indexical reference in which the demonstrative this makes reference to a car. In contrast, (8) is a case of predicate transfer where a property typically attributed to cars is applied to a person. Besides, (8) fulfils the two conditions that Nunberg (1995) postulates for predicate transfer to occur: (a) there must exist some kind of correspondence between the property denoted by the derived predicate and the one denoted by the original predicate; and (b) the property added by the new predicate must be noteworthy. It may be interesting to note that there seems to be a correlation between our two basic metonymic types and the phenomena of deferred indexical reference and predicate transfer. Thus , (7) is a source-in-target metonymy, where ‘ key' stands for ‘ car' , while (8) is a target-in-source metonymy in which by means of the matrix domain ‘ owner/driver' (identified with the speaker) we metonymically refer to one of its subdomains (i.e. ‘ car'). The domain relationship can be be忧er observed in figures 3 and 4 below. Nunberg (1995) further claims that the distinction between deferred indexical reference and predicate transfer is useful in order to adequately account for anaphoric reference in cases in which the anaphoric device is found in a conjoined predicate. In this connection, he argues that in de':'
302 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco
Figure 3.
KEY FOR CAR
metonymy
OWNERlDRIVER
Source
Figure 4.
OWNERlDRIVER FOR CAR
metonymy.
ferred indexical reference , a conjoined predicate must be semantically connected to the deferred referent, as is the case with ‘ the car' in (例， whereas in predicate transfer the conjoined predicate must express a proper句， ofthe element that receives the prope吗" i.e. ‘ the driver/owner in (1 0). If we consider the metonymies involved in these sentences, it may be noted that deferred indexical reference allows the conjoined predicate to describe the target domain and predicate transfer licenses co-reference to the source domain: (9)
a. This is parked out back and may not start. b. ??This only fits the leftfront door and is parked out back.
(1 0)
a. 1 am parked out back and have been waitingfor 15 minutes. b. *1 am parked out back and may not start.
Metonymic motivation in anaphoric r，价'rence 303 Nunberg's explanation would seem to account adequately 岛r anaphoric reference in cases of predicate transfer; however, it is striking to find that a 民ntence such as (11) is not considered an example of predicate transfer: (1 1)
Th e ham sandwich is waitingfor his check.
Nunberg claims that (11) is not a case ofpredicate transfer but ofmeaning However, he does not clearly state the conditions under which meaning transfer takes place. Besides, (11) fulfils the two requisites he imposes on predicate transfer: (i) there exists a correspondence between the ham sandwich and the property of the customer ordering/eating a ham sandwich, and (ii) the property added (i.e. ‘ to be waiting for his check') is just as noteworthy in the customer-order relationship as the property of being parked is in the driver-car relationship. Nunberg's problem, if he were to regard this example as a case of predicate transfer, is that anaphoric reference is not made to 吁lam sandwich' , as his the。可 of predicate transfer would predict, but to ‘ customer':
缸ansfer.
(12) a. The ham sandwich is waiting for his check and he is getting restless. b. *1页e ham sandwich is waiting for his check and it is getting restless. Ifwe consider again (11) from a different point ofview , we observe that it contains a source-in-target metonyrny where ‘ ham sandwich' is a subdomain of the matrix domain ‘ customer' , as represented in the following figure:
Figure 5.
ORDER FOR CUSTOMER metonymy
304 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco
The comparison of examples (9) , (1 0) , and (12) reveals that source-intarget metonymies make use of the target domain of the mapping for anaphoric reference , while target-in-source metonymies prefer the source domain. However, closer examination of figures 3 , 4 , and 5 shows that both metonymic types coincide in selecting the matrix domain for reference , no matter whether it is the source or the target of the mapping. In order to account for the principles that govern this aspect of metonymic anaphora , Ruiz de Mendoza and P的z (2001: 351) have put forward the Domain Availability Princ收le or DAP , according to which only the matrix domain of a metonymic mapping is available for anaphoric reference. With regard to the activity of this principle , we additionally suggest that the preference for the matrix domain as the domain of reference may be due to the greater degree of development this domain possesses: (i) the source domain of source-in-target metonymies must be 臼lly developed into its corresponding target domain by means of a process of domain expansion; and (ii) it is often the case that the target domain of a target-in-source metonymy is not well-defined (see section 2). Finally, we observe that this a1ternative account provides a more elegant explanation of metonymic anaphora than Nunberg's since it allows us to deal with cases of deferred indexical reference , predicate transfer, and meaning transfer in terms of just one principle. A different account of anaphoric relations in metonymic contexts , which is s仕ongly inspired by Fauconnier's (1 985) well-known Ident i.fication Principle , has been provided by Stirling (1 996). According to the Identification Principle , whenever two objects , a and b , are connected by means of a pragmatic function , a description of a may serve to identify b. In her proposal , Stirling (1 996) builds a model of possible 可pes of metonymic anaphora, which is exemplified below: (13)
Plato is on the top shelf. It is bound in leather.
Type (i)
(1 4)
Plato is on the top shelf. You 'll find that he is a very interesting author.
Type (ii)
(1 5)
Plato
is α great
author. He is on the top shelf.
Type (iii)
Each of these types is graphically represented in a different way , as the following figures show:
Metonymic motivation in anaphoric r，φrence 305
antecedent
anaphor
designates b via a, i.e. ac tI vates:
a
R 一~
intended to refer to b
\〈士c li工一/
Figure 6. Type (i) of anaphoric relation involving metonymy
anaphor
antecedent designates b via a, i.e. activates:
a
intended to refer to a
b
b anaphoric link
Figure 7. Type (ii) of anaphoric relation involving meton沪ny
antecedent
anaphor
designates a
intended to refer to b
b
a
R
a 一一一一书， ~ b Figure 8. Type (iii) of anaphoric relation involving metonymy
306 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco
In her analysis Stirling (1996) makes a distinction between types (i) and (ii) , where there is a metonymic mapping in the antecedent of the anaphoric pronoun , and type (iii) , in which the metonymy develops in the a司joined sentence. Besides, she notices that only some metonymic relations can produce acceptable anaphoric connections. In order to study in depth the principles that guide the selection of anaphoric pronouns , she elaborates on Fauconnier's (1985) analysis ofmetonymic anaphoric relations. Fauconnier claims that the choice of the anaphoric pronoun is greatly influenced by its reflexive or non-reflexive status, and by the animate or inanimate nature of the trigger and targe t. However, Stirling partially criticizes Fauconnier's treatment of metonymy and anaphora by arguing that Fauconnier does not take into account the grammatical features of the anaphoric expression. The fulllist of examples provided by Stirling (1996: 82) is reproduced below: Inanimate trigger - Animate target (order - customer metonymy) (16) Type (i) a. The mushroom omelet a• b left without paying his bill. Heb jumped into a tax Ï. b. ?The mushroom omelet a• b left without paying its bill. Itb jumped into a taxi. (17) Type (ii) a. ?The mushroom omelet a• b left without paying. It a was inedible. b. *The mushroom omelet a• b left without paying. He a was inedible. (1 8)
Type (iii) a. *The mushroom omeleta was too spicy. It b left without paying. b. The mushroom omeletawas too spicy. Heb left without paying.
Animate trigger - Inanimate target (author - oeuvre metonymy) (1 9) Type (i) a. Plato a• b is on the top shelf. Itb is bound in leather. b. ?Plato a• b is on the top shelf. He b is bound in 1eather.
5. The trigger in Stirling's account call the source domain.
coπesponds
to what other cognitive linguists
Metonymic motivation in anaphoric r，矿每rence 307
(20) Type (ii) a. Plato a • b is on the top shelf. You'll find that he a is a very interesting author. b. *Plato a • b is on the top shelf. It a is a very interesting author. (21) Type (iii) a. Plato a is a great author. Heb is on the top shelf. b. Plato a is a great author. It b is on the top shelf. From her analysis Stirling (1996) conc1 udes that in 可pes (i) and (ii) there is a preference for the pronoun to agree with the intended referent, whereas type (iii) prefers the animate pronoun. In our view, Stirling's account fails to provide a consistent explanation of anaphoric reference to metonymy in at least two ways. Firstly, her proposal is not a fully principled one since she only postulates the existence of some general tendencies, which are easily falsified by examples like (17b) where the expectation would be for the anaphoric pronoun to agree with ‘ customer\In order to account for exceptions like this , it would be necessary to have a complementary account of other factors which may have a role in determining the selection of anaphoric pronoun and which may even override a purported general tendency. Secondly, her analysis is ultimately more descriptive than explanatory as she does not search for the motivation behind the preference for one anaphoric pronoun over another. In dealing with Nunberg's (1995) treatment of this controversial issue , we have previously referred to the DAP. In the rest of this paper we shall show how the choice of anaphoric pronoun is guided by a set of principles which, together with the DAP, provide a systematic explanation for this problem. We start our analysis with examples (16) to (18). All of them are cases of source-in-target metonymy where the meal ordered is a subdomain of the customer in the restaurant frame. As we mentioned when dealing with (12) , the DAP only licenses the ma位ix domain (i.e. ‘ customer') for anaphoric reference , which explains why (16a) is acceptable , whereas (16b) is no t. In contrast, the examples under (17) are slightly different. (1 7a) is odd because, according to the DAP , the metonymic noun cannot be the antecedent for the pronoun it. Moreover, the frame of reference is changed from the first to the second c1 ause in such a way that the hearer is led to ignore the metonymy and look for an adequate referent for it. More interesting is the case of (17时， which is ruled out even though it follows the DAP (i.e. it chooses the matrix domain ‘ customer' for the anaphoric opera-
308 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco
tion). The problem here is that the pronoun licensed by the DAP is semantically incompatible with the predicate of the adjoined sentences (i.e. being edible is not a property of customers). As a resu1t, the processing of (17b) would need the activation of a second metonymic mapping in the adjoined clause from the customer to the meal he has ordered. This operation would involve too much processing effort on the part of the hearer since it would require to reverse the previous metonymic operation carried out in the first clause. In order to account for this restriction in the selection of the predicate of a conjoined or adjoined sentence in a metonymic context, we propose the Constraint on Metonymic Anaphor，αor CMA. According to this constraint, whenever anaphoric reference is made to a metonymic noun phrase , the anaphoric pronoun cannot have an independent metonymic interpretation, different from the one assigned to its antecedent. In most cases this formulation of the CMA amounts to stating that the anaphoric pronoun cannot be metonymic itself. However, consider the following example , in which the predicate of the second clause does require a metonymic reading of the anaphoric pronoun: (22)
Nixon bombed Hanoi; later, he bombed Cambodia.
This sentence is an instantiation of the RULER FOR ARMY metonymy, in which the matrix domain ‘ Nixon' stands for ‘ the air force under his control' (see figure 9). As we have noted , the motivation underlying the CMA is the enormous processing effort that is required to activate a different metonymic meaning of the anaphoric pronoun in the second clause in order to make it compatible with the predicate. However, this processing effort is considerably reduced if the interpretation of the predicate of the second clause does not require the hearer to activate a different metonymy, but just to maintain active the same metonymy of the first clause, as is the case in (22). This situation is not ruled out by the CMA , which only hinders the possibility of having a different metonymic reading of the anaphoric pronoun. Sentence (22) is a case of target-in-source metonymy. The following related examples show that the interaction between the DAP and the CMA also regulates the selection of anaphoric pronoun and its compatible predicate in target-in-source metonymles: (23) a. Nixon bombed Hanoi and he did not know what he was doing. b. *Nixon bombed Hanoi but they were under orders. c. *Nixon bombed Hanoi and he was a special unit.
Metonymic motivation in anaphoric 叫声rence 309
to the DAP, only ‘ Nixon' , the source domain, is available for maphoric reference.ηlis explains why (23a) is acceptable while (23b) , w也ich chooses one of the subdomains for anaphoric reference , is not. In 阳rn， (23c) , which does follow the DAP , is not possible because it would require a different metonymic reading of the pronoun he in order for it to be . I:Ompatible with the predicate of the co时 oined sentence , a situation which sωbe discarded by the CMA. ~ccording
NIXON
Source
Figure 9. NIXON FOR AIR FORCE metonymy.
Finally, consider the following pair of examples , which are of particular interest in showing the irrelevance of the animate/inanimate character of the domains involved in a metonymic mapping: (24) a. Chrysler has changed its employment policy and it will lay off a hundred workers as a result. b. *Chrysler has changed its employment policy and helshe will lay off a hundred workers as a result. These sentences contain another example of a target-in-source metonymy (namely, COMPANY FOR MANAGEMENT) where only the company, i.e. ‘ Chrysler' , can be chosen for anaphoric reference. The comparison of (24a) and (24b) shows that the DAP only licenses the ma创x domain as the antecedent for reference in spite of being inanimate, which hinders the possi-
310 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco
bility of selecting other domains to function as possible antecedents for the anaphoric pronoun. 6 Now, we turn our attention to the ORDER FOR CUSTOMER examples in (1 8). According to Stirling (1996) , these sentences represent the only cases of true metonymic anaphora as the metonymic mapping is entirely located in the anaphoric pronoun. (18a) is ruled out because the anaphoric it , which can only make reference to the ‘ mushroom omelet' , is semantically incompatible with the predicate of the subsequent sentence left without paying. It is worth noting , however, that the first clause in (18a) does not involve any metonymy at all , and that the unacceptability of the second clause relies solely on the semantic anomaly that results from the aforementioned impossible combination. In contrast, sentence (1 8b) is totally correct. The feasibility of this example has nothing to do with the existence of a metonymic mapping in the second sentence. On the contrary, (18b) can hardly be regarded as an instantiation of a metonymy since the compatibility between the pronoun he and its predicate is achieved without the need of a metonymic shift ofreferent. 7 Examples (1 9) - (21) require a different analysis since the metonymic processes involved in them are of a different nature. First of all , consider the following sentences: (25)
1 enjoyed Plato.
(26)
Plato is on the top shelf.
At first sight, it may seem that both (25) and (26) contain examples of the same metonymic mapping. However, a closer examination reveals that this is not the case. In (25) , Plato stands for ‘ Plato's work'. It is a source-intarget metonymy where an author and his work stand in a domainsubdomain relationship. The problem arises when we try to account for the meton严ny in (26). This metonymy cannot be explained by means of the
6. In (24) it is also possible to choose the pronoun they as the subject of the conjoined sentence; the motivation underlying the plausibility of this pronoun , which is not a matter ofmetonymy, will be explained in detail below. 7. N otice should be taken that (l 8b) only seems possible in narrative contexts like the following: John ordered a mushroom omelet and a beer. The mushroom omelet was too spicy. He (=John) left without paying.
Metonymic motivation
切 anaphoric r，价rence
311
mapping alone, since in this example we are not making reference to the direct outcome ofPlato's activity as a writer but to the representation of his ideas in a specific medium (e.g. a book, a rare manuscript, a CD ROM, etc.). Thus , one further mapping is necessary 仕om ‘ Plato's work' to ‘ a sample of his work'. In Ruiz de Mendoza and Pérez (2001) , and in Ruiz de Mendoza and Díez (2002) metonymies like the one in (26) are called double metonymies (see figure 10): AUTHOR FOR WORK
Figure 10.
AUTHORFOR WORKFORMEDIUM
metonymy
Because of the double mapping, the metonymy in (26) has two different matrix domains. The same is the case in examples (1 9) 一 (21)， where there are two matrix domains available for reference: ‘ author' and ‘ medium\In order to account for the selection of the domain of reference which works as the antecedent for the anaphoric pronoun we further postulate the Domain Combinability Principle or DCP. According to this principle, whenever two domains are available for anaphoric reference to a metonymic noun, we tend to select the domain that is semantically more compatible with the predicate of the sentence containing the anaphoric pronoun. The DCP explains why in (19a) the second matrix domain is chosen for anaphoric reference , since ‘ being bound in leather' is a prope此y of books and not of writers. Likewise, (20a) selects the anaphoric he because the first
312 Francisco José Ruiz de Mendoza /báñez and Olga /sabel Díez Velasco
matrix domain is semantically more compatible with the adjoined predicate ‘ be a very interesting author' than the second matrix domain. Compare now (l 9b) and (20b). Although both examples violate the DCP, they hold different degrees of acceptab i1ity: (1 9b) is odd but possible, whereas (20b) is completely ruled out. The joint activity of the DAP and the DCP cannot account for this. In effect, (20b) chooses the second or final matrix domain in the metonymic chain for reference (i.e. ‘ medium') , but is semantically compatible with the first one (i.e. ‘ author') , while (19b) involves the opposite operation, that is , it selects the first or initial matrix domain for reference though it is semantically compatible with the second one. The difference in the degree of acceptability between these two sentences calls for the formulation of a third principle, the Domain Precedence Principle or DPP , according to which, in cases of double metonymic mappings , unless the predicate combines better with the final matrix domain , reference is preferably made to the initial matrix domain. Accordingly , (20b) , which violates both the DPP and the DCP , is completely incorrect , whereas (l 9b), in following the DPP although not the DCP , is partially acceptable. Finally , the tension between the DPP and the DCP also accounts for the acceptab i1ity ofboth (21a) and (21b). In (21 时， although the adjoined predicate is semantically more compatible with the final matrix domain , the selection of the initial matrix domain for anaphoric reference is sti11 possible on the basis of the DPP. In (21 b) , on the other hand , it is the DCP that applies and licenses the selection of the final matrix domain for anaphoric reference. In the light of the previous analysis we observe that the DCP is stronger than the DPP , at least in cases in which the metonymy develops in the antecedent. Thus, if (l 9b) followed the DCP it would be fully acceptable, but since it only follows the DPP it is only partially so. On the other hand, if the metonymy develops in the anaphoric pronoun , as in (21a) and (21 时， the two principles have the same strength. Perhaps , this can be explained by the distinct nature of metonymic anaphors. When the metonymy is in the antecedent as in (1 9) , we have a situation in which the DAP needs to apply first. However, the DAP does not apply in (21a) and (21b) simply because the antecedent is not metonymic; that is to say, the antecedent plays no role whatsoe
Metonymic motivation in anaphoric r，矿èrence 313
whatsoever - whether metonymic or not 一 in the previous c1 ause. At first sight, the second c1 ause would not be metonymic; however, in the context of the first c1 ause we understand that it refers to the ‘ medium' in which Plato's work has been conveyed to us. This interpretation process requires the implicit activation of the three domains of the metonymic chain and the anaphoric link to an implicit domain. All 由is explains why the greater strength of the DCP plays no role in selecting (21b) over (21a). The anaphoric pronoun (21a) has an explicit antecedent with which it agrees in gender and number. The DPP is not required to exert any significant pressure to preserve the form of the pronoun, which more or less naturally agrees with the antecedent. Interestingly enough, when no grammatical agreement is required because the antecedent is implicit, as in (21 时， the DCP demonstrates its greater strength by calling for the selection of the matrix domain which is conceptually compatible with the predicate of the clause to which the anaphoric pronoun belongs. In short, the previous discussion suggests that the above-mentioned principles vary in terms oftheir strength ofapplication. Firstly, we have the DAP , which is the strongest one, since failure to apply it results in fully anomalous sentences, and it is the only one that holds both in cases of single and double metonymies. Secondly, we have the DCP, which is only found in cases of double metonymy , and is stronger than the DPP because of its semantic basis. Thirdly, the DPP is the least strong of the three principles probably because of its formal nature. Finally, the CMA is not strictly a principle but a filter for a忧empts to generate metonymies that would cancel out a metonymy which is already active in the antecedent. We shall now explore whether these three principles, together with the CMA, can also adequately explain anaphoric reference to relative sentences. Take the following examples 企om Fauconnier (1985: 9): (27) a. *The omelet, which left in a hurry, was too 甲icy. b. *The omelet, which was too 早icy， 1，笔斤的 a hur，η. (28) a. Plato, who was a great man, is on the top shelf. b. Plato, who is on the top she~汇 was a great man. The sentences in (27) are instantiations of the ORDER FOR CUSTOMER metonymy. As has been noted above , this is a source-in-target metonymy in which the food ordered by a customer is a subdomain of the customer, which is the ma创x domain. According to the DAP, the matrix domain
314 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco ‘ customer' is the only domain available for anaphoric reference. Relative
pronouns in English do not distinguish number or masculine/feminine gender, but they signal the difference between animate and inanimate entities. Thus , the examples in (27) are ruled out because they break the DAP: the relative which is not referential to the matrix domain ‘ customer' , which is an animate being, but to one of its subdomains (i.e. ‘ omelet'). The examples in (28) contain instances of the AUTHOR FOR WORK FOR MEDIUM metonymy. Since it is a double metonymy , there are two matrix domains ('author' and ‘ medium') available for reference. First, (28a) is acceptable because the predicate of the relative clause is semantically compatible with the initial matrix domain. Therefore, the DCP and the DPP coincide in selecting the initial matrix domain ‘ Plato' as the antecedent for anaphoric reference. In (28b) we find a slightly different case since the choice of who as relative pronoun violates the DCP. Thus , the prope此y of ‘ being on the top shelf cannot be congruently applied to 'people' but is semantically compatible with ‘ books\However, this sentence is not ruled out because it follows the DPP , since the domain selected for reference in (28b) is the initial matrix domain. The reason for the greater strength of the DPP in cases of anaphoric relative pronouns is to be sought in the position of the relative clause in the sentence. Careful examination of (28b) shows that the relative pronoun appears before the second matrix domain of the metonymy has been activated, which naturally prevents this domain from being chosen for anaphoric reference. These examples are evidence that anaphoric reference to meton严nic noun phrases in relative clauses is not random and that it can be consistently explained in terms of the same principles which we have put forward to explain other cases of metonymic anaphora. Take now this apparently problematic example: (29)
1 called the garage and they will have the car rea命 by tomorrow.
In (29) we find an example of a target-in-source metonymy where the matrix domain ‘ garage' stands for ‘ the worker who answered the phone'. However, the use of anaphoric they instead of it in (29) suggests that the matrix domain has not been selected as the antecedent for anaphoric reference. At first sight, this fact may suggest that this example flouts the DAP; on closer inspection, however, we find that this is not the case since reference is not made to 也e subdomain of the metonymic mapping either. Thus , in (29) they does not refer to the person who answered the phone, but to the
Metonymic motivation in anaphoric r，价rence 315 group of people who work in the garage. The reason for this lies in the fact garage activates a frame or scenario in such a way that all the elements belonging to this 仕ame are eligible for reference. Therefore , (29) cannot be considered a case of anaphoric reference to a metonymy, but a case of anaphoric reference to a 仕ame element in which no metonymy is involveQ; 也at is to say, the metonymy is not present in the conjoined sentence. Additional evidence for 由is claim is found in the analysis of the following 臼n tence: 也at
(30)
1 took my car tomorrow.
ω the
garage and they wil/ have it
ready 妙
Although in this example garage is not used metonymically , the COI飞joined sentence selects the same frame element (i.e. the workers) as the ant臼e dent. This reinforces our view that the metonymy has no role in the choice of 捕。J in (29) since neither the source nor the target, but a different 企缸ne element works as the antecedent. Accordingly, both (29) and (30) 缸'e examples of what may be termed implicative r，价'rence， where the pronoun they refers to an implicated collective noun. 8 Cases of implicative reference are fairly usual when dealing wi也∞1lective nouns as (31) illustrates: (31) a. Wall Street is in panic and they wil/ keep in panicfor some time. b. Washington has ordered a new attack this morning and now th句1 regret ìt. In these two examples, the pronoun of the conjoined sentence makes implicated reference to one of the frame elements that Wall Street and Washington activate , and not to their corresponding metonymic targets. Nevertheless , it is also possible to select the metonymic noun as the antecedent as in Wall Street is in panic and it wil/ keψ in panic for some time. In this c臼e， the selection of the pronoun follows the requirements of the DAP, the ma位ix domain being the only domain available for anaphoric reference. 8. In Clark (1977) and Clark and MarshaIl (1 981) the process by means ofwhich reference to a noun phrase can be inferred from other entities mentioned explicitI y in the discourse is caIled bridging. However, we prefer the term implicative rE作rence since it brings into focus the twofold inferentiaI and referentiaI nature of this phenomenon.
316 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco The notion of implicative reference allows us to account for other apparently problematic examples. Consider the following pair of examples from Panther and Radden (1 999: 10): (32)
a. The harpsichord has the flu. His part has been taken over by the grand piano. b. The harpsichord has the flu. Its part has been taken over by the grand piano.
Although Panther and Radden (1 999) have noted the problem these two examples pose for a theory of anaphoric reference to metonymic nouns , they do not attempt to offer a solution. In our view , the difference between (32a) and (32b) lies in the fact that in the former the antecedent is the matrix domain of the metonymy (i.e. ‘ the harpsichord player') , whereas in the latter it is no t. Its in (32b) does not refer to the subdomain of the metonymy but it directly refers to ‘ harpsichord' as part of the conceptual frame activated in the interpretation of the first sentence. This means that the metonymy has been abandoned once we reach the adjoined sentence. This analysis is also substantiated by the possibility of saying Peter has the flu; the part 0/ the sax wi/l be taken over by the trumpet tonight (where Peter is a sax player), which shows that the adjoined sentence can make reference to any frame eI ement as long as it is relevant enough. To sum up , in this section we have argued that anaphoric reference to metonymic nouns is subjected to three principles and a constraint, which together offer a consistent explanation of one of the most problematic areas within the cognitive the。可 ofmetonymy.
5. Conclnsion In this paper, we have analyzed another area of linguistic structure in which metonymy plays an important role. This has served to provide additional evidence in favour of the c1 aim that metonymy has an important impact on grammar. More specifically, we have studied the metonymic motivation which underlies the selection of the relevant anaphoric pronoun. In order to show how metonymy lies at the basis of some problematic cases of anaphoric reference , we have started by providing a definition of metonymy. In this connection , we have argued that the only defining feature of metonymy is the domain幽internal nature of the mapping while the rest of charac-
Metonymic motivation in anaphoric rl价rence 317
teristics typically attributed to metonymy are only a natural consequence of the domain-subdomain relationship that exists between the source and target domains. Moreover, we have observed that the ma位ix domain of a metonymy must always be a clearly delineated domain, which can be easily identified and accessed; we have further noted that in cases in which the . two domains involved in a metonymic mapping would quali命 to become the matrix domain , there are other cognitive principles (e.g. controlling entity over controlled entity) which guide the status oftheir relationship. Our analysis has also shown that anaphoric reference to simple metonymies (i. e. those in which there is only one matrix domain involved) is govemed by the Domain Availability Principle and the Constraint on Metonymic Anaphora , while the animate/inanimate nature of the domains of a metonymy is not relevant at all. Besides, in order to explain anaphoric reference in cases of double metonymic mappings (i.e. those in which there are two matrix domains) we have put forward two additional principles, the Domain Precedence Princ伊le and the Domain Combinability Principle , which have shown to work in all cases , and we have ranked them according to their degree of applicability. Our view has also proved valid to offer a consistent explanation of some cases of metonymic anaphora which have been regarded as problematic by other authors. Finally, we have made a distinction between reference to a metonymic noun and implicative reference with the pu甲ose of adequately accounting for examples which appear to involve metonymy but which, as a matter of fact , do not.
References Barcelona, Antonio 2000 Metaphor and Metonymy at the Crossroads. BerlinINew York: Mouton de Gruyter. Bybee, Joan, Revere Perkins and Wiliam Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: The University of Chicago Press. Clark, Herbert H. 1977 Bridging. In: Johnson-Laird and Wason (eds.) Th inking: Readings in Cognitive Science , 411 -432. Cambridge: Cambridge University Press.
318 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco Clark, Herbert H. and C. R. Marshall 1981 Defmite Reference and mutual knowledge. In: Webber and Sagi (eds.) , Elements 01 Discourse Understanding. Cambridge: Cambridge University Press. Croft, William 1993 The role of domains in the interpretation of metaphors and meωn归国es. Cognitive Linguistics 4 , 4: 335-370. Dik, Simon 1989 The Theory 01 Functional Grammar, Vol. 1: The Structure 01 the Clause. Dordrecht: Foris. Dirven, René 1993 Metonymy and metaphor: Different mental strategies of concepualisation. Leuvense Bijdragen 82: 1-25. Dirven, René and Ralf Pöring (eds.) 2002 Metaphor and Metonymy in Comparison and Contrast. Berlin: Mouton de Gruyter. Fauconnier, Gilles 1985 Mental Spaces. Cambridge , Mass.: MIT Press. Fillmore, Charles 1985 Frames and the semantics of understanding , Quaderni di Semantica Vo l. VI, 2: 222一254. Goossens , Louis 1990 Metaphtonymy: The interaction of metaphor and metonymy in expressions for linguistics action. Cognitive Linguistics 1, 3: 323-340. Meton严nic bridges in modal shifts. In: Kl aus-Uwe Panther and 1999 Günter Radden (eds.) , Metonymy in Language and 1页 ought， 193-211. Amsterdam: John Benjamins. Heine , Bemd 1995 Agent-oriented vs. epistemic modality: Some observations of German modals. In: J. Bybee and S. Fleischman (eds.) , Modali.伊 and Grammar in Discourse , 17-53. Amsterdam: Benjamins. Johnson, Mark 1987 The Body in the Mind: The Bodily Basis 01 Meaning, Reason and Imagination. Chicago: University of Chicago Press. Kövecses , Zoltán 1990 Emotion Conc写pts. New York: Springer-Verlag. 2000 Metaphor and Emotion: Language, Culture, and Bo命 inHuman Feeling. Cambridge: Cambridge University Press. Kövecses , Zoltán and Günter Radden 1998 Metonymy: Developing a Cognitive Linguistic View. Cognitive Linguistics 9, 1: 37-77.
Metonymic motivation in anaphoric r，价'rence 319
Lak:off, George 1987 Women , Fire, and Dangerous Things: Wh at Categories Reveal About the Mind. Chicago: University of Chicago Press. 1993 The contemporary theory of metaphor. In: A. Ortony (ed.), Metaphor and Thought , 2nd. ed. Cambridge: Cambridge University Press. Lak:off, George and Mark Johnson 1980 Metaphors W注 Live By. Chicago , London: The University of Chicago Press. 1999 Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western Thought. New York: Basic Books. Lak:off, George and Mark Tumer 1989 More than Cool Reason: A Field Guide to Poetic Metaphor. Chicago: The University of Chicago Press. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Vo1. 1. Theoretical Prereqisites. Stanford, Cal.: Stanford University Press. 1993 Reference-point constructions. Cognitive Linguistics 4 , 1: 1-38. Nunberg, Geoffrey 1995 Transfers ofmeaning. Journal ofSemantics 12: 109-132. Panther, Klaus-Uwe and Günter Radden (eds.) 1999 Metonymy in Language and Thought. Amster由mlPhiladelphia: John Benjamins. Panther, Klaus-Uwe and Linda Thomburg 2000 The EFFECT FOR CAUSE meton严ny in English grammar. In: A. Barcelona (ed.) , Metaphor and Metonymy at the Crossroads: A Cognitive Perspective , 215-232. BerlinINew York: Mouton de Gruyter. 2002 The roles of metaphor and metonymy in English -er nominals. In: René Dirven and RalfPörings (ed.) , Metaphor and Metonymy in Comparison and Contrast, 279-319. Berlin: Mouton de Gruyter. PérezHem缸dez， Lorena and Francisco Ruiz de Mendoza Ib姐ez 2002 Grounding, semantic motivation, and conceptual interaction in indirect directive 叩eech acts. Journal of Pragmatics 34, 3: 259284. Radden, Günter 2000 How metonymic are metaphors? In: A. Barcelona (ed.) , Metaphor and Metonymy at the Crossroads: A Cognitive Per.再pective， 93一 108. BerlinINew York: Mouton de Gruyter.
320 Francisco José Ruiz de Mendoza Ibáñez and Olga Isabel Díez Velasco Radden, Günter and Zoltán Kövecses 1999 Towards a theory of metonymy. In: K. Panther and G. Radden (eds.) , Metonymy in Language and Th ought, 17-60. Amsterdaml Philadelphia: John Benjamins. Ruiz de Mendoza Ibáñez, Francisco J. 1997 Cognitive and pragmatic aspects of metonymy. Cuadernos de Filología Inglesa 6, 2: 161 一 178. 1999 The role of cognitive mechanisms in making inferences. Journal of English Studies 1: 237-255. 2000 The role of mappings and domains in understanding metonymy. In: A. Barcelona (ed.) , Metaphor and Metonymy at the Crossroads: A Cognitive Perspective , 109-132. BerlinINew York: Mouton de Gruyter. Ruiz de Mendoza Ibáñez, Francisco J. and Olga 1. Díez Velasco 2002 Pattems of conceptual interaction. In: R. Dirven and R. pörings (eds.) , Metaphor and Metonymy in Comparison and Contrast, 489-532. Berlin/N ew York: Mouton de Gruyter. Ruiz de Mendoza Ibáñez, Francisco J. and Lorena Pérez Hemández 2001 Metonyrny and the grammar: Motivation , constraints and interaction. Language & Communication , Vol. 21 , 4: 321-357. Seto , Ken-ichi 1999 Distinguishing metonymy from synecdoche. In: K. Panther and G. Radden (eds.) , Metonymy in Language and Thought , 91 一 120. AmsterdamIPhiladelphia: John Benjamins. Stirling, Lesley 1996 Metonymy and anaphora. In: W. Mulder and L. Tasmowki (eds.) , Belgian Journal of Linguistics J0 (Coherence and Anaphora) , 69-88. AmsterdamIPhiladelphia: John Benjamins. Traugott, Elizabeth C. and Ekkehard König 1991 The semantics-pragmatics of grammaticalization revisited. In: E.C. Traugott and B. Heine (eds) , Approaches to Grammaticalization , Volume 1: 189-218. Amsterdam: John Benjamins. Waltereit , Ri chard 1999 a Grammatical constraints on metonymy: On the role of the direct objec t. In: K. Panther and G. Radden (eds.) , Metonymy in Language αnd Thought , 233-255. AmsterdamIP hiladelphia: John Benjamins. 1999 b What means to deceive yourself: The semantic relation of French reflexive verbs and their corresponding transitive verbs. In: Z. Franjzyngier and T. Curl (ed时 ， Reflexives. Forms and Functions , 257-272. AmsterdamIPhiladelphia: John Benjamins.
Predicative adjectives and grammatical-relational polysemy: The role of metonymic processes in motivating cross-linguistic differences Rita Brdar-Szabó and Mario Brdar
Abstract Some of the contrasts between English, on the one hand, and German, Croatian and Hungarian, on the other, conceming the availability of certain constructions with predicative adjectives may be motivated by the interaction of cognitive and structural factors. Specifically, we c1 aim that grammatical-relational polysemy in these English constructions with predicative adjectives may rest in some cases on metonymic processes. In other words , English is shown to rely heavily on metonymic processes in rearranging its predicate argument structures while at the same time keeping 岛rmally one and the same form of the predicative expression. The other three languages we are interested in here tend to formally indicate different arrangements in predicate argument structure by using formally different predicative expressions. Keywords: active zone specification, ascriptive construction, contrastive typology, grammatical constructions, grammatical relations, meton严ny， polysemy, predicative adjectives, raising construction, subject
1. Introduction Traditional contrastive studies on predicatively used adjectives have established that English predicative adjectives often translate into other word c1asses in Croatian, Hungarian and German, particularly verbs and adverbs, although direct structural correspondents do in fact exist in the languages in
322 Rita Brdar-Szabó and Mario Brdar question. This phenomenon is illustrated below on some parallel examples from English and Croatian, respectively:
(1)
a. whαt was she afraid 旷? b. Cega se bojala? What REFL fear-PART
(2)
a. 1 am very adaptable 一 ω circumstances. b. Vrlo se lako prilagoâavam okolnostima. very REFL easily adapt circumstances-DAT
(3)
a. ... and 1 am most appreciative ofthe trouble he took. b. ... 1 jako cijenim njegov trud. and very appreciate his trouble
(4)
a. James was bri，吃fαbout his adventures. b. James je vrlo sazeto govorio 0 svojim James AUX very concisely spoken about his avanturama. adventures.
(5)
a. The Square was baking hot. b. Na trgu je bilo strasno vruée. on square COP been terribly hOtADv
(6)
a. He is certain to show up. b. Sigurno ée se pojaviti. certainly will REFL appear
Explanations , if any are offered at all , chiefly focus on contextual andlor stylistic factors (where the pressure oftranslational needs may count among either of these). Even when there is no structural/contextual pressure , nonadjectival predicates are very frequent as counterparts of English predicative a司jectives， regardless of whether the English adjective is deverbal or not (cf. Ivir 1983: 105). Among the conditions conducive to such differences in the dis位ibution of predicative adjectives between English and Croatian, and most Slavic languages , we could perhaps suggest the fact that the two languages belong to diametrically opposed types as far as the verbal dynamism of the predi-
Predicative adjectives and grammatical-relational polysemy 323
is concemed. 白lÏs p缸ameter， f lfSt discussed by Curme (1931: 22), and later taken up by Praguian scholars such as Mathesius (1961), and Vachek (1961: l3 5), distin伊ishes between languages favouring centripetal predicates, where the semantic core is a verb , and languages heavily using centrifugal predicates, where the semantic core of the predicate is displaced 齿。m 由e finite verb to some nonverbal elements (predicative adjectives , nouns, e也). Croatian is thus considered to e对libit centripetal, dynamic , verbal predicates, while Germanic languages exhibit a larger amount of cen创臼.gal， non-dynamic and verbo-nominal predicates. Some predictions 也at 岛llow from this parameter appear to be corroborated by the dis位ibu tion of constructions with so-called functional verbs. A large number of semantically bleached German verbs appe缸 in such constructions wi也 deverbal nouns (e.g. ins Rollen geraten , in Bewegung kommen , etc.). Comparable constructions in English are somewhat less frequent, and predominantly contain NPs, in contrast to German, where the nominalized nouns are introduced by prepositions. These constructions 町'e， on the whole, certainly less 仕equent in Croatian, but some types 町e very common. We could therefore assume that 由1S P町创neter is responsible for most of 也e contrasts exhibited in (1-6). However, such an explanation would be unsatisfactory 岛r a number of reasons. Firstly, it is too generalωit does not explicitly link the two phenomena. Secondly, it is falsified by some cross-linguistic facts. German, which is assumed to be on the cen位ifugal pole of the parametric continuum, and in some respects (cf. functional verb cons住uctions) is even more centrifugal than English, is shown below to 阴阳m with Croatian and Hungarian, since it systematically fails to e油ibit adjectival synt缸 in the predicate. Further, the fact th创 predicative adverbs (or predicatively used prepositional phrases) frequently appear as counterP町ts of adjectives is left unexplained, since these are centrifugal constructions too. Finally, cen位ifugal predicates with functional verbs usu创ly contain deverbal nouns or adjectives, however, we find that a司jectival predicates are avoided even when there is no such morphological link. In other words , if we say 由at English centrifugal predicates 缸'e "derived" from verbal ones, and that Croatian (and other languages) simply fall back on simpler 岛rms， it is not c1ear why this should happen when there
臼ite
324 Rita Brdar-Szabó and Mario Brdar
is the case in "explanations" invoking stylistic and contextual factors. We will also show that these facts cannot be handled by means of the parametric contrast ‘ centrifugal' vs. ‘ centripetal\Not an insignificant number of such cases, along with certain other constructions involving predicative adjectives , can be well motivated by the interaction of some cognitive and structural factors. Specifically, we shall claim that grammatical-relational polysemy in these English constructions with predicative adjectives may rest in some cases on metonymic processes , and that some other languages like German , Croatian and Hungarian fail to tolerate grammatical polysemy to the same extent, in general , and in particular when it is based on metonymy. By ‘ grammatical-relational polysemy' we mean the occurrence of several diι ferent argument-structures linked with a single predicative expression and presenting basically the same state of affairs from different points of view. In other words , English will be shown to rely heavily on metonymic processes in rearranging predicate argument structures while at the same time keeping formally one and the same form of the predicative expression. The other languages we are interested in here tend to formally indicate different arrangements in predicate argument structure by using a formally different predicative expression, which may be morphologically more or less transparently related to the direct counterpart of the English a吗jective. Metonymic processes will be shown to provide one of the links missing in the account based on the centrifugal vs. centripetal parameter. The organization ofthe article is as follows. In Section 2 , we briefly explain what we exactly mean here when we claim that some contrastive phenomena can be well-motivated by relating this particular sense of the term to a more general issue of motivation in grammar and in language in general. After a short discussion of the nature of metonymy in Section 3, we turn in Section 4 to grammatical-relational polysemy. Having thus established the general framework , we demonstrate in Section 5 the ubiquity of the phenomenon in English on a range of constructions , while at the same time uncovering conceptual metonymies that underlie these cases of polysemy and then proceed to uncover some limitations on the availability of these constructions imposed in the other three languages. Our findings are sum
Predicative adjectives and grammatical-relational polysemy 325
2. 00 motivatioo io grammar, cogoitive aod cootrastive Motivation of linguistic phenomena, al也ough it is , informally speaking, a relatively simple task of searching for some meaningful links between linguistic expressions and their contents and contexts of usage, and therefore an integral p町t of any linguistic enterprise, turns out, when it comes to its execution, to be largely a matter of ideological position. What is actually meant by motivation is determined by one's language philosophy and the actual grammatical model adopted. In a generatively-oriented model , the task is reduced to establishing the set of possible constructions (sentences, u忧erances， etc.), and thus indirectly the set of impossible ones, as well. It could also be the other way round; the model may set itself as its goal the specification of the set of constraints which filter out the unacceptable constructions for example, as 扭曲e 企amework of Op timality Theory. In such a model, "motivation" is reduced to the interaction of purely system-internal cons位'aints. Cognitively and functionally oriented linguists seem to have reached a broad coosensus on the issue of motivation with respect to at least two of its aspects (cf. Lakoff 1987, Langacker 1987 and 1991 , Haiman 1980, 1983). First1y , motivation is a phenomenon exhibited by a range of linguistic structures 由at are neither wholly arbitrary nor fully predic阳.ble. Motivation is also seen as a matter of degree. Cf. Langacker (1 987: 48) and Lakoff (1987: 346 and 493) , who speak oflevels ofpredictability and relative motivation leading to res位icted predictions, respectively. Secondly, linguistic structures seem to be chiefly motivated by an interplay of external factors such as cognitive structures and communicative needs. As Lakoff(1987: 539) puts it: People seem to leam and remember highly motivated expressions better than unmotivated expressions. We thus hypothesize that the degree of motivation of a grammatical system is a measure of the cognitive efficiency of that system relative to 也e concepts the system expresses. There 町e，
of course, scholars who disagree with this position in some respects. Bybee (1985) is thus of the opinion 由at， when it comes to grammaticalization processes, only cognitive processes can have motivating force. Hopper and Traugott (1993: 67) concur with 也is position and suggest 由at communication strategies "世aw upon general cognitive processes."
326 Rita Brdar-Szabó and Mario Brdar
In the present paper we shall show how the amount of apparent idiosyncracies in the complementation system of adjectives can be reduced if a basic cognitive process such as metonymy is assumed to link alternative valency frames of adjectival predicates. We shall also be able to show that these same metonymic processes can be invoked in some cases to motivate the form of the complements of predicative adjectives. The choice of prepositions after predicative adjectives in the following examples may at first blush appear quite arbitrary: (7)
a. At least, I've been open about it. b. The witness wasn 't veηd笔finite on the point.
but we shall see below that their systemtic appearance in such contexts is motivated by an underlying metonymic model. However, cognitive linguists have always been aware that the whole story cannot be this simple, for at least two reasons. For one thing , cognitive structures and communicative factors need not work in unison. They are on occasion even likely to work in quite opposite directions. The expressive power of a language , defined informally as "the collection of concepts in that conceptual system that the language can distinctively 饵' press" (Lakoff 1987: 539) , may be constrained to a degree by some requirements of communication. The principle of economy, at work in processes such as routinization and idiomatization of expressions , leads to simplicity, i.e. minimal differentiation of linguistic expressions. Languages can thus be regarded as "gigantic expression幽compressing machines" (Langacker 1977: 106). The fact that one set of these factors may gain primacy over the other in different languages in general an d/or in specific linguistic structures , may help determine the shape of smaller or larger portions of the grammatical systems involved. In this paper we shall examine the interplay of metonymic processes and topicalization as a case in point. The other element that complicates the pic阳re is the source of crosslinguistic differences. Cognitive as well as functionallinguists seem to have concentrated so 岛r primarily on cross-linguistic similarities. If both cognitive structures and communicative needs are assumed to be universal and more or less shared by humans , e.g. metonymy and metaphor indeed seem to be universal phenomena, we should expect human languages to be, if not the same, then at least extremely similar. However, while some linguistic phenomena are ubiquitous , many others are specific to only some lan-
Predicative adjectives and grammatical-relational polysemy 327
guages. Obviously, the picture painted by the simple interplay of cognitive structures and communicative needs lacks certain crucial details, some of which may be supplied by detailed cross-linguistic comparisons. The magnitude of cross-linguistic differences cannot be fu11y motivated just by reference to these two sets of factors. In 由e present paper, we show that whether a given language employs metonymic processes in the formation of altemative predicate 町伊ment structures in order to achieve specific communicative goals may also have to do with how other areas of the language in question are structured, i. e. with the shape of its cuηent grammar (Mi也iun 1991: 160). 四is is also in line with Lakoffs (1987: 537!) characterization of motivation in terms of, among other things, global ecological location within a grammatical system. In this paper we sha11 specifica11y point out the role of some morphosyntactic factors in constraining 由e application of metonymic processes. This means 由at the present paper will be dealing with motivation in grammar in several senses and at several levels. To sum up , we shall attempt to isolate factors that shape the complementation system of adjectives in an intralingual perspective by motivating links between altemative valency 企'ames， and by showing that the 岛rm of complements need not be tota11y arbi饲厅， and fina11y we show how cross-linguistic differences 血at 町'e at first blush qui始 bewildering nevertheless make sense.
3. 00 metooymy As we have at位ibuted p缸t of the ability of adjectival predic创es to become grammatica11y polysemous, i.e. to assume altemative/additional valence 仕ames， to some underlying metonymic models , it appears necessary to review some basic 位aditional insights into metonymy and compare them with some more recent suggestions. A fairly 丘'equent way of defining metonymy has been to contrast it with metaphor (and occasionally with synecdoche) and focus on two central points of difference, although it has been repeatedly claimed that the borderline between the two is blurred (cf. Barcelona 2000, Ruiz de Mendoza 2000, and Radden 2002). It is widely accepted that meton归ny is based on contiguity, whereas metaphor is based on similarity (cf. U11mann 1962: 212, Taylor 1989: 122). Contiguity is taken in its broadest sense to cover a11 associative relations except simil缸ity， and may 由us be too vague a notion, making metonymy almost a cognitive wastebasket. Panther and
328 Rita Brdar-Szabó and Mario Brdar Thornburg (2002: 282) a忧empt to constrain the scope of metonymy, proposing that the relation between the metonymic source and the target is contingent, i.e. it is not concep阳ally necessary. When a nurse , for example , refers to a patient as the ulcer in room 506 , it is not conceptually necessary for the ulcer to belong to the patient in room 506. Metonymic relations are thus in principle defeasible or cancelable, because the source concept is still usually retrievable (though backgrounded), even if the target concept is conventionalized in the lexicon. The other important point of contrast concerns the number of conceptual domains involved. Unlike metaphoric mapping , which takes place across two discrete domains , metonymy is "a cognitive process in which one conceptual entity, the vehi c1 e, provides mental access to another conceptual entity , the target, within the same domain , or ICM" (cf. Kövecses and Radden 1998: 39). This is illustrated by a simple example such as: (8)
阳ne experts say the bubbly w il/ being opened.
like命 go
flat in minutes
q斤er
Here the property of being bubbly is such an important feature of champagne that it makes it sufficiently different from other types of wine. The concep阳al domain within which the mapping takes place is explicitly named in the nominal expression wine experts. However, in the case of metonymy, it is important to bear in mind that the single concep阳al domain involved is structured by an ICM (Lakoff 1987: 288) and may exhibit some internal complexity. It is thus possible for metonymic mapping to occur within a single domain matrix which involves a number of subdomains (cf. Croft 1993: 348). In other words , metonymic mapping across different domains within a single domain matrix , involving the concep阳al effect of domain highlighting , is also possible. According to Radden and Kδvecses (1999) , there are three general types ofmetonymic mappings. In two ofthese, the whole ICM and its part(s) are involved , i. e. part-for-whole and whole-for-part metonymies. In the third type parts of an ICM are involved so that a part stands for another part. Ruiz de Mendoza (2000) and Ruiz de Mendoza and Díez de Valesco (This Volume) , however, suggest that these can be reduced to just two types: either the source of the metonymic mapping is contained in the target (source-in-target metonymy) or the target is in the source (target-in-source metonymy).
Predicative adjectives and grammatical-relational polysemy 329
In terms ofits 缸nction， metonymy is seen 副 "natural1y suited 岛r 岛cus sing" (Lakoff and Johnson 1980: 37ft). It maximizes economy (because no new expressions are created or needed), but it also enhances expressiveness because it can provide very direct access to that p町t of the whole we are focussing on. The point may again be illus回ted by (8): bubbly is an al- . ready existing adjectival expression, but its conversion to a noun results in a ve叩 effective way of referring to champagne as a specific type of sp町 klingwine. Although metonymy is without doubt a universal cognitive process, it may work differently in different languages and cultures. As Lakoff (1 987: 78) wams while discussing metonymy, "... such general principles are not the same in alllanguages , one cannot simply say that any也ing can stand for any由ing else in the right context. One needs to distin伊ish which principles work for such languages."
4. Grammar and polysemy We would like to hypothesize here that the contrasts that we are about to present can be at least in p缸t linked wi由 the variable propensity of the languages in question 岛r grammatical-relational polysemy. By ‘ grammatical-relational polysemy' we mean the occurrence of several different argument structures linked with a single predicative expression and presenting basically the same state of affairs from different points of view: (9)
a. 1 don 't have any pictures of my mother, and it is difficult for me to remember what she looked like. b. Th ese German words will, ther，φre， be d，伊icult for an English 再peaker to leam to use. c. For that reason, losing my audiencefor such a long time here has been ve，η， difficultfor me.
difference in the set of arguments can materialize as a formal one, i. e. the altemating 仕ames can have a different number of arguments and/or be morphosyntactically different (e.g. be introduced by different prepositions, occur in different case-forms, come in different order, etc.). 四is aspect of the phenomenon has been recognized and referred to as polyadicity within 也e framework of Lexical-Functional Grammar (cf. Bresnan 1982). The difference can also be a qualitative one, i. e. arguments in altemating 企ames 白lis
330 Rita Brdar-Szabó and Mario Brdar
can be of different kind in terms of their semantic roles. The two types of differences can , of course , co-occur. It will be seen that the existence of altemative frames in effect means that one and the same predicative expression can partake of different construction types. The reasons why we consider this phenomenon to be a special instance of polysemy are threefold. Firstly, if we adopt the more or less standard Construction Grammar view that morphemes and lexical items can also be treated as constructions sui generis , then there is every reason to consider our predicative a司jectives to be constructions as well. Part of their collocational or combinatorial meaning is the specification of the range of wider constructions in which they can be embedded. Predicative expressions appearing in a broader range of constructions are consequently richer in grammatical meaning , and thus grammatically more polysemous. Just as the appropriate interpretation of a polysemous item can be enforced or brought about by another item entering a construction with it that exhibits certain semantic features (in the sense of traditional selectional restrictions) , e.g. when rough is combined with nouns denoting a location such as town or area , its sense of involving violence and crime becomes focused , we see that the selection of a particular frame may be licensed by a given argument, its morphosyntactic form or semantic type , and vice versa. In both cases , an element' s entering into a construction results in restricting its meaning , in the former example it is the lexical meaning of the premodi马Ting adjective , in the latter it is the predicative expression's grammatical meaning. Secondly, the existence of a1temative argument structures with a single predicative expression may be accompanied by some more or less subtle shi丘s in the meaning of the predicative expression , i.e. such a1temation may correlate with lexical polysemy proper. Discussing the ubiquity of polysemy, Langacker (1 995: 22) points out "that the senses of a relational expression 0白en differ in the choice of focal participants (' argument structure') , even within the same conceptual base." Thirdly, the phenomenon we are concemed with here could be seen as a type of constructional polysemy. The tenηhas， of course, been used in a different way by Goldberg (cf. 1995: 3 月， where a given construction type , e.g. ditransitive , is associated with a set 0
Predicative adjectives and grammatical-relational polysemy 331 The phenomenon we study here is in a sense the mirror-image of the Goldberg-type constructional polysemy. While her constructional polysemy presupposes a top-down perspective, our grammatical polysemy is rather a bottom-up phenomenon.. The former focuses on constructions: it starts from a construction type and looks at its more or less formally diver- . gent variants, as well as at the range of lexical items that appear in these variants. The latter focuses on a lexical item with a predicative function , say difficult, and looks at various constructions it can partake of (the impersonal construction with or without extraposition of c1 ausal subject, toughconstruction, etc.). The linguistic and cognitive reality and robustness of the grammatical 可pe of polysemy is borne out, among other things, by the existence of syntactic blends in the sense of Bolinger (1961). Deviant conS位uction blends such as: (10) a. Well, 1 think, regardless ofthe margin ofthe election, which was, as 1 indicated, settled quite some time ago, the president is well aware, as we 've just talked previously, about the difficulties in some parts of our American community. [CNN, Special Event, Wh ite House Press Secretary Ari Fleischer Holds Daily Press Briefing, February 26, 2001] b. This eliminates frequent retyping, minimizes inadvertent t)伊ing errors, and enables users who are unfamiliar with a program to play back sets of instructions pre-recorded by someone more adept with the application. [Encarta Encyclopedia Deluxe 2000] arise due to the parallel off-line presence and availability of various valency frames , and their occasional simultaneous on-line activation. In (10a) the speaker produces at one point the predicative a司jective aware , but before its complement PP is provided, there is a parenthetic comment c1 ause containing the verb talk. By the time the speaker again picks up the main line of thought, the valency frame of talk, requiring the prepositional object introduced by about, is activated, and due to the proximity principle about is selected instead of oJ, resulting in a syntactic blend of two distinct constructions.
332 Rita Brdar-Szabó and Mario Brdar 5. A cross-linguistic perspective on metonymically based grammaticalrelational polysemy Let us now examine the phenomenon of grammatical-relational polysemy from a cross-linguistic point of view. We first consider some fairly simple a司jectival predicative constructions that exhibit what has been termed semantic diversity of subjects. Semantic diversity of subjects is for us a convenient1y broad concept, encompassing two tight1 y related phenomena that have to do with qualitative aspects of valence. First, there is the range of semantic or thematic roles found as subjects with certain predicates. Secondly, we extend the term diversity to cover the selectional restrictions that hold between predicates and their arguments. In this part we shall first consider locative subjects with meteorological predicates , and then move to experiencer subjects , and finally tum to constructions with possessor subjects and some constructions where predicates refer to the manner of an activity. These two last construction types exhibit, in addition to the somewhat unusual distribution of semantic roles assigned to the subjects , some expansion of the predicative complex by complements. More or less the same holds for constructions with predicative adjectives complemented by infinitive clauses in so-called raising constructions. As we go along we shall try to show that these cases of semantic diversity of the grammatical relation subject in clauses with adjectival predicates are based on metonymic processes. At the same time we shall show that there are significant cross-linguistic differences that will later be shown to be tallying with the distribution of raising structures with adjectival predicates. Table 1 gives an overview of the types of constructions to be discussed in Section 5.1 - 5.5 , as well as the type of metonymic mapping and the specific metonymy involved. Where two types of metonymic mappings are listed, one occurs in the subject (having to do with the semantic diversity of subjects), and the other in the predicative expression (accommodating the subject NP in terms of selectional restrictions). These construction types are illustrated by the sentences below. (11)
a. b. c. d.
Chicago was rainy yesterday. Jane is cold. The old man is hard ofhearing. He was not very clear about the consequences.
(5.1) (5.2) (5.3) (5 .4)
Predicative adjectives and grammatical-relational polysemy 333
(5.5) (5.5)
e. He is certain to win. f. Th e book is difficult 归 read.
Table 1. Overview of constructions exhibiting metonymic shifts that are discussed in Section 5
Construction type
Typeof metonymíc
Specific metonymies
Section
旦纽旦旦
Setting (locative) subject construction
WHOLEFOR PART
Experiencer subject + PART FOR adjective ~~{~
predicati~e
cons位uctíon
(a) COMPLEX ENTITY FOR ITS LOCATION (b) PERMANENT STATE FORTEMPORARYSTATE
5.1
PERMANENTSTATEFOR TEMPORARY STATE
5.2
(a) PARTFOR (a) POSSESSOR FOR POSSESSED PART 牛e+a硝 3 dje蚓 ctive 阳 poωsse伽(份 b盼)P础 ARTFO 侃 R (b) ATTRIBUTEOFPART slve cons住uct1OII WHOLE FOR ATTRlBUTE OF WHOLE Be + adjective manner PART FOR MANNER FOR ACTIVITY cons位uctíon WHOLE SALIENT PARTICIPANT PARTFOR Raising constructions FOR WHOLE EVENT WHOLE
5.1.
5.3
5.4 5.5
Setting (locative) subject construction
One of the several groups of English predicates that permit se忧ing subjects, more precisely, locative expressions, to surface as subjects comprises meteorological predicates, i.e. adjectives following linking verbs. We note the parallel existence of impersonal sentences with dummy-it subjects in which location-denoting nouns are introduced by prepositions, thus functioning as place adverbials. (12) a. London wasfoggy today. b. It was foggy in London today.
334 Rita Brdar-Szabó and Mario Brdar (13)
a. The Square was baking hot. b. It was baking hot in the Square.
We normally expect the nouns that surface as subjects here , unlike corresponding adverbials , not only to refer just to some locations affected by particular meteorological conditions (or exhibiting a given type of atmosphere in general) named by the predicative adjective , but rather to multifaceted geo-cultural entities. Such nouns are often subjected to metonymic shifts again narrowing down their reference to a single facet , e.g. to people, or an institution located at the place in question. In (1 2a) and (1 3 时， however, the metonymic shift goes into another direction , towards a reading under which these entities are stripped of all other a抽出utes to mere physical locations open to certain meteorological influences (COMPLEX ENTITY FOR ITS LOCATION). Note that the sentences in (1 2b) and (13b) are impersonal and that the meteorological conditions denoted in the predicate are in fact their topics ，时， in Langacker' s terms (1 987 , 1991), their trajectors. In (12a) and (13a) we note a shift in focal prominence from the condition to just one salient participant, i.e. to the locations at which the conditions obtam. At a more abstract level , we may note another metonymic shift (which may in this specific instance be provisionally glossed as PERMANENT STATE FOR TEMPORARY STATE) here that is related to the fact that (1 2a) and (13a) are potentially ambiguous. Parallel to the interpretation we have so far concentrated on , we observe that these utterances can also receive the plain ascriptive interpretation , which is usually associated with the construction in question. In other words , temporaη， states are presented here in terms of 臼irly stable or pe口nanent ones. This type of locative subject is at best a marginal phenomenon in German. The most usual way of rendering the English examples in German is to use the impersonal construction with es as a formal subject, and a locative adverbial which may be fronted , thus approaching the topical status of the English subjectivized locative. The acceptability of (1 4的， on the other hand, is rather questionable: (1 4)
a. In Berlin wird es wieder In Berlin will it again b. ?Berlin ist zur Zeit kalt Berlin is currently cold
ziemlich heiß sein. rather hot be und regnerisch. and rainy
Predicative adjectives and grammatical-relational polysemy 335
One of the possible reasons for a more restricted use of this structural pattem in German (and in Croatian and Hungarian, too) might be a lower number of meteorological adjectives , since some concepts are expressed by verbs rather than by adjectives (cf. E. freezing , adj. vs. G. frieren , v.; E. thunder, v. , thundery , adj. vs. G. es donnert; v.; E. snow, V. , snOl1岁， adj. vs. G. es schneit, v. , etc.). There are admittedly word-for-word equivalents for some of the above English examples, of course without prepositionally-marked time adverbials , but German native-speakers interpret them as descriptive statements referring to inherent qualities of places, (14b) thus not being synonymous with (14c): (14) c. Berlin Berlin
ist is
kalt. cold
A similar situation obtains in Croatian and Hungarian, too. Non-impersonal sentences with meteorological a司jectives tend to get a descriptive reading. As in German, Croatian prepositionally marked locatives (15) , and Hungarian adposition-marked locatives (16) may be 企onted: (15) a. Bilo je strasno vruée COP AUX terribly hot b. Na trgu strasno Je on square COP terribly c. *?Trgje strasno vruc. square COP terribly hot (16) a. Nagyon meleg very hot b. A téren D~F square-on c. *?A tér DEF square
5.2.
na tl宫u. square on vruce. hot
téren. volt a COP DEF square-on nagyon meleg volt. very hot COP nagyon meleg volt. very hot COP
Experiencer subject + predicative adjective construction
With certain English a司jectival predicates expressing some physical and emotional states or conditions such as hot, cold, sad, happy， 卢rious， glad, alarmed, amazed, amused, annoyed, depressed, etc. the subject can be
336 Rita Brdar-Szabó and Mario Brdar
identified in semantic terms as experiencer or conceptualizer. These experiencers are nowadays almost regularly coded as subjects in English: (17)
1 am
hungrylcold伊eezing/hoti4ψ'aid.
It is interesting to point out that 01d and Early Middle English relied more on their transparent coding of experiencers by means of case-marking. It is contrastively significant that a situation almost identical with that characteristic of older stages of English is found in some cases in the other three languages we are concemed with. As for metonymic shifts that can be observed here, we again note that temporary states are presented here in terms of fai r1y stable or permanent ones. This type of metonymic mapping, PERMANENT STATE FOR TEMPORARY STATE , also accounts for potential ambiguities (e.g. 1 am hot) and subt1 e changes in meaning that go hand in hand with different construals. The construal typical of Old and Early Middle English , as well of the other three languages under study, can be said to give more prominence to external influence on the experiencer, and it appears that it is only the experiencer's body, chiefly its peripheral parts , that are most affected. When the experiencer is presented as subject, the state resulting from the influence appears to be more stable (notice th创 it is presented as if it were permanent), whi1 e the experiencer appears to be more affected (as a whole , i.e. both body and mind). Although this construal is also common in the other three languages , there are certain c1 asses of predicates which prefer the transparent strategy of experiencer marking and take obliquely, i.e. non-nominatively coded phrases. There are thus some a司jectival predicates in German that allow only oblique experiencers: (1 8)
Es tut
mzr weh/ Mir tut es weh/ 乒Jch tue me-DAT ache me-DAT does it ache I-NOM do weh. ache
It does ihm him
kalt/heiß/ Mir (1 9) a. Es ist mir It is me-DAT coldlhot me-DAT b. ~Ich bin kalt/heiß. am coldlhot
ist (es) is it
kalt/heiß. cold/hot
Predicative a，咿'ectives and grammatical-relational polysemy 337 (20) a. Dieses Haus glφllt mirl Mir g，φ llt dieses this house likes me-DAT me-DAT likes this b. *Ich g，φ lle dieses Haus. 1 like this house-ACC
Haus. house
Unlike English hot and cold, which allow for a certain degree of 创nbiguity， German kalt and heiß in (19b) do not refer ωthe tempora可 physical condition ofthe subject referent, but rather to a more or less permanent quality. Their subject should not be regarded as an experiencer but as an entity characterized by the quality denoted by the adjectival predicates. Croatian and Hungarian, naturally, have nominatively coded experiencers with some predicate classes, as in (21a) and (22时， respectively, but, on the whole，也ey allow even less morphosyntactic collapsing of experiencers with agents onto a common grammatical relation and surface 岛rm. Both languages have numerous impersonal constructions with experiencers in the dative or accusative (in Croatian), in the dative (in Hungarian) , and finite verb forms in the 3rd person singular. Note 也at copulas 町e followed by predicative adverbs in (21 b) in Croatian, and in (22e) in Hungarian. Elsewhere the predicative expression is a noun following a copula or just a verb, e.g. in (21 哟， (21 e) and in (22d). (21)
a. Ivan je gladanl Ona Ivan-NOM COP hungry she-NOM
Meni
je
se
smrzaval
REFL 丘eezes
zima.
me-DAT COP cold b. Hladnol toplol vruéellosel dobrolteskoldrago cold warm hot il1 well hard glad c. ~Hladanltopaolvruél losl dobarlteiakl drag cold warm hot bad good heavy dear
d. Boli
mel
hurt-3SG me-ACC
e. Stidl sram shame shame ml
mi
me-DAT COP
sam. COP
Strah me
jel
fe町 me-ACC
COP fear-1SG REFL
me
jel
Stidim
Strasim se. sel
me-ACC COP shame-1SG REFL
}e.
me-ACC COP (22) a. Éhesl hungry
szomjas vagyokl Félekl thirsty
je.
COP
fear-1SG
Fázok. 仕'eeze-1SG
Muka
i11
338 Rüa Brdar-Szabó and Mario Brdar b. Hideg vagyok. cold am-lSG c. (Nekem) tetszik ez a me-DAT like-3SG this DEF d. Melegem van/ Nekem hot-POSS-lSG is me-DAT Fφ
ház. house-NOM melegem van/ hot-POSS:lSG is
nekem.
hurt-3SG me-DAT e. Rosszul vagyok/ Jól ill-ADVam-lSG well-ADV
vagyok. am-lSG
We note the absence of notoriously ambiguous constructions with predicates ofthe cold/hot-type , even ifthe experiencer is in the nominative case, as in the last two examples in Hungarian: the predicate is clearly adverbial (it ends in -1 viz. -ul, one of the suffixes used to derive adverbs from a司)ec tives) , and the potential ambiguity with Rossz vagyok ‘ 1 am bad/ naughty' or Jó vagyok ‘ 1 am good' , where the predicates are adjectival does not obtain. Croatian adjectives , when used predicatively in personal constructions , agree with their subjects. It appears that the contrasts between the four languages are less sharp in this area. All the languages exhibit personal constructions with experiencer subjects (based on metonymy) , but they are far more common in English than in the other three languages. Such less conspicuous contrasts could perhaps be linked to the quite general nature ofthe metonymy involved.
5.3.
Be + adjective possessive construction
Among adjectival predicates in English that exhibit semantically unusual subjects we also find some peculiar possessive constructions: ì
a. Seeing how sel严d矿òrmed he was, fat but elegant; short of leg and ham, on platform shoes ... b. The family of gods, all quite short in the legs and long in the trunk, wαs very tolerant about these abuses. c. He was so amiable, reddish ofhair, and rudl命 ofskin，... d. He was already dry ofthroat and hot of eye. e. The most numerous were the tall stately merchantmen, broad in the beam and with capacious holds, driven by sails.
Predicative adjectives and grammatical-relational polysemy 339
f. Owing ω the vision of the pink light 1 was firm 'O f purp'Ose... ... and because 1 am hard 'Ofhearing on the right side... At first blush, these l'O'Ok just like 'Ordinary c'Opular c'Onstructi 'Ons with predicative adjectives f'Oll'Owed by prep'Ositi'Onal c 'Omplements intr'Oduced by of 'Or 切. H 'Owever, a cl'Oser l'O'Ok reveals that these c'Onstructi'Ons 缸e peculiar in 由at the prep'Ositi'Onal phrases in questi'On 町'e very p 'O'Or candidates f'Or c'Omplements 'Of adjectives and, even m 'Ore significantly，也at the adjectival predicates refer here in fact t'O s'Ome qualities that characterize NPs that 町'e preceded by 由e prep'Ositi'On and n 'Ot directly their subject NPs. Subject NPs 町'e p 'Ossess'Ors 'O f what is den'Oted in the PP.ηle p 'O ssessed items are m'Ore 'Or less inalienable, either c'Oncrete n 'Ouns referring t 'O b 'Ody p缸ts， 'Or p缸ts 'O f certain 'Objects，。民 on the 'Other hand, abstract nouns den 'Oting perceptual abilities, such as senses, or some inherent and fairly stable aspects 'Of one's personality. The predicate 也at primarily characterizes 也e p 'Ossessed underg'O es a met'Onymic shift 由at may be termed Aπ阳B盯'E OF PART FOR ATTRIBUTE OF WHOLE (an instance of a m 'Ore general met'Onymic m'Odel acc'Ording to which a p缸t stands for a whole) and is now applied to the p 'Ossessor with which it may have been semantically more 'Or less incompatible. Cf. the case of heavy, which can be predicated of animate subjects, but then has a meaning utterly different 企om the one intended in the foll 'Owing examples: (24) a. Tomorrow, heavy 'Ofheart, 1 was going to the Prado... At first he was rather heavy of expression and 1 thought he might be lookingfor trouble,... Interestingly, the active zone, i.e. the intended target of the met'Onymic shift, is obligat'Orily specified by means of prepositi'Onal phrases. Otherwise, ambiguity, or even obscurity 'Of expression, might ensue. A similar phenomenon 'Of specificati'On 'Of active zones by means of nounincorporation is discussed in Tuggy (1 986). It is tempting t'O think of this specification 'Of the active z'One as a result of an interplay 'Of metonymy with another basic conceptual process, i. e. as a result 'Of blending an ascriptive constructi'On wi由 another marginal possessive constructi'On (e.g. to be of no value， ω be ofimportance, etc.). This would then m 'Otivate the ch'O ice 'Ofthe most frequent prep'Osition, i. e. oJ, but it would require a careful examinati'On 'O fthe interplay 'O fblending with metonymy，岛r which more synchr'Onic as
340 Rita Brdar-Szabó and Mario Brdar well as diachronic data are needed (e.g. the historical development of the two constructions should be plotted against each other). The other three languages we contrast here with English exhibit largely uniform behaviour in this respect: there are only a few instances of comparable constructions with adjectival predicates and possessor subjects apart from some idiomatic constructions that are extremely restricted in terms of their distribution. German also boasts a handful of possessive constructions with abstract possessed entities in addition to the few with body parts. Croatian (26) and Hungarian (27) data are similar. Again , the specification of the active zone is a rule: (25)
a. Leicht im light in b. Das Jacketf DET jacket
Geschmack. taste ist in den Schultern COP in DET shoulders
(26)
Ivan je brzna jeziku. Ivan COP quick on tongue
(27)
A
eng/weit. narrow/wide
ruha széles vállban. DET dress wide shoulder-in
Further, there are either predicative expresssions with the copula verb followed by a partitive expression (Croatian example (28)) , comparable to the be 01 X possessive construction in English , or morphologically complex predicative adjectives (Hungarian example (29)) comparable to English a司jectives ofthe type blue-eyed:
(28)
Ivan je Ivan COP
siroke ruke. broad-GEN hand-GEN
(29)
János bõkezü. János broad-handed
In sum, German , Croatian and Hungarian may exhibit some such marginal constructions but they do not readily allow this type of metonymy with a司jectival predicates and , consequently , there is less grammatical-relational polysemy here because the adjectives in question normally do not acquire additional valence 仕ames.
Predicative adjectives and grammatical-relational polysemy 341
5.4.
Be + adjective manner constructions
The last type of adjectival predicate constructions in English exhibiting similar metonymy effects to be discussed in this section are cases where manner (or circumstance) stands for action. Predicative adjectives refer, here to the manner in which various activities referred to explicitly in PP complements of adjectives , but more frequently only inferable 仕om the PP, are performed. For example, the adjective clear, as in: (30)
... one should be as clear as possible about historical facts.
may refer to the manner of the understood cognitive activi守 or to the manner of some understood linguistic activity such 邸，岛r example , speaking. These actions range 仕om rather physical ones to cognitive and linguistic actions (but these are frequently extremely difficult to keep apart, as illustrated in the following set of examples): (31) a. b. c. d. e. f. g.
Arthur was brief about his other teachers in his recollections. Sheila wasn 't very definite on the point. Yes - and be direct about the 笔fJect of his work. At least, I've been open about it. The Italian was energetic in examining his rings, ... Be firm about turning down invitations ifbeating deadlines. The English aristocracy was always ruthless in letting its members gently decline.
The MANNER FOR LINGUISTIC ACTION 可pe of metonymy, exemplified in the first few examples in (31) , has been discussed in detail in Brdar & Brdar-Szabó (2003). It will be seen that subjects ofthese predicative adjectives are typically animate. The predicative adjectives are not always inherently compatible with such subjects since they primarily refer to aspects of activities. In other words , these conventionalized metonymies result in some additional polysemy of the predicative expressions in questíon. Most ofthe a司jectives allowing MANNER FOR LINGUISTIC ACTION type of metonymy take prepositional complements introduced by about, occasionally on, with or in. The choice of the most frequent prepositions , about, and on , is motivated by the fact that the most prominent verbs of linguistic
342 Rita Brdar-Szabó and Mario Brdar action , such as speak and ωlk， also take prepositional complements of the same form. The other three languages do not allow such predicative a司jectives 缸' panded by prepositional phrases implying active zones , but opt for more transparent coding of these events , where active zones are made quite explicit. There are either predicative adjectives followed by adverbial subordinate clauses containing the verb referring to the activity whose manner of performance is denoted by the adjective , or the a司jective is rendered as an adverb modifying the verb , which unlike in English , now cannot remain implicit (cf. Brdar & Brdar-Szabó 2003). This is illustrated in the Croatian counte叩arts of some of the English examples above:
训 E
咄咄 N
m俨G
ω由
。
DA
0 ，胁
旷
5.5.
vcn .现 -o
。由
(32) a. Arthur nije bio kratak kad se prisjeéao Arthur NEG-AUX been brief when REFL reminisced drugih nastavnika. other teachers b. Sheila nije bila vrlo odreâenakada je 0 tome Sheila NEG-AUX been very definite when AUX about that govorila. spoke -ud vc ovd am Gnr u muh ue k1 nd c. 1 da - govori 4··· o u and yes speak djela. work-GEN vn obLK ov'-t ru-03 nuok OO e mM d. Konαcno sam otvoreno 0 u6··finally AUX openly
Raising constructions
Now that we have shown that there are significant differences between the four languages conceming the semantic diversity of subjects with adjectival predicates , we tum to some infinitival complement clauses with a司jectival matrix predicates that have traditionally been referred to as raising and pseudo-raising constructions. English raising constructions are discussed in great detail in Langacker (1 995) , and the present section will build on many of his insights. Although two types of constructions , subject-to-subjectraising and tough-movement , occur with a司jectival predicates, for reasons of space , we shall concentrate chiefly on subject-to-subject-raising.
Predicative adjectives and grammatical-relational polysemy 343
The metonymic processes that we can observe with raising structures are analyzed by Langacker (1995)ωcases of active-zone/profile discrepancy, i.e. 副 specific instances of more general reference-point phenomena. What happens here 扭曲at one of the participants of the embedded event (either functioning as its subject or object argument) is picked out and presented in such a way 也at the adjectival predic创.e， expressing the narrator's perspective, point of view, judgement or opinion, is now predicated of this designated participant. In short, one part(icipant of the event) stands for the whole (eventlstate ofaffairs) as the subject ofthe a司jective. With both types of raising, we note the existence of some core constructions, whose predicates can also appe缸 in impersonal constructions, and a cluster of related constructions e对libiting only some behavioural 位aits of the core constructions. It will be seen that one of these, a pseudosubject-to-subject-raising, provides us with an interesting link between, on the one hand, simple ascriptive constructions discussed above , and prototypical raising constructions, on the other hand. Most of the English adjectival predicates in subject-to-subject-raising constructions express modal concepts such as potentiality and certainty, e.g. likely, sure, certain , able, apt, bound, etc.: (33) a. … with the operation, which was sure ω stir new trouble in Lebanon. b. ... whose editor is certain ω reject it because the subject matter does notfall within thejoumal's scope. adjectives 缸'e
generally available in simple ascriptive constructions 臼 well， but they then require that their subjects refer to an event, at least implicit坊， whose likelihood or certainty is asserted or denied. In raising constructions, on the other hand, these predicative adjectives may have apparently non-event-denoting subjects as well 也at are logically p町tici pants of the embedded events. The metonymic process enabling the shift of perspective, i. e. its 岛cussing from the whole event to a salient participant, (and which may be best termed SALIENT PARTICIPANT FOR WHOLE EVENT) may be accompanied by a subtle change in the modal meaning involved. While the modal concept expressed by the matrix adjective is still interpreted as refeηing to the embedded event, the certainty or potentiality is no longer purely objective, i. e. it is now seen as stemming in part, at least, 企om some inherent qualities of the subject, possibly involving some degree These
344 Rita Brdar-Szabó and Mario Brdar of control by the subject or a related entity. In other words , epistemic modality is now fused with an element of dynamic modality. As for the syntactic behaviour of these adjectival predicates, we note that likely , certain , and sure allow extraposed finite that-clauses and infinitive clauses with raising , but comparable impersonal constructions with a prepositional phrase preceding the infinitive clause are not acceptable:
(34)
a. It is likely that John w il/ come. b. John is likely to come. c. ~It is likely for John to come.
(35)
a. It is certain that John w il/ come. b. John is certain to come. c. *It is certain for John ω come.
Apt, able , bound and liαble do not appear in impersonal constructions with finite that-complements: (36)
a. We a陀 liable to forget. b. ~It is liable that we w il/ forget.
What these matrix predicates have in common with the previous group is that they too do not allow prepositional phrases introduced by for or of in front of infinitives: (36) c. *It is liable of!for us to forget. (37)
*It is bound ofyou份r you to make a mistake.
Turning to German translational equivalents of raising adjectives , we find that, as König observed (1 971: 68) , predicates like sicher, gew树， wahrscheinlich , etc. can occur only in impersonal constructions. Plank (1984: 353) points out that German , instead of argument-raising , rather favours the lowering of epistemic or attitudinal matrix predicates to the rank of nonsubordinating adverbs , particles or parentheticals:
(38) (39)
Bloggs is likely/c ertain to come. a. *Bloggs ist wahrscheinlichlbestimmt zu kommen. b. Bloggs wird wahrscheinlichlbestimmt kommen.
Predicative adjectives and grammatical-relational polysemy 345
Croatian does not appear to have any adjectives like likely 也at would take infinitival complement clauses in a raising environment. Closest Croatian counte甲町ts of English raising adjectives 缸e morphologically related adverbs modifying the contents of the proposition that is coded as an embedded clause in English. In the latter case we have an ex创nple ofmatrix predicate lowering comparable to that found in German: (40)
He 's sure to come.
(41)
Sigurno ée surely will
(42)
He is likely to win.
(43)
Vjerovatno ée
doéi. come
pobijediti. likely-ADV will win
In Hungarian we find that, quite in accordance with our expectations based on German and Croatian da钮， Stephanides (1980: 127f) cites cases ofHungarian sentences in which the equivalents of English raising-predicates are rendered as adverbials modi马ring， within a simple clause, the contents of what is in English expressed in the embedded clause: (40)
He 's sure ω come.
(44)
Biztosan el}δn. surely comes
We see here that German, Croatian and Hungarian fail to make use of metonymic processes and thus reduce the polysemy in this area as much as possible. Unlike English, in which predicates like likely or certain can appe町 in three valence frames , simple ascriptive , impersonal , and raising construction, the other 由ree languages have just 由e frrst two frames. Meton归ny is not invoked in these languages to derive the third type. Instead of raising constructions we find so-called predicate-lowering, characterized by the use of more or less distinct adverb forms , derived from adjectives by productive morphological processes. Pseudo-subject-to-subject raising constructions 缸e illustrated in the following set of examples:
346 Rüa Brdar-Szabó and Mario Brdar
(45) a. John was stupid to rφse that ~加r. b. It was stupid 01 John to r，φ'se that offer. (46)
John was quick ωr矿山 the offer.
The first subtype , exemplified in (45坊， differs from the core type of raising constructions in that it exhibits parallel impersonal constructions with a prepositional subject. The second subtype, illustrated in (46) , has no impersonal counte甲art. None ofthe other three languages has anything like these pseudo-raising constructions with infinitives. Most frequently , there are impersonal constructions with prepositional phrases followed by infinitives (comparable to (45b)) , or by subordinate (finite) clauses specifying the reason for ascribing the property in question to the subject: (47)
a. Es war dumm von Johann , diesen it was stupid of Johann this b. Es war dumm von Johann , dass it was stupid of Johann that Vorschlag αbgelehnt hat. suggestion rejected has
Vorschlag abzulehnen. suggestion to reject er diesen he this
(48)
a. ?Bilo je glupo od Ivana odbiti taj prijedlog. COP AUX s阳pid o fI van-GEN to reject that suggestion b. Bilo je glupo od Ivana da/sto je COP AUX stupid of Ivan-GEN that AUX odbio taj prijedlog. rejected that suggestion
The closest these languages come to pseudo-raising is in personal constructions in the matrix followed by these same subordinate (finite) clauses , which is illus位ated with a Croatian example corresponding to (45a). Note that the conjunction sto is the one that is used to introduce both nominal and adverbial clauses and thus partly corresponds to English that. It can also be expanded to zato sto (l iterally because what/thαt). (48)
c. Ivan je glup (zato) sto John COP stupid because that/what
je odbio ponudu. AUX refused offer
Predicative adjectives and grammatical-relational polysemy 347
We might assume that Ivan is the salient pa民icipant standing here for the whole situation and that the finite clause specifies the active zone. The diι ference between English and Croatian would in this case be in the form of the element speci马ring the active zone , i. e. the target of the metonymy. But since an ascriptive construction, possibly involving a metonymic shift (Ivan je glup ‘ Ivan is stupid') , is available on its own, we could also assume 由at we have a case of syntactic blending, not so much blending in Bolinger's original sense of erroneous hybrid structures, but rather as an outcome of conceptual blending (instantiated as a syntactic blend of personal and impersonal construction) teamed up with metonymy. We cannot discuss in any detail tough-constructions , but the situation that we may observe there is very similar to what we found with subject-tosubject raising, particularly the contrastive pa忧em ofpredicate-lowering, as demonstrated in Brdar and Brdar-Szabó (1992) , Brdar (1994) , and Brdar (1 996). In fact , tough-constructions are practically impossible in German, Croatian and Hungarian if there is no inherent compatibility between the subject and the predicate of the matrix clause in terms of selectional restrictions, in marked contrast to English prototypical tough-constructions. There are , however, occasionally structures in Croatian in which the matrix subject in the nominative case (the predicative a司jective agreeing with it in gender and number) is followed by a prepositional infinitive (introduced by za): (49)
a. Knjiga book b. Vijest news
je COP je COP
laka za easy for ugodna pleasant
citati. read-INF za cuti. for hear-INF
This construction seems to have evolved as a blend between prepositional phrases with deverbal nouns (laka za ci阳nje ‘ easy for reading') expressing pu甲ose and simple prepositional infinitives with modal meaning illustrated below: (50)
za vjerovati. To nije 岛r believe-INF this NEG-COP ‘It is impossible to believe that'
‘
348 Rita Brdar-Szabó and Mario Brdar Such instances of prepositional infinitives have been marked in Croatian prescriptive grammatical tradition as a substandard phenomenon (consid制 ered to be an instance ofGerman influence , but in view ofthe existence and frequency of prepositional infinitives in many Romance languages , it might also be argued that the Mediterranean Sprachbund assisted here the evolving Danubian Sprachbund) , but the cons位uction is gammg ground. However, this Croatian construction is subject to severe restrictions regarding the range of predicative expressions so that it chiefly occurs with adjectives like 1，αk( ‘ easy')， teiak ('difficu1t'), and ugodan ('pleasant'). German exhibits some comparable cases of this construction:
(51)
Das Buch ist interessant zu lesen. DEF book COP interesting to read
Let us , however, consider now a more pseudo-tough-like German example like:
(52) a. Der Film war DEF film was b. Braune Mädels brown gir1 s c. Dein Bri写f war your letter COP
traurig anzuschauen. sad to-watch sind hübsch anzuschauen ,... COP pre忧y to-look-at rührend zu lesen. moving to read
While we could of course c1 aim that the subjectivized participant stands in these pseudo-tough-constructions for the whole event, it is telling that the predicative a司jective must be compatible with the subject, and that there obtains a meaning difference between the simple ascriptive construction and the corresponding impersonal construction. It rather appears that a shift of meaning has taken place from the meaning of rührend in the ascriptive construction towards the meaning in which the activity of reading is evaluated. The construction above seems thus to be half-way between the 阳'om that respect, which may be taken as supporting a blending approach.
6. Concluding remarks We hope to have shown that some apparent1 y arbitrary contrastive facts can be accounted for in a systematic and principled fashion by assuming an
Predicative adjectives and grammatical-relational polysemy 349
interplay of some cognitive and structural factors. We have provided evidence for our claim that certain cases of grammatical-relational polysemy of some English constructions with predicative adjectives rest on metonymic processes, and 由at these processes help shape the complementation system of a司jectives， partly by motivating the morphos严ltactic form of . complements. A comparison of these English constructions with their counte甲arts in languages like German, Croatian and Hungarian has shown that the latter languages regularly fail to tolerate this type of polysemy based on metonymy. English has been demonstrated to rely heavily on meton严nic processes in rearranging predicate argument structures enabling different construals while at the same time keeping formally one and the same form of the predicative expression. The other languages involved, which in fact make use of metonymy elsewhere, tend to formally indicate different arrangements in predicate argument structure by using formally different predicative expressions, particularly Croatian and Hungarian. This appears to go hand in hand with conspicuous differences in the kind and degree of the schematicity of the structure specifying the active zone between English, on the on hand, and the other three languages , on the other. Unlike in English, these structures are far more elaborated in the other three languages and tend to assume the form of a司juncts and not that of arguments. The results of this cross-linguistic comparison are summed up in Table 2 below. Brackets are used in the table to indicate that a given construction is in principle possible but less frequently made use of. The fact that a certain construction is available in a fairly limited range of contexts is expressed by%. Focussing on the form ofthe element speci马ring the active zone , we can present the possibilities on a continuum as follows (omitting zero specification as the limiting case, which is more or less frequently found in all languages): infinitive (53)
p町ticle
10
infinitive > complement clause
finite finite > complement > adverbial clause clause
We see that the most 企'equent specifications of active zones in English tend to be items on the le白 of the continuum, while the other three languages prefer the specification by items on the right. Notice that the Croatian examples (32a) and (32b) exhibit practically the same pa忧em as (48c). It is
350 Rita Brdar-Szabó and Mario Brdar
relevant that, since these adverbial clauses are not presented as arguments , they do not impose a new valency frame on the predicative expression in question and thus do not lead to an increase in grammatical polysemy (while they may lead to some lexical polysemy). Table 2.
Availability of types of adjectival constructions and metonymies in a cross-linguistic perspective
(a) Setting (l ocative) subject con- (b) structíon Experiencer subject + predicative adjective constructlOn Be+a司jec-
(a) tive (b) possesslve construct lOn Be + adjectIVe manner
E
户UW 『
Metonymy type(s)
u 一
tlOn 句rpe
u-G n-ob-
吃一c
Construc-
H
COMPLEX ENTITY FOR ITS LOCAτION
4
一…一
^I/
(^I/)
POSSESSOR FOI主 POSSESSED ATTRIBUTE OF PART FOR ATTRIBUTE OF WHOLE
^I/
(%) (%) (%)
MANNER FOR ACTIVITY
4
SALIENT PARTICIPANτFOR WHOLEEVENT
^I/
PERMANENTSTATEFOR τEMPORARYSTAτE
PERMANENTSτATEFOR
TEMPORARY STATE
(^I/)
(^I/)
cons位uctíon
Raising constructlOn
%
%
Our final point concems the interplay of metonymy and communicative goals. The phenomenon of grammatical polysemy may be thought of as one of the devices invoked to express pragmatic functions such as topicalizing. This goal may of course be achieved in a number of ways: a language like English can topicalize by using special topicalizing constructions, or by promoting semantically non-typical arguments to the subject position associated with topic, which results in more grammatical polysemy. On the other hand, it generally appears that languages with richer morphological systems, notably with rich case systems and productive morphological ap-
Predicative adjectives and grammatical-relational polysemy 351
paratus that may be used to derive formally distinct predicative expressions, such as German , and particularly Croatian and Hungarian , can achieve these same communicative goals in a different way. Since they tend to have more flexible word orders , they can topicalize by placing the item in question in 出e appropriate position while retaining the morphological marking. of its syntactic function , or by using an alternative derived predicate with a different array of arguments. To give an example, Hungarian and Croatian do not really need the tough-construction , since they can achieve more or less the same functional effect of topicalization by simply fronting the accusative object. This means that the 邵阳al shape of the current grammar makes it almost unnecessa可 to use certain topicalizing constructions or to promote certain arguments to the subject status. There is no need to make extensive use of alternate valency frames of predicative expressions, i.e. these languages will exhibit far less grammatical polysemy, and therefore rely less on metonymic processes. On a more general level , we hope to have demonstrated the importance of metonymy as one of the mechanisms in the management of the flow of information in discourse , i.e. enabling the topicalization of certain participants. In addition to its referential function , which is 仕'aditionally recognized by most accounts , metonymy must thus be recognized not only as having a serious impact on grammatical structures, but also as being allpervasive in the organization of linguistic structures, because we have seen it spanning a broad range of phenomena, from the more locally syntactic phenomenon of predicate-argument structure organization to the macrosyntactic and pragmatic level of discourse management, thus blurring the putative boundaries between grammar and pragmatics.
References Barcelona, Antonio 2000 On the plausibility of claiming a metonymic motivation for conceptual metaphor. In: An tonio Barcelona (ed.) , Metaphor and Metonymy at the Crossroads. A Cognitive Perspective , 31-58
(Topics in English Linguistics 30). BerlinINew York: Mouton de Gruyter.
Bolinger, Dwight 1961 Syntactic blends and other matters. Language 37 .3: 366-38 1.
352 Rita Brdar-Szabó and Mario Brdar Brdar, Mario 1994 a
11矿ìnitival
Complement Clauses and Core Grammatical Relations and Object in English , German , Croatian and Hungarian. PhD dissertation , University of Zagreb. 1994 b Object-assignment in a functiona1 grammar of Croatian revisited. In: E1isabeth Engberg-Pedersen, Lisbeth Falster-Jakobsen and Lone Schack Rasmussen (eds.) , Function and Expression in Functional Grammar, 109-126. BerI inINew York: Mouton de Gruyter. 1996 Raising cross-1inguistic dus t: What can contrastive 1inguistics do for typo10gy? Suvremena Lingvistika 41 -4 2: 63-80. Brdar, Mario and Ri ta Brdar-Szabó 1992 How tough is tough-movement to typo10gize? In: Christian Mair, and Manfred Markus (eds.) , New Departures in Contrastive Lingustics (Innsbrucker Beiträge zur Ku1turwissenschaft, Ang1istische Reihe 4-5). Vo1ume 1, 105-114. Innsbruck: Institut rur Ang1istik der Universität Innsbruck. 2003 Meton归口ic coding of 1inguistic action in Eng1ish , Croatian and Hungarian. In: K1aus-Uwe Panther and Linda Thomburg (eds.), Metonymy and Pragmatic 1nferencing, 241-266. Amsterdam/ Phi1ade1phia: John Benjamins. Bresnan, Joan 1982 Po1yadicity. In: Bresnan, Joan (ed.) , The Mental Representation of Grammatical Relations, 149一 172. Cambridge, Mass.: M.I.T. Press. Bybee, Joan L. 1985 Morphology. A Study of the Relation between Meaning and Form. AmsterdamIPhi1ade1phia: John Benjamins. Croft, William 1993 The role of domains in the interpretation of metaphors and metonymies. Cognitive Linguistics 4 .4: 335-370. Curme , G.O. 1931 A Grammar of the English Language. Vo l. 3: Syntax. Boston: Heath. Goldberg, Adele E. 1995 Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Haiman, John 1980 The iconicity of grammar: Isomo甲hism and motivation. Language 56.3: 515-540. 1983 Ic onic and economic motivation. Language 59 .4: 781-819. Su句;ect
Predicative adjectives and grammatical-relational polysemy 353 Hawkins, John A. 1986 A Contrastive Typology of English and German: Unifying the Contrasts. London: Croom Helm. Hoppe巳 Paul J. and Elizabeth Closs Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Ivir, Vladimir 1983 A Contrastive Analysis of English Adjectivl臼 and Th eir SerboCroatian Corre.甲'ondents. Zagreb: Institute of Linguistics , University of Zagreb. Kδnig， Ekkehard 1971 Adjectival Constructions in English and German: A Contrastive Analysis. Heidelberg: Julius Groos. Kδvecses， Zoltån and Günter Radden 1998 Meωn沪ny: Developing a cognitive linguistic view. Cognitive Linguistics 9.1: 37-77. Lakoff, George 1987 Women , Fire, and Dangerous Th ings. What Categories Reveal about the Mind. ChicagolLondon: The University of Chicago Press. Lakoff, George and Mark Johnson 1980 }.命taphors We Live by. ChicagolLondon: The University of Chicago Press. Langacker, Ronald W. 1977 Syntactic reanalysis. In: Charles N. Li (ed.), Mechanisms of Syntactic Change , 57-139. Austin: University ofTexas Press. 1987 Foundations ofCognitive Grammar. Volume 1. Theoretical Prerequisites. Stanford: Stanford University Press. 1991 Foundations ofCognitive Grammar. Volume 2. Descriptive Application. Stanford: Stanford University Press. 1995 Raising and transparency. Language 71.1: 1-62. Mathesius , Vilém 1961 Obsahový rozbor soucasné anglictiny na základe obecne lingvistickém [A Functional Ana1ysis of Contemporary English on a General Linguistic Basis]. Prague. Mithun, Marianne 1991 The role of motivation in the emergence of grammatical categones: 币le grammaticization of subjects. In: Elizabeth Closs Traugott and Bemd Heine (eds.), Approaches ω Grammaticali zation. Volume 2: Focus on 巧伊es of Grammatical Markers. (Typological Studies in Language 19.2), 161-184. Amsterdam/ Philadelphia: John Benjamins.
354 Rita Brdar-Szabó and Mario Brd，αr Panther, Klaus-Uwe and Günter Radden 1999 Introduction. In: Klaus-Uwe Panther and Günter Radden (eds.) , Metonymy in Language and Thought. (Human Cognitive Processing 4) , 1-14. Amsterdam!Philadelphia: John Benjamins. Panther, Klaus-Uwe and Linda Thomburg 1999 The potentiality for actuality metonymy in English and Hungarian. In: Klaus-Uwe Panther and Günter Radden(eds.) , Metonymy in Language and Thought. (Human Cognitive Processing 4) , 333-357. Amsterdam!Philadelphia: John Benjamins. 2002 The roles of metaphor and metonymy in English -er nominals. In: René Dirven and Ralf Pörings (eds.) , Metaphor and Metonymy in Comparison and Contrast. (Cognitive Linguistics Research 20) , 279-319. BerlinINew York: Mouton de Gruyter. Plank , Frans 1984 Verbs and objects in semantic agreemen t: Minor differences between languages that might suggest a major one. Journal 01 Semantics 3: 305-360. Radden , Günter 2000 How metonymic are metaphors? In: Antonio Barcelona (ed.) , Metaphor and Metonymy at the Crossroads: A Cognitive Perspective, 93一 108. BerlinINew York: Mouton de Gruyter. Radden, Günter and Zoltán Kövecses 1999 Towards a the。可 of metonymy. In: Klaus-Uwe Panther and Günter Radden (eds.) , Metonymy in Language and Thought. (Human Cognitive Processing 4) , 17-59. Amsterdam!Philadelphia: John Benjamins. Ruiz de Mendoza Ibáñez, Francisco José 2000 The role of mappings and domains in understanding metonymy. In: Antonio Barcelona (ed.) , Metaphor and Metonymy at the Crossroads: A Cognitive Perspective. (Topics in English Linguistics 30), 109-132. BerlinINew York: Mouton de Gruyter. Stephanides, Eva H. 1980 A contrastive study of the English and the Hungarian infinitive. In: László Dezsö and William Nemser (eds.) , Studies in English and Hungarian Contrastive Linguistics, 93一 144. Budapest: Akadémiai Kiadó. Taylor, John R. 1989 Linguistic Categorization: Prototypes in Linguistic Theory. Oxford: Clarendon Press. Tuggy, David 1986 Noun incorporations in Nahuatl. Proceedings of the Annual Meeting ofthe Pacific Linguistics Conference 2: 455斗 70.
Predicative adjectives and grammatical-relational polysemy 355
Ullmann , Stephen 1962 Semantics: An Introduction to the Science 01 Meaning. Oxford: Blackwell. Vachek, Josef 1961 Some less familiar aspects of the analytical trend in English. Brno Studies in English 3: 31 -44.
Metonymy behind grammar: The motivation of the seemingly "irregular" grammatical behavior of English paragon names Antonio Barce/ona
Abstract 1 The paper argues for the metonymic motivation of the grammar of paragon names. Paragons are well-known individual members of a category that represent an ideal of the category. Names mentioning paragons regularly violate the standard grammatical cons仕aints on names (inability to take the plural, etc.). Two metonymies motivate paragons and the behavior ofparagon names. First, a social metonymic model of the paragon yields the common property shared by all category members. Second, the paragon metonymically stands for the category, directly motivating the grammatical behavior of the name (as in That young man is a real Shake.再peare i.e. a gre创 playwright).
Keywords: categories, constraints on names , grammar, mapping , meton-
ymy , origin of names , paragon name , proper name, schematic metonymy, stereotype
1. 1 am grateful to the editors and reviewers of this volume for their useful suggestions. The research reported was partially funded by the Secretaría de Estado de Universidades. Investigación y Desarrollo of Spain (State Secretary for Universities, Research and Development), Project no. PB-98-0375.
358 Antonio Barcelona 1. Introduction The motivation of the original meaning of names is normally metonymic. Names often originate historically in definite descriptive noun phrases (see Lyons 1977: 179). These phrases have specific reference and they include , as their main descriptor, a circumstance closely connected with their referent, or an aspect of this referent. 2 For instance, place names typically originate in descriptive noun phrases whose descriptor mentions a nearby geo2. An exhaustive, systematic classification of reference types has proved extremely difficult in standard semantic theory. Therefore , 1 must clarify my use ofthe terminology in this area. Lyons (1977: 177-197), in his discussion ofreι erence, distinguished "singular definite reference" (reference to a uniquely identifiable referent, achieved by singular definite noun phrases) , "distributive and collective general reference" (reference to each member of a set or to the set as a whole) , "specific and nonspecific indefinite reference" (the former , when the speaker of a sentence such as Every morning at six 0 'cZock a heron flies over the chalet can be taken to refer to a particular, though unidentified, referent; the latter, when the speaker cannot be taken to refer to one particular referent) , and "generic reference" (reference to a class in a generic proposition; Lyons suggested several possible subtypes of it). These distinctions presuppose the distinction between singular and plural definite NPs , indefinite NPs (which are either an indefmite pronoun or are introduced by an indefinite article) and what Lyons called "non definite NPs" (he gave no examples of the latter, but he said 也at they are noun phrases that are non-definite , that all indefinite NPs are non definite , but that the converse is not necessarily true). Other semantic textbooks , like Saeed (1 997) , provide no systematic classification of reference types. Two standard grammars ofEnglish , Quirk et al (1985: 265-287) and Quirk and Greenbaum (1 990: 78-93) distinguish "specific reference" (reference to one or a set of particular referents , apparently subsuming Lyons's "singular definite reference", and his two types of "general reference") from "generic reference". They also distinguish "definite reference" (the referent(s) is (缸e) uniquely identifiable) from "indefinite reference" (the referent(s) is (are) not uniquely identifiable). The classification proposed by Quirk et al., though not 企ee 企om disadvantages , is sufficient for my purposes. It implies that definite and indefinite reference (not necessarily achieved by NPs introduced by definite or indefinite determiners) are subclasses of specific reference , in the sense used by these gramrnars. After all , according to Lyons , it is quite doubtful that "nonspecific indefinite reference" should be regarded as a type of reference at all.
Metonymy behind grammar 359
graphical feature: thus the place name Merseyside refers to the region surrounding the river Mersey, and the place name 0矿ord derives from a 岛rd where oxen crossed a river. The descriptor may also mention a related historical or political circumstance, as in Jerez de la Frontera ("Jerez on the Frontier") , one of the many Spanish towns bearing 由at prepositional phrase as part of their name because they once were on the borderline between Christian Castile and the Moorish kingdoms. The descriptor of the original noun phrase may also make reference to people somehow related to the place (Washington) , to some outstanding buildings and architectural features revealing its original function (Chester, from Latin castra ‘ (military) C创np')，3 or to many other possible related circumstances. Surnames often originate in the male parent's name (Johnson ‘ John's son' , Sánchez ‘ Sancho's son'). The names ofmonths and days originate in names of divinities, kings, st町s， etc. , somehow associated to them in a given culture: January (ultimately from Latin Januarius mensis ‘ Janus' month') , March (企om Martius mensis ‘ Mars' month') , Sunday (from OE sunnon daeg ‘ sun's day') , Monday (企'Om O.E. monan daeg ‘ moon's day'). Thousands of similar examples could easily be added. Thus names often seem to originate as descriptions of a given entity on the basis of a different but related entity or circumstance. These 可pes of description constitute s位aightforward instances of metonymy, as we shall see presently. They eventually become, in the course of history, rigid designators with unique reference, i.e. they become real names. This semantic fact constrains their grammatical behavior to a greater or lesser extent in many languages, English among them. The standard morpho-syntactic constraints in English on names that consist of a single proper noun such as Paris prevent their combination with plural morphemes (1 时， determiners (lb) and restrictive modifiers (lc): (1)
a. *Parises b. *the Paris, *a Paris, *some Paris c. ?1 like the beautiful Paris, but not the ugly one.
3. The Old English word ceaster (which leads to Modem English chester and which originates in Latin castra) , was used by Old English speakers, in an early meωn归回c extension, to designate any enclosed space intended for habitation, and was attached to many other place names like Colchester, Gloucester, Dorchester, etc. See Baugh (1959: 93).
360 Antonio Barcelona
Similar constraints apply to complex names consisting of a determiner and a proper noun such as The Hague:
(2)
a. *The Hagues b. *Hague, *a Th e Hague, *some (The) Hague c. *Ilove the modern The Hague, and hate the old The Hague.
These morpho-syntactic res位ictions of proper nouns are due to the fact that names have unique reference. These restrictions do , however, not apply when names exceptionally are used as common nouns. Standard grammars , such as Quirk et a l. (1985: 288-297), Quirk and Greenbaum (1 990: 87-88) or Jespersen (1949: 42 6--429 , 488-4 91 , 544-580), devote plenty ofspace to the numerous cases in which the above constraints are violated. Quirk et al. (1985) and Quirk and Greenbaum (1990) distinguish five types of names which can be reclassified as common nouns (see Table 1): Table 1.
Types of names undergoing conversion into common nouns. Source: Quirk et al. (1985) and Quirk and Greenbaum (1990)
(a)
Different referents with the same name: Th ere is a Richmond in the south 01 England and a Richmond in the north, not to mention a dozen Richmonds outside the British Isles (p laces with the same name).
(b)
Different referents with the same name which also constitute a conceptual class: She always spends her Mondays 1 Octobers 1 Chris伽ases here.
(c)
Uses due to "informal conventions": Ma 缸rr 咀ied couples and 岛 fh 泊m 油回 1咀ili 世臼 es 缸: The W协 1 Use of a famous name ωmean the type 由 tha 创t made it famous: There were no Shakespeares in the ninenteenth century [p araphrased by Quirk and Greenbaum as "[...] writers who towered over contemporaries as William Shakespeare did over his," and by Quirk et al. as "authors like Shakespeare"].
(ωc-斗1)
(ωc-2)
(d) Names subject to modification: (d-l) Uses due to "partitive restrictive modification": 1及e young Joyce already showed signs 01 the genius that was ω be卢伊lled in Ulysses. (d-2) Uses due to nonrestrictive modification: Poor old Mrs Fletcher 1 The fondly remembered John F. Kennedy (e)
Miscellaneos: A Rolls Royce 11页is museum has several Renoirs. 1 a pack 01 Cheste巧fìelds.
Metonymy behind grammar 361
In this paper, 1 will concentrate on one ofthese 守pes ofname use (type c-2 in .Table 1) and will claim in Section 3 (where the corresponding names will be called "paragon n缸nes"， or simply "paragons") that metonymy h臼 a crucial role in the motivation of this use. But, before presenting and discussing 也is claim, 1 find it necess缸Y to briefly characterize the notion of metonymy assumed in this paper. This is done in Section 2.
2. Metonymy as a cognitive phenomenon The concept of metonymy assumed in the present paper is similar given by Kövecses and Radden (1998: 39):
to 也at
Metonymy is a cognitive process in which one concep阳al entity, the vehicle, provides mental accessωanother conceptual entity, the 饱rget， within the same domain, or ICM. In Barcelona (in press) 1 proposed a general notion of meton严ny， which 1 call schematic metonymy. A schematic metonymy is a meton严ny that satisfies the minimal requirements 岛r every conceptual metonymy, irrespective of the neat distinctiveness of source vis-à-vis target, and irrespective of the referential or nonreferential nature of the metonymy. These minimal requirements 町e: intradomain mapping and activation of a target by a source. Every metonymy involves a schematic metonymy.4 In order to explore the grammatical relevance of metonymy 1 will make use of 由is notion of metonymy, since it is broad enough to cap阳re any meton严nic mapping which motivates , or interacts wi曲， grammatical structure. There岛re， 1 will be using the term metonymy in the present article as equivalent to schematic metonymy, which is defined as follows:
4. In Barcelona (in press , 2002) , 1 propose and discuss at
leng世1 two additional notions of metonymy, progressively more constrained, which, however, presuppose the schematic notion:ηpical metonymies are schematic meton归ùes whose target is neatly distinct from the source (distinctiveness being a ma忧.erof degree, and measured in terms of a set of factors that cannot be dealt with here); in She ￥ just a pretty face , FACE is the source and the target is PERSON; notice 也剑出.e NP where the meton严ny occurs is not referential, but predicative. Proωtypical meωn沪回es 缸'e typical (hence, also schematic) metonymies with a referential pu叩ose， and with individuals as targets and as referents: My Picasso is not庐r sale.
362 Antonio Barce/ona
A schematic metonymy is a mapping, within the same overall cognitive domain, of one cognitive (sub)domain, called the source, onto another cognitive (sub)domain, called the t缸get， so 出at the latter is mentally activated. This definition assumes 也at metonymy is fundamentally a "reference point" phenomenon (Langacker 1993: 29-35). It also inc1udes the c1 aim that metonymy is a conceptual projection or mapping (see Lakoff and Tumer 1989: 103). 1 prefer using the notion of domain rather than 由at of conceptual entity as done by Kövecses and Radden (1998) , because even "entities" are understood in terms of complex domain matrices (Langacker 1987: Chapter 4). Moreover, when metonymy is involved in category structure, it is complex categories, rather than individual entities, that are source and target，邸， for example, in the stereotype-based metonymic models of "bachelorhood" or "motherhood" (Lakoff 1987: Chapter 8). However, there is no harm in talking about metonymic sources or targets being entities in cases in which they are c1 early unitary physical objects (inc1uding people),5 or their c1 early delineated parts. 6 The other differences 町e simply a matter of personal preference. 7 Given this notion of metonymy, it should be easy to understand why names are c1 aimed to be normally meton严nic in origin: Within an overall 5. For instance, PRESIDENT (a person with a role) FOR ARMY, as in Nixon bombed Hanoi.
6. E.g. SALIENT BODY PART FOR PERSON , as in We need strong arms. 7. Unfortunately, both definitions raise the further problem of deciding when and on the basis of which criteria the two entities or (sub )domains are included in the same overall domain. The "within one domain" relationship or mapping is supposed to be the property that distinguishes metonymy 企om metaphor (which implies a mapping between two domains "located" in separate overall domains).ηlis issue is discussed in Barcelona (2002, n.d. b), where 1 suggest that the overall domain in which meton严ny occurs is a "functional domain" (a "企ame" or "ICM" , as Kövecses and Radden' definition quoted above also suggests) , where the roles of the source and the 饱rget are linked by a "pragmatic function" (like AUTHOR-WORK, PRODUCER-PRODUCT, etc.). The domains linked by metaphor are not connected by a pragmatic function. In Barcelona (in press and n.d. 时， 1 suggest that an additional criterion (which further research may find to be more reliable) to distinguish metaphor 企om metonymy is the systematic matching 01 counterparts (i.e. the systematicity of submappings) encountered in metaphor, versus the absence of such systematic counte甲art matching mmetonymy.
Metonymy behind grammar 363 domain，也e
referent ofthe name is a metonymic target (sub)domain which is mentally activated by a source (sub)domain consisting of a circumstance or an aspect closely connected with 世le referent. For example, wi也in the overall domain of PLACES, the referent of the name Q矿ord is, from a historical point of view, the metonymic target domain which was mentally activated by a source domain, namely the region where oxen used to cross the ford of a river, around which the city seems to have grown. Similarly, in the FAMILY domain, the name of the male parent (source subdomain) can be mapped onto his children (target subdomain) as a way to concep阳alize and refer to the latter，也is mapping being manifested, among other 也ings， in the conventions of family naming in numerous cultures (Johnson , Sánchez, Peters). Metonymy thus accounts 岛r the development of names from descriptive phrases. As shown by the ex缸nples in Table 1, names may also be used as common nouns. 1 argue in the following section that this shi缸， in the case ofparagon names, is also fundamentally motivated by metonymy.
3. Metonymic analysis of the grammar of paragon names A paragon is an individual member or a set of individual members of a category "who represent either an ideal or its opposite" (Lakoff 1987: 87). As an ideal member of a category, a paragon may metonymically stand for the category 臼 a whole. Lakoff gives the example of the baseball player Babe Ru曲， who may as a paragon stand 岛r 也e category of great baseballplayers. Of the uses identified by the standard grantrnars mentioned above , 由e ''use of a famous name to mean the type 也at made it famous" (Table 1, c-2) is a paragon use of a name. As Lakoff says, "paragons are made use of in constructions in the language: a regular ßabe Ruth, another Willie Mays, the Cadillac ofvacuum cleaners , etc." The use ofparagon names in some such constructions is illustrated in the following sentences, where the p缸a gon name Shake，再peare is used as a common noun 扭曲e singular (3) and plural (4) , preceded by a determiner (3) - (6) and, optionally, by a modifier (3) , (6):
(3) (4) (5) (6)
Lope de Vega was not the 牛'anish Shake.甲eare. There were some Shake.吃peares in the 阳lentieth century. Lope de Vega was not a Shake.再peare. Lope de Vega was not a real Shake，再peare.
364 Antonio Barce/ona
As a common noun , the paragon name Shake.再peare can be pluralized as in (4) , although there is only one authentic person of this name , and can have reference as in (3) and (4) or lack reference in predicate position as in (5) and (6). Grammatically, paragon nouns thus seem to allow the whole range of structural possibilities as common nouns. Semantically, however, they are different from common nouns. In particular, paragon nouns do not denote the standard referent of that name: thus the paragon noun Shake.再peare does not denote the playwright Shakespeare but a class of individuals having a characteristic property associated with Shakespeare. The special paragon sense of the name Shake.再peare can be accounted for by the following conceptual factors: (a) A conceptual model ofthe standard referent ofthe paragon name , i.e. Shakespeare the writer, which presents Shakespeare as a writer endowed with immense literary talent. This model of Shakespeare is a stereotypical model of this individual, and, as a stereotypical model , it is a metonymic mode 1. 8 (b) The process consisting of the mental creation of a class of individuals characterized by one or more of the relations and properties imported from conceptual model (a). In this case , the common relation in which all the members of the class participate is HAVING IMMENSE LITERARY TALENT. This class includes one (p ossibly more) ideal members , as Shakespeare is socially seen as an ideal - a paragon - for the class of immensely talented writers. (c) A metonymy (a conceptual process) which connects model (a) as the source domain to class (b) as the target domain. This metonymy arises on the basis of the fact that Shakespeare is socially regarded as an ideal for the class and that membership in the class depends on how close a particular writer comes to the ideal of immense literary talent set by Shakespeare. The preceding claim is represented graphically in Figure 1. Box (a) represents the mapping of Shakespeare as endowed with his characteristic properties onto the rest of our knowledge network about him. This mapping results in a stereotypical model of Shakespeare as primarily, in fact exclusively, a writer with immense literary talen t.
8. This use of the term stereotype is justified below, as is my claim that a stereotypical model involves a metonymic model.
Metonymy behind grammar 365 (a)
S队KESPE战ENETWORK
biography
\•
PTcüon /
IMMENSE LITERARY TALENT (CHARACTERISTIC PROPERTY)
actor
.
/•
playwright
•
\ portralts , image
j
Metonymy
(b) CLASS OF WRITERS WITH IMMENSE LITERARY TALENT Tarso de Molina
/
\ SHAKESPEARE NETWORK
Lopez de Vega
司
'
/
George Bernhard Shaw
\
(a) Stereotypical conceptual network associated with Shakespeare. Stereotype arising by virtue of the metonyrny CHARACTER1 STIC PROPERTIES FOR INDIVIDUAL (b) Figurative class of Shakespeare-style writers with unrivalled talent. Organized by a stereotypical model by virtue of the meton严ny STEREOTYPICAL MEMBER / SUBCATEGORY FOR CATEGORY. F部lre
1. Factors in the grammar of paragon names
366 Antonio Barcelona
Box (b) represents the mapping of this stereo叩pical view of Shakespeare onto the class of writers with immense literary talent. Both are metonymic mappings and are symbolized by the arrows. Both mappings are connected by the fact that the concep阳al model of Shakespeare resulting 仕om the first metonymy is the source in the second: The line connecting the big box in (a) with the small box in (b) indicates that the stereotypical model of Shakespeare, which highlights the characteristic salient relation "Shakespeare having immense literary talent", corresponds to the source in the metonymy, which creates a metonymic model of the mental class WRITERS WITH IMMENSE LITERAY TALENT.
Thus , the combination of processes (a) , (b) and (c) motivates the existence of paragon names and their grammatical behavior as common nouns. In the rest of this section 1 will a忧empt to justify this claim. The existence of paragons depends , first of all , on the conceptual network associated with the individual that names the paragon. In the example under discussion, this conceptual network is constituted by Shakespeare's known biographical data, his literary production and its reception , his activity as a playwrlght and actor, his well-known portraits and images, etc. However, what is paramount in our common knowledge about William Shakespeare, even for many people that have never read his works , is his IMMENSE LITERARY TALENT.
Most people have a stereotypical understanding of Shakespeare almost exclusively as an enormously talented writer, even as the best writer of his time. For these people, this is Shakespeare's characteristic property, which distinguishes him 仕om other writers. 1 claim that this is a stereotypical model of Shakespeare because it backgrounds other facets of the author which may or may not be known to these people, such as the fact that he was a married man and a father , that he was an actor, that he was educated at Stratford grammar school, that he retired relatively early to live on the money he had earned through his theatrical activity, etc. Some readers may be surprised to find the term stereotype or social stereotype applied to models for individuals. Lakoff (1987: 85-86) applied the term "social stereotype" to subcategories standing for whole categories (e.g. the housewife mother is a 可pe of mother that stereo可pically stands or used to stand? - for the whole mother category). However, it is not unreasonable to posit th
Metonymy behind grammar 367
and relations and a network of subdomains, some of which may be more representative than others. If the housewife-mother subcategory may socially stand for the whole mother category, then there is nothing inherently wrong in suggesting that the conceptual subdomain SHAKESPEARE AS AN IMMENSELY TALENTED WRITER (constructed on the basis ofhis characteristic property) may stand for our whole knowledge network for Shakespeare. 9 We constantly use these (often inaccurate and unfair) stereo可pes for individuals. For instance, if a person stutters noticeably or gossips a lot, these properties will often prevail in people's views of that person and in the language used in talking about him or her, irrespective of the fact that s/h e may also be extremely intelligent and a warm, good-natured person. Similarly, writers that first became famous after publishing book X are often angered when they find themselves repeatedly referred to as "the author of X" , as if they had written nothing else of value after that book. Many other examples could be adduced. Social stereotypes often arise on the basis of a conceptual metonymy, since stereotypes are a type of metonymic model , as Lakoff (1987: 85) points out. He also claims in the same place that most metonymic models are models for individuals , so stereotypes for individuals should not be rare. The stereotypical model of the individual Shakespeare is also created by a concep阳al metonymy. This metonymy will be called CHARACTERISTIC PROPERTY OF AN INDIVIDUAL FOR THE INDIVIDUAL , which is a PART FOR WHOLE metonymy. It is thus not simply a property (IMMENSE LITERARY TALENT) that gets mapped onto the whole of the conceptual model for Shakespeare, but the relation involving this property and Shakespeare. That is , the source in the concep阳al metonymy creating this metonymic model is the relation linking SHAKESPEARE to the property (HAVINo) IMMENSE 9. By the way, the term socia/ in the expression socia/ stereotype does not mean that the stereotype has to apply to a social group (as a reviewer of this paper has pointed out): When Lakoff (1 987: 85) talks of "social stereotypes" he refers to a stereotypical model of a category that arises 切 socie秽， not to stereotypical views restricted to categories necessarily consisting of social groups. For instance, lots of people have a stereotypical view of the car category MERCEDES (not a social group) as a highly expensive and luxurious type of c缸， on the basis ofthe upper segment ofthe Mercedes range, even though we know that other Mercedes cars are not expensive or luxurious.
368 Antonio Barcelona
LITERARY TALENT; and the target is the whole knowledge network about Shakespeare. This mapping simplifies the network, as it gets virtually reduced to this relation , hiding other relations (such as the f民t that he was an actor, etc) which would have equal status in a non-metonymic model of Shakespeare. 10 A common prope此y or a common relation is always the potential basis for a mental class, i.e. the class constituted by entities linked to the same prope此y in the same relation. In this case , Shakespeare's stereotypical model is used as the basis for setting up a mental class of individuals who are characterized, according to the speaker, by displaying the same property, that is , by participating in the same type of relation , which might be called X HAVING IMMENSE LITERARY TALENT. This mental class is the class of"writers with immense literary talent". But this mental class is itself normally understood in terms of an ideal model , in this case by a paragon , i. e. Shakespeare. His characteristic salient relation both underlies the mental class and makes Shakespeare the ideal model or paragon for the class , to which he , furthermore , lends his name. This name thus becomes a class-name, i.e. a common noun. This understanding of the class is conceptually possible thanks to another PART FOR WHOLE metonymic mapping of the stereotypical model of Shakespeare onto the whole class of writers with immense literary talent lO.In earlier papers (Barcelona n.d. a,), 1 have called this meton严ny DEFINING PROPERTY FOR INDIVIDUAL , using the term d，听ning as equivalent to "characteristic" , "quintessential". This label of the metonymy is less accurate than the one now proposed, for two reasons. First, because the source in the metonymy is not just a property, but a relation involving the property, as stated above. Second, because, as a reviewer pointed out, only common nouns , not proper nouns , can be defined (in a strict sense of d，听ne and d，听ning); therefore , defining has been replaced by characteristic, which is a less problematic adjectlve. The same reviewer suggested salient as an altemative to d，听ning. But not all salient facets in people's knowledge about Shakespeare are characteristic as well. The cI aim made by certain literary historians that Shakespeare had homosexual tendencies has reached the general pub Iic. This is a salient relation in their knowledge about Shakespeare, but not a characteristic one , and this is shown by the fact that it is not seized upon toωm Shakespeare into a homosexuality paragon: John is a Shakespeare is normally interpreted as meaning that John is an extremely talented writer, not as meaning that John is a perfect homosexuaI.
Metonymy behind grammar 369
(IDEAL MEMBER FOR CLASS). This is the metonymy that directly motivates , at the syntagmatic level , the use ofShakespeare's name as a common noun. It is important to insist upon the idea that a prerequisite for the application of this metonymy in this case is the creation of the mental class of immensely talented writers (b in the figure) , and that a prerequisite for the creation of this class is the existence of the culturally entrenched meton严nic model of Shakespeare (a in the figure). The metonymic model of Shakespeare is stored "paradigmatically" in the speaker's network of cognitive models. It seems to operate at the purely concep阳al， prelinguistic level , to make the existence of paragons possible , and it does not "show up" directly in linguistic behavior. Only the mental class and its paragon-based metonymic model seem to have overt grammatical manifestation, as we will show beloW. 11
11. Kövecses and Radden say that the phenomena studied in this paper are based on just one metonymy, which they call CATEGORY FOR DEFINING PROPERTY, whereby well-known individuals "are metonymically recategorized as a class on the basis of their defining, stereotypical property. Thus, in calling a person a Judas , we are describing him or her as ‘ treacherous' , and in referring to an upcoming star in linguistics as a second Chomsky, we have in mind his or her intellectual brilliance." (Kövecses and Radden 1998: 54) In my view, they oversimpli马r the complex conceptual connections operating in these cases. Their ve叩 description of the functioning of this metonymy makes it clear that the well-known individual in question is recategorized as a class on the basis of his defining property. But this can only be done by first stereotypically (hence metonymically) mapping this defining property (or rather, this characteristic relation) onto the individual (and, in a way, "downplaying" all ofhis / her other properties). This is done by what 1 call CHARACTERISTIC RELATION FOR INDIVIDUAL. The stereotypical model of Judas , like that of Shakespeare, is stored paradigmatically in people's "knowledge base". When the same relation is discovered in other people, a mental class is created. Only then does that relation d e become a defining property for the mental class (in the Judas example, the 由 fining 町 0 r characteristic prope 町 r飞吵 y would be "t位 reacherous member of a group") thi岱s class, it can act as a metonymic Since Judas is one of the social ideals for 由 reference point for the whole class (IDEAL MEMBER FOR CLASS); as 1 said, it is this meton严ny that directly licenses, at the syntagmatic level , sentences like He was a Judas (i. e. ‘ he was a treacherous member of a group'). But the target of this metonymy is not the defming or characteristic property "treacherous member ofa group" itself, but the c1ass of"treacherous members ofa group", which includes that property.
370 Antonio Barce/ona
The grammatical behavior of paragon names can now be explained quite easily. Once we have a mental class such as that of immensely talented writers , it is possible to count its members , which is evidenced in the possibility of pluralizing the noun describing the class as in We have had many Shakespeares in the past century. We can also make reference to some of them by means of determination - as shown by the use of articles and other determiners , as in (3) 一 (6) 一， by the use of pre-head (either modi马ring or determining) genitive NPs , as in (7) , (7)
Cervantes
is 牛ain
's Shakespeare.
or by a combination of determination and modification , as in (3) and (6). Of course, a paragon name can appear in NPs with generic reference to the whole class: (8)
A real Shakespeare would never use those trite images.
As a class-label, a paragon name can also be used in predicate nonreferential NPs , as in (5) and (6). In (6) and (町， the degree of fit between the referent ofthe NP and the paragon is measured by means ofa hedge like real. In its role as a paragon, the individual referent "Shakespeare" is not really split up into a number of entities to construct the mental class (i. e. Shakespeare is not figuratively split up into a class of "Shakespeares" embodying several aspects of Shakespeare), but mapped , as a model , onto a class, of which he is the most prominent member, and to which he lends his name. In other words , in a sentence like (9) 。)
There are three real Shake.司peares in my college.
the NP three real Shake再peares refers to three different individuals , not to the same individual from three different perspectives. This is an important difference from "partitive restrictive modification".12
12. See paragraph d-1 in Table 1. In The young Joyce a/rea命 showed signs of the genius that was ω be fulfilled in the Joyce that wrote U/ysses , the speaker / writer does not refer to two different people in the mental space of reality, but to the same person viewed from two different perspectives, or involved in two different situations. Joyce is counterfactually, and figuratively , "split" into a series of inviduals with the same name , on the basis of a number of relations in
Metonymy behind grammar 371
An interesting aspect of the application of the IDEAL MEMBER FOR CLASS metonymy is that the target (the figurative class) can be construed more or less rigidly. The use ofhedges like real in ex创nples (6) , (8) and (9) seems to attest to a rigid construal of the class. In other words , saying that X is a Shakespeare, can mean , or be interpreted to mean , either that X is an immensely talented writer, or that X is an immensely talented writer with some of the specific features that characterized Shakespeare as a writer, i.e. not just an immensely talented writer, but a writer who "towered" over his contemporaries as Shakespeare did. This more rigid scope is apparently represented by the different paraphrases offered by Quirk and Greenbaum and by Quirk et al. (see Table 1, c-2). The more rigid construal invokes a special subclass of the mental class, for which Shakespeare is the only possible paragon. This subclass is constituted only by "writers with as immense a literary talent as Shakespeare's". If the construal of the categ。可 does not invoke this special subclass , then saying that X is a Shakespeare means , or is interpreted as meaning, simply that X has immense literary talent. In this case, Shakespeare would be just one of the various conventional paragons for the whole class of immenseIy talented writers , together with Cervantes, Dante, Virgil , etc. Finally, another interesting aspect of paragons is that they can be embedded in metaphorical mappings , as in ex臼nples like Michael Jordan is the Shakespeare 01 basketball. Here , the metonymically understood domain of immensely talented writers is metaphorically mapped onto the domain of immensely talented basketball players.
4. Conclusion Metonymy has been shown to be operative in the creation of names. In this 町ticle， meton严ny has been argued to be crucially involved in the conceptual motivation for the seemingly irregular grammatical uses of English
which the real Joyce was involved. A detailed discussion of this phenomenon was originalIy included in the present paper, but had to be left out, given the s位ict space limitation set by the editors. For an initial exploration, see Barcelona (n.d. a).
372 Antonio Barcelona
paragon names , consisting of a transient re c1 assification \3 of these names as common nouns. At a purely conceptual level , a metonymy creates a stereotype of the individual acting as paragon. This stereotypical model highlights the paragon's characteristic relation and property. On the basis of this relation, a mental c1 ass is set up with those individuals participating in a similar relation with the same characteristic property. And a second metonymy maps the paragon onto the class, which licenses the use of the paragon name as a c1 ass name, hence as a common noun. One of the most interesting aspects of this research is that it provides evidence for the c1aim 也at names are not merely unique designators of entities whose grammatical behavior remains unaffected by the concep阳al networks associated with their referents. Quite the opposite, the grammatical behavior of names is constantly govemed by our rich knowledge network about their referents , as Christine Michaux has c1 aimed in a recent artic1 e (Michaux, 1998).14
References Barcelona, Antonio 2002 Clarifying and applying the notions of metaphor and metonymy within cognitive linguistics: An update. In: René Dirven and Ralf Poerings (eds.) , Metaphor and Metonymy in Comparison and Contrast, 207-277. Berlin: Mouton de Gruyter.
13. The reclassification of paragon names as class names may become permanent in time , and be registered in standard dictionaries: a Galahad is "any man regarded as very pure and noble" , a class ofmen whose paragon is Galahad, the knight in late Arthurian legend characterized by his purity and nobility of spirit (Webster's New Twentieth Century Dictionary, Unabridged, 2nd edition). 15. The conceptual motivation for all of these uses of names can probably be represented in much greater detail in terms of Turner and Fauconnier's "conceptual integration" model , also known as "blending" (see, e.g. Fauconnier 1997). In paragon names we seem to have a blend between Shakespeare the writer and the particular writers included in the mental class. However, the goal of this paper was simply to demonstrate the metonymic motivation of these uses , rather than to r写present in a detailed elaborate model the web of connections resulting 企om these special uses. This latter task is in itself a little piece of research which awaits the work ofblending specialists.
Metonymy behind grammar 373 2003
On the dominant notion of metonymy in cognitive linguistics: An analysis and a few modest proposals. In: Hubert Cuyckens , Thomas Berg, René Dirven and Klaus-Uwe Panther (ed时， Motivation in Language: Studies In Honor 01 Günter Radden , 223255. Amsterdam: John Benjamins. La gramática de los nombres propios: un viaje con retomo (por n.d. a buenos motivos). (Unpublished manuscript.). Plenary lecture presented at the 11 Intemational Conference of the Spanish Cognitive Linguistics Association , Madrid, Spain , May 17, 2000. n.d. b The difference between metaphor and metonymy: A question of asymme位y? (Unpublished manuscript.). Paper presented at the Fourth Conference on Researching and Applying Metaphor. Tunis, 5-7 April 2001. Baugh, Albert C. 1959 A History 01 the English Language. Second Edition. London: Routledge and Kegan Paul Ltd. Fauconnier, Gilles 1997 Mappings in Th ought and Language. Cambridge: Cambridge University Press. Jespersen, Otto 1949 A Modern English Grammar on Historical Principles. Part VII: Syntax. (Completed and edited by Niels Haislund.) London: George Allen and Unwin Ltd and Copenhagen: Ejnar Munksg臼rd.
Kövecses, Zoltán and Günter Radden 1998 Metonymy: Developing a cognitive linguistic view. Cognitive Linguistics 9: 37-77. Lakoff, George 1987 Women , Fire and Dangerous Things: What Categories Reveal about the Mind. Chicago: Chicago University Press. Lakoff, George and Mark Turner 1989 More than Cool Reason: A Field Guide to Poetic Metaphor. Chicago: University of Chicago Press. Langacker, Ronald W. 1987 Foundations 01 Cognitive Grammar. Vol. 1: Theoretical Prerequisites. Stanford: Stanford University Press. Lyons, John 1997 Semantics. Vo l. 1. Cambridge: Cambridge University Press. Michaux, Christine 1998 Dénominations et répresentations conceptuelles. In: José Luis Cifuentes (ed.) , Estudios de ling说ística cognitiva 11, 755-765. Alicante: University of Alicante, Spain.
374 Antonio Barcelona Quirk, Randolph and Si也ey Greenbaum
The Student's Grammar 01 the English Language. London: Longman. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar 01 the English Language. London: Longman. Saeed, John 1. 1997 Semantics.Oxfo时， Cambridge， Mass.: Blackwells. 1990
Åuthor index
Abraham, Roy C. 197, 214 Abrah缸n， Wemer 87 , 89 , 93 , 98 Akmajian, Adrian 25 , 42 Alm- Arvius, Christina 197, 214 Anderson, John M. 126, 152 Anderson, John R. 273 , 290 Antis, S.M. 234, 246 Arnheim, Rudolf 234, 243 Ball, T. 236, 245 Barcelona, Antonio 30, 41 , 294, 317, 319, 320, 327, 351 , 354, 357, 361 , 362, 368, 371 , 372 Barlow, Michael 79 , 122, 152 Barsalou, Lawrence W. 236, 243 Basson, Joey 273 , 290 Baugh, Albert C. 359, 373 Bebr, lrmtraud 100 Beitel, Dinara A. 2, 27, 42 , 188, 215 Bereiter, Carl 273 , 290 Berg, Thomas 11 , 42 , 43 , 373 Bergen, B. K. 221 , 240 , 243 Be而ne胁， Pier M. 125-127, 129, 152 Besnier, Nico 76, 98 Black, John B. 247 Black, Tamara 236 , 247 Blake, Renée 32, 43 , 217 Blansitt, Edward L., Jr. 126, 152 Bloomfield, Leonard 158, 188 Bolinger, Dwight 20, 43 , 131 , 132, 152, 331 , 347, 351 Boroditsky, Lera 222, 243 Botha, Rudi P. 270, 272 , 281 , 282 , 284, 290 Bouman, A.C. 271 , 290 Bransford, J.D. 273 , 290
Brdar, Mario 30, 41 , 321 , 341 , 342, 347, 352 Br，也r-Szabó， Rita 30, 41 , 321 , 341 , 342, 347, 352 Bresnan, Joan 329, 352 Brugman, Caudia 159, 160, 188, 189 B创ller， Karl 76 , 98 Bybee, Joan L. 106， 117， 118 ， 122一 124, 12ι128， 152, 163 , 189, 200, 206 , 214, 222 , 244, 317, 318, 325 , 352 Cafti , Claudia 77 , 98 Campe, Petra 91-93 , 98 Casati, Roberto 234 , 244 Chang, N.C. 240, 243 Chemicky, Brian 234 , 246 Chomsky, Noam 158, 189, 292 , 369 Clark, Eve 173 , 189 Clark, Herbert H. 221 , 222 , 235 , 241 , 242 , 244 , 315 , 317, 318 Claudi , Ulrike 8, 9, 43 , 44, 106, 119, 153 , 190, 215 , 222 , 245 Cocude, M. 236 , 244 Cohen, David 118, 120 Comrie, Bemard 123 , 124, 126, 143 , 152 Cooper, William E. 33 , 43 Craig, Colle忧eG. 204, 214 Croft, William 8, 18 , 33 , 43 , 57, 72 , 105, 106, 118, 159, 189, 191 , 241 , 294, 318, 328, 352 Cruse, Alan D. 159, 189 Crystal, David 30, 43 Curme, G.O. 323 , 352 Cutler, Ann e 61 , 72
376
Author index
Cuyckens , Hubert 373
42， 43 ， 188一 190，
Dahl, Östen 125 , 126, 128 , 152, 153 Davies, W i1liam D. 212 , 215 Davis , Philip W. 212 , 215 De Schutter, G. 87, 88, 98 Delfitto , Denis 125 , 152 Demers , Richard A. 25 , 42 Den Dikken, Marcel 87, 89, 98 Denis , Michel 152, 236 , 244 Díez Velasco, 01ga 1. 30, 40 , 293 , 320 Dik, Simon 318 Dirven, René 43 , 73 , 99 , 167, 189, 250, 251 , 257 , 258 , 265-267 , 294, 318-320, 354, 372 , 373 Dixon, Robert M.W. 210 , 215 DuBois , John W. 118 Durie , Mark 212 , 215 Ebert, Karen H. 144, 152, 153 Emanatian, Michele 222 , 244 Englebert, Annick 90, 98 Evans , Vyvyan 2， 21 ， 27， 37， 157一 160, 163 , 166一 168 ， 187, 189, 192 Everaert, Martin 邸， 87， 98 Fauconnier, Gi1les 89 , 98 , 294 , 300, 304, 306, 313 , 318 , 372 , 373 Fi1lmore, Charles 294 , 318 Fleischman, Suz缸me 163 , 189 , 318 Foley, W i1liam A. 204 , 215 Foolen, Ad 19, 26 , 35 , 56, 72 , 75 , 77 , 99 Franks, J. 1. 273 , 290 Freeman, N. 183 , 189, 191 Geeraerts , Dirk 2, 32, 43 , 272 , 290 Geiger, Ri chard A. 73 , 194, 215 , 217
Giacalone Ramat, Ann a 106, 119 Gibbs, Raymond W. 2, 27 , 42 , 188 , 190, 205 , 215 , 221-223 , 233 , 244, 295 Giv侃， Talmy 105 , 106, 119 Glenberg, Arthur M. 236 , 244 Godfrey, John 122 , 153 Goldberg, Adele E. 10, 43 , 56, 57, 61 , 73 , 83 , 99 , 225 , 244 , 265 , 266 , 273 , 275 , 290 , 330, 352 Goldin-Meadow , Susan 244 Goossens , Louis 217 , 294 , 295 , 318 Grady, Joseph 27 , 43 , 175 , 177, 190 Green, Melanie 166, 189 Greenbaum, Sidney 154, 191 , 358 , 360, 371 , 374 Greenberg, Joseph H. 107, 119, 154, 291 Grice, H. Paul 30, 44, 252 , 265 Grondelaers , Stefan 32 , 43 Groot, Casper de 152 Güldemann, Tom 215 Gumperz, John 162 , 190, 247 Haas , Mary R. 215 Haiman, John 2, 16- 18, 33 , 44 , 105 , 106, 118-120, 325 , 352 Hardy, Heather K. 212 , 215 Hamish, Robert M. 25 , 42 Harrington, Michael 215 Haspelmath, Martin 117, 119 Haugen, Einar 128, 153 Hawkins , Bruce 167 , 190 Hawkins , John A. 14, 44 , 119, 353 Heine, Bemd 2, 8, 9, 26 , 27 , 30, 36, 43 , 44 , 103-107, 109, 111 , 113117, 119, 120, 122 , 126, 127, 153 , 163 , 190, 200 , 206-208 , 214 , 215 , 222 , 245 , 295 , 318 , 320, 353 Herskovits , Annette 159, 167 , 190 Hi1l, Clifford Alden 190
Author index Hiraga, Masako K. 2, 9, 20 , 44 , 191 Hoeks位a， Teun 11 , 44 Hoiting, Nini 241 , 247 Holisky, Dee A. 211 , 215 Holliman, Edward 153 Holyoak, Keith J. 273 , 290 Hopper, Paul J. 10, 22 , 44, 104, 106, 119, 163 , 190, 325 , 353 Horn, Lawrence R. 31 , 44 Hottenro由， Priska-Monika 167, 190 Hudson, Richard 59, 73 Hui忧， W.3 ， 44
Hünnemeyer, Friederike 8, 44, 106, 119, 153 , 190, 215 , 222 , 245 Yoshihiko 251 , 252 , 265 , 266 Ivir, Vladimir 322, 353 Ikeg缸田，
Janney, Richard W. 77 , 98 Jespersen， O忱。 131 ， 153 ， 360， 373
Johnson, Mark 27 , 135 , 153 , 172, 175 , 177, 190, 195 , 196, 215 , 216, 223 , 245 , 270, 291 , 294, 296 , 318, 319, 329, 353 , 359, 363 , 365 Johnson-Laird, Philip N. 115 , 120, 167, 191 , 222 , 234, 246 , 317 Josey, Meredith 32, 43 Keegan, John M. 206, 207, 216 Keller, Rudi 270 , 290 Kelly, S.D. 235 , 244 Kempen, W i1lem 271 , 290 Kessakul , Ruetaivan 241 , 245 Kilian-Ha钮， Christa 215 Kimenyi, Alexandre 270, 290 Kirsh, David 234 , 245 , 246 Köhler, Oswin 207 , 208 , 216 König, Ekkehard 295 , 320, 344, 353 Kooij , Jan G. 11 , 44
377
Koops, Christian 27 , 36, 121 , 126, 153 , 206, 216 Kosslyn, Stephen 236 , 245 Kövecses, Zoltán 32 , 44, 232 , 246 , 270 , 287 , 288 , 291 , 294 , 300, 318, 320, 328, 353 , 354, 361 , 362, 369, 373 Krauss, Robert M. 235 , 245 Krych, Meredith 235 , 244 Kuteva, Tania A. 103 , 105 , 106, 116, 117, 119, 120, 122, 13~ 153 , 159, 160, 191 Labov, W i1liam 20, 32, 43 , 44 Lakoff, George 2 , 4 , 24, 27 , 28 , 45 , 67, 73 , 135 , 153 , 159, 160, 165 , 166, 175 , 177, 189, 190, 193 , 195一 197， 216， 217， 223 ， 245 ，
270 , 291 , 294 , 296 , 319, 325329, 353 , 362, 363 , 366, 367 , 373 Lambrecht, Knud 77 , 83 , 99 Lane, Jonathan 204, 205 , 217 Langacker, Ronald 5, 21 , 28 , 29, 32, 34, 45 , 50, 51 , 54, 59, 67 , 71 , 73 , 90, 91 , 99 , 125 , 153 , 159, 171 , 179, 185 , 191 , 194, 216 , 223-226 , 228 , 239 , 240, 241 , 245 , 266 , 272-275 , 277 , 281 , 285 , 286, 289 , 291 , 294, 300, 319, 325 , 326, 330, 334, 342, 343 , 353 , 362, 373 Lass, Roger 120 Lawler, John 70, 73 Leech, Geo flTey 154, 191 , 374 Lessau, Donald A. 215 Levinson, Stephan S. 31 , 45 , 247 Leys, Odo 77 , 81 , 87 , 88 , 99 Li, Charles N. 124, 153 , 353 Lichtenberk, Frantisek 210 , 211 , 216 Lindner, Susan 174, 187, 191
378
Author index
Lindstromberg , Seth 167 , 191 Ll oyd , S. 183 , 189 , 191 Lyons , John 104, 105 , 120, 358 , 373 Maglio , Paul P. 221 , 234 , 239 , 245 , 246 Malchukov , Andrej L. 95 , 96 , 99 Marillier, Jean-François 100 Marshall , C. R. 315 , 318 Martin , Samuel E. 饵， 119 ， 133 ， 154, 216 Mathesius , Vilém 323 , 353 Matlock, Teenie 29 , 38 , 67 , 190, 221 , 227 , 234 , 237 , 238 , 239 , 240 , 245 , 246 Matsumoto , Yo 224 , 226 , 246 McCarthy, J.E. 273 , 292 McCawley , James D. 88 , 99 McDaniel , Jane 153 Meillet, Antoine 25 , 45 孔1er1an， Francesca 210 , 216 Metzler, Jaqueline 236 , 247 Michaelis , Laura A. 77 , 83 , 99 , 217 Michaux , Christine 372 , 373 Miller, George A. 115 , 120, 167, 191 , 222 , 234 , 246 Milner, Jean-Claude 89 , 99 Mithun, Marianne 210 , 216 , 327 , 353 Moravcsik, Edith A. 120, 285 , 291 Murphy, Gregory L. 273 , 291 Newman , John 9 , 28 , 38 , 45 , 128 , 151 , 154 , 193 , 196, 197, 216 , 217 Newmeyer, Frederick J. 14, 45 Norv毡， Peter 197 , 217 Nunberg , Geoffrey 301 , 303 , 304 , 307 , 319 Nusbaum, H. 235 , 244
0lson , Mike 204 , 215 Onions , C. T. 17, 45 Ostergard, Svend 94 , 99 Paardekooper, P. c. 82 , 98 , 99 Pagliuca, Wiliam 106 , 117 , 118 , 119 , 120, 152 , 153 , 189 , 214 , 222 , 244 , 317 Palmer, Frank R. 210 , 217 Panther, Klaus-Uwe 1, 43 , 121 , 246 , 249 , 293 , 295 , 300, 316, 318-320 , 327 , 352 , 354 , 373 Pauwels , Paul 197, 217 Pawley, Andrew 204 , 205 , 217 Peirce , Char1 es S. 16, 45 Pérez Hemández, Lorena 319 , 320 Perkins Revere 106 , 117 , 118 , 152 , 189, 214 , 222 , 244 , 317 Pinker, Steven 273 , 291 Plank, Frans 344, 354 Popper, Karl R. 11 , 45 Pörings , Ralf 43 , 294 , 319 , 320 , 354 Pustejovsky, James 159 , 191 Pustet, Regina 30, 45 Quirk , Randolph 81 , 88 , 99 , 137 , 154 , 167 , 191 , 358 , 360 , 371 , 374 Radden , Günter 1, 43 , 59 , 67 , 73 , 121 , 197 , 217 , 221 , 222 , 232 , 246 ， 249-251 ， 257 ， 258 ， 265一
267 , 269 , 270 , 287 , 288 , 291 , 294 , 300 , 316, 318-320 , 327 , 328 , 353 , 354, 361 , 362 , 369, 373 Radfo时，Andrew 56 , 73 Raidt, Edith H. 271 , 291 Ramachandran, V.S. 234 , 246 Raphaely, Dorth 234 , 246 Reiser, B. 236 , 245
Author index Flhodes， ~chard ~ce，
70, 73
Sally 77 , 79, 165 , 189, 191 ,
249, 266 Nick 68, 73 Roberg, Heinz 215 Roberts, John R. 201-203, 217 Rohde, Ada 29, 39, 249 , 265 , 266 Ross, John R. 33 , 43 Ross , Malcolm 95 , 96 , 99 Rudzka-Ostyn, Brygida 190, 192, ~emer，
193 , 215 , 217, 291 Ruiz de Mendoza Ib姐ez， Francisco José 30, 40, 99, 293 , 319, 320,
354 Rumelhart, David E. 222 , 246 Saeed, John 1. 358 , 374 Sætherø, Eli 199, 217 Sanders, Gerald A. 120, 269 , 292 Sanders, Paul E. 2, 27 , 42 , 188, 215 Sandra, Dominiek 1 日， 165 , 189, 191 , 247, 266 Saussure, Ferdinand de 1, 4, 45 ,
104, 120 Scardamalia, Marlene 273 , 290 Scheffer, Johannes 125 , 130, 154 Schiffrin, Deborah 179, 191 Schladt, Mathias 215 Schmidt, Hans-Jδrg 251 , 252 , 263 ,
264, 267 Scholtz, Johannes du P. 271 , 292 Schwartz, Daniel L. 221 , 236, 247 Seiler, Hansjakob 1 日， 120 Serr冠 Bometo， Carlo 197, 218 Seto, Ken-ichi 320 Shen, Ya-Ming 197, 218 Shepard, Roger N. 236 , 247 Shisler, Benj缸ninK. 18, 45 Simon-V andenbergen, Anne-Marie
197, 217 Sinha, Chris 159, 160, 183 , 189,
191, 272, 292
379
Slobin, Dan 1. 231 , 241 , 247 Solso, R.L. 273 , 292 Sperber, Dan 31 , 45 , 252 , 266 Stefanowitsch, Anatol 29 , 39, 90,
91 , 100, 149, 154, 249, 265 , 266 Stegmüller, Wolf阴19 4, 45 Stephanides，直va H. 345 , 354 Stirling, Lesley 301 , 304, 306, 307,
310, 320 Stolz, Thomas 215 Svartvik, Jan 154, 191 , 374 Svorou, Soteria 163 , 192, 200 , 218 Swadesh, Morris 218 Sweetser, Eve E. 21 , 46 , 159, 162,
192, 195 , 197, 218, 222, 223 , 225, 247 Tai , James H.Y. 270, 292 Talmy, Leonard 28 , 119, 192, 221 , 224一226， 231 ， 233 ， 241 ， 247，
250, 251 , 256, 257, 266 Taylor, John R. 5, 16, 17, 18, 19,
24, 26, 32, 34, 35 , 43 , 44 , 46, 49 , 61 , 73 , 77, 78 , 83 , 100, 160, 165 , 192, 274, 292, 294, 327, 354 Thompson, Sandra A. 124, 153 , 247, 266 Thomburg, Linda 293 , 295 , 319, 328, 352, 354 Torres Cacoullos , Rena 129, 154 Traugott, Elizabeth Closs 10, 22 , 44, 104, 106, 119, 120, 126, 127, 154, 163 , 190, 192, 214, 295 , 320, 325 , 353 Tuggy, David 269 , 339, 354 Tumer, Mark 189, 195 , 197, 216 , 218, 270, 291 , 294, 319, 362, 372, 373 Tversky, Barbara 237 , 248 Tyler, Andrea 2, 21 , 27 , 37, 157160， 163 ， 16ι168， 187, 189, 192
380
Author index
Ullmann , Stephen 17, 46 , 327 , 355 Ungerer, Friedrich 251 , 252 , 263 , 264 , 267 Vachek, Josef 323 , 355 van der Leek, Frederike 56, 72 , 77 Van Huyssteen , Gerhard B. 40 , 269 , 273 , 282 , 292 Van Niekerk, Frans 273 , 290 Vandeloise , Claude 167 , 169, 170, 192 Vanrespaille , Mia 191 Varzi , Achille C. 234 , 244 Verhaar, John W.M. 87 ，邸， 100 Verhagen , Arie 13 , 46 , 269 Verspoor, Marjolijn 98 , 250 , 251 , 257， 258 ， 265一267
Vuillaume , Marcel 82 , 86 , 93 , 100
Wagner, Günter 208 , 218 Wagner, S.235 , 244 Waltereit, Ri chard 295 , 296 , 320 Watkins , Laurel 1. 208 , 218 Waugh , Linda R. 46 Welmers , William E. 200 , 218 Wierzbicka, Anna 104 , 120 , 197 , 218 Wilcox , Sherman 191 , 272 , 292 Wilson , Deirdre 31 ，侣， 187 , 252 , 266 Yu , Ning 31 , 46 Zeki , Semir 234 , 248 Zipf, George 30 , 45 , 46
Subject index
Action Schema 26 , 103 , 116 active zone specification 321 Afrikaans 30， 39， 269-272， 274一 276 , 278-280, 282 , 289-292 anaphor 294 antecedent 30, 178 , 293-295 , 300, 306, 307, 309 , 311-316 ascriptive construction 321 , 339, 343 , 347 , 348 aspect 24, 36, 38 , 39, 50, 56, 58 , 70 , 94 , 116, 117 , 121 , 123 , 133 , 134, 143一 145 ， 149-152 ， 158 ， 172 ，
184, 185 , 194, 197 , 201 , 204 , 206 , 209 , 210 , 216 , 254 , 262 , 280 , 289 , 304, 329, 358, 363 , 371 attribute 75 , 95 , 96 , 297 , 333 , 339, 350 autonomy 35 , 50, 112, 269 , 270 , 274 , 285 , 289 basic verbs 28 , 3ι38 ， 151 ， 193 ， 197, 200 , 203 blending 29 , 75 , 94 , 95 , 99 , 192 , 339 , 347, 348 , 372 categories 29 , 36, 52, 54, 57, 59, 70 , 88, 94 , 103 , 107 , 116 , 117 , 124, 160, 165 , 166, 189, 208 , 213 , 243 , 258 , 353 , 357 , 362 , 366, 367 cognitive grammar 39, 269 , 270 , 272 , 273 , 275 Companion Schema 103 complementarity 193 , 204 composition 39, 192, 269 , 270 , 272 , 273 , 275 , 276 , 285 , 289
conceptual semantics 222 concretization 103 , 106 constituency 143 , 269 , 270 , 274 , 275 , 282 , 283 , 289 constraints on names 357 construction 19 , 26 , 29 , 34-36, 41 , 49-58 , 61 -6 4, 66-6 8, 70-72, 75-79, 81-92 , 94-97 , 105 , 107, 113 , 114, 116, 123-127, 129 137, 140, 142-145 , 147-150, 154, 159, 161 , 200 , 201 , 203206 , 225-227 , 243 , 249 , 265 , 272 , 275 , 276 , 280 , 284 , 285 , 288 , 289 , 321 , 33 0-335 , 338340, 347-350 contextualization 103 , 106 contrastive typology 321 co甲us linguistics 249 correspondence 18 , 269 , 270, 274 , 276 , 289 , 301 , 303 deferred indexical reference 294 , 301 , 302 , 304 dependence 270, 275 , 285 , 286 distributed semantics 157, 159, 160 domain expansion 294 , 297-299 , 304 domain highlighting 294 , 297 , 328 emergent grammar 121 entrenchment 49 , 61-63 , 165 , 166 , 272 experiential reality 193 , 195 , 211 , 214 expressivity 26 , 75一77 ， 96， 97 extralinguistic motivation 193 , 201 , 211
382
Su句;ect
index
fictive motion 29 , 38 , 221 , 222 , 224 ， 225 ， 228-233 ， 237 ， 240一
243 , 246 figurative language 197 , 222 , 238 , 239 , 244 frequency data 249 gapping 249 , 250 , 251 genetic motivation 26 , 27 , 34, 36, 103 , 105 , 106 , 117 Germanic languages 75 , 81 , 84 , 93 , 144, 153 , 323 goal bias 39, 249 , 250 , 252 , 255 , 257-262 ， 264一266
Goal Schema 103 , 116 , 117 grammar 4 , 11 , 12 , 34 , 36 , 37 , 41 , 43 , 44 , 49-51 , 57 , 63 , 71 , 77 , 78 , 82-84 , 97 , 103 , 117, 119 , 122 , 150, 151 , 192 , 194, 201 , 202 , 209 , 211 , 212 , 214 , 216-218 , 243 , 247 , 251 , 269 , 270 , 275 , 289 , 291 , 292 , 316, 319 , 320, 324 , 325 , 327 , 351 , 352 , 357 , 363 , 365 , 366 grammatical constructions 19, 122 , 126, 132 , 150, 151 , 240 , 270 , 273 , 275 , 289 , 321 grammatical relations 321 grammaticalization 8, 16, 22 , 23 , 27 , 36, 37, 73 , 103 , 106, 111 , 117, 119, 121 , 122, 126, 128 , 129, 134, 135 , 137, 138 , 144, 150, 151 , 154, 163 , 193 , 200 , 320, 325 grammaticalization path 121 , 151 habitual aspect 121 head 7, 19 , 35 , 68 , 75 , 78 , 81 , 82 , 85 , 87-90 , 95-97 , 112 , 229 ideophone 34 , 49 , 68 , 70 idiomaticity 49 , 61
imagery 221 , 222 , 229 , 233 , 236238 , 242 , 247 implicative reference 294 , 315-317 indefinite article 36 , 81 , 82 , 93 , 103 , 107 , 11 0- 113 , 117 , 358 linguistic motivation 4 , 10, 23 , 24 , 193 , 196, 201 , 211 , 222 Location Schema 26 , 103 locative 36 , 67 , 108 , 109 , 114, 117 , 121 , 123 , 126 , 127 , 130-133 , 135 , 140, 147 , 150, 202 , 206 , 209 , 249 , 332-334, 350 locative constructions 67 , 121 , 123 , 126, 127, 130-132 , 150 mapping 40 , 94 , 150 , 199, 293 , 29 6-299 , 304, 306, 308-311 , 314, 316, 328 , 332 , 333 , 336 , 357 , 361-364, 366, 368 , 369 markedness 75 ma仕ix domain 40 , 294 , 298-301 , 303 , 304 , 307-309, 311-317 meaning extension 9, 19 , 26 , 37 , 157, 158 , 162 , 163 , 165 , 167 , 188 , 282 mental simulation 222 , 233 metaphor 2 ， 7 ， 17 ， 29 ， 31 ， 32 ，侣， 75 , 82 , 83 , 92 , 94 , 135 , 175 , 177 , 195 , 197 , 216 , 294 , 296 , 297 , 318 , 319 , 326, 327 , 351 , 354, 362, 372 , 373 metonymy 2, 5, 8, 16-18 , 29 , 30 , 34, 39, 40 , 42 , 43 , 59, 231 , 232 , 246 , 269 , 270 , 280 , 282-284 , 287-289 , 293-321 , 324, 326329 , 332 , 338-341 , 347, 349351 , 354 , 357 , 359 , 361 , 362 369, 371 373 motion event 39, 145 , 146 , 231 , 234 , 245 , 249-252 , 254 , 255 , 257-259 , 262-265 •
Su句;ect ind，臼
motion verb 36, 39, 121 , 123 , 129, 130, 145 , 146 , 148-151 , 221227 ， 231 ， 237-242 ， 249， 251一
253 , 255-260 , 263 , 264 Noun Phrase 75 , 98 numerals 36, 103 , 107-110, 112 , 113 , 117, 284
of 75 , 76 , 81 , 89-92 origin of names 357 paragon name 41 , 357 , 361 , 363366, 370, 372 path 38, 39, 59, 67 , 197, 221 , 22 4226 , 23 0-234 , 238-241 , 243 , 249-251 , 253 , 254 , 256 , 257 , 262-264 , 266 polysemy 16, 20 , 21 , 27 , 33 , 38, 41 , 42 , 59 , 157 , 158 , 160, 161 , 165一 167 , 172 , 177 , 188 , 190, 192, 195 , 215 , 321 , 324, 329-332, 340, 341 , 345 , 349, 350 possession 36, 103 , 107 , 113一 117 posture verbs 38, 121 , 128 , 129, 137-141 , 143 , 144, 146, 150, 193 , 206 , 208 , 210-213 pragmatic strengthening 37, 157 , 163 , 167, 177 , 178 , 181 , 182 , 185 predicate transfer 294 , 301-304 predicative adjectives 321 , 322, 324, 326, 330, 332, 339, 340343 , 349 preposition 9, 21 , 27 , 35 , 37 , 42 , 59 , 62 , 75 , 76 , 81 , 86 , 87 , 89 , 90 , 92 , 131 , 132 , 157-160, 165 , 166, 175 , 176, 178 , 188 , 192 , 253 , 339 principled polysemy 37, 157, 158 , 165 , 177 productivity 49 , 61 , 158 , 197
383
profile determinacy 269 , 270 , 274 , 289 progressive aspect 36, 121-123 , 125-130, 135 , 137, 143 , 144, 149 , 207 , 210 , 282 proper name 30, 41 , 357 proto-scene 28 , 37, 157 , 166- 168 , 170- 172, 174, 175 , 178 , 181 , 185, 186 psycholinguistics 222 radial categ。可 157， 166 raising cons位uction 321 , 332, 342 , 343 , 345 , 346 reduplication 17 ， 39， 269-272 ， 274一 276 , 280, 282 , 284 , 285 , 288290 Romancelanguages 35 , 75 , 76 , 80 , 81 , 86 , 92 , 93 , 348 sanctioning sense 37, 157, 165-167 schema-instance relation 49 , 51 , 54 schematic metonymy 357 , 361 , 362 semantic network 157 , 158 , 163 , 165 , 16~ 172 , 175 , 177 serial verb construction 9 , 193 , 20 4-206 situated implicature 157 , 164 source domain 32 , 40 , 94 , 199, 294 , 297 , 302 , 304, 306, 309, 363 , 364 Source Schema 103 , 115 spatial models 222, 236 spatial scene 157, 159, 166, 168 , 170, 171 , 179, 182, 184, 223 , 224 stereotype 357 , 364, 366, 367, 372 structural motivation 36, 78 , 103106 subject 13 , 19, 38 , 39, 56, 63 , 64 , 66， 67， 70， 87-90， 93 ，饵， 11 4117 ， 126， 136一 138 ，
140, 142,
384
Su苟ect
index
145 , 148 , 194 , 198 , 202-204 , 207 , 212 , 215 , 225 , 226 , 230 , 310, 321 , 331-337 , 339 , 343 , 346-348 , 350, 360 syntagmatic relation 5, 49 , 51
target domain 40 , 94 , 294 , 297 , 300 , 302 , 304 , 317, 363 , 364 windowing
249， 2日， 266
List of contributors
Antonio Barcelona Departamento de Filología Inglesa Universidad de Murcia E-30071 Murcia Spain e-mail: [email protected]
Mario Brdar Department of English Language and Literature Josip Juraj Strossmayer University Lorenza Jägera 9 HR-31000 Osijek Croatia e-mail: [email protected]. hr
Rita Brdar-Szabó Germanistikai Intézet Eötvös Loránd University 19-21 Ajtósi Dürer sor H-1146 Budapest Hungary e-mail: ritamario@dravanet. hu Olga Isabel Díez Velasco Universidad de La Ri oja Departamento de Filologías Modemas Edificio Filología C/ San José de Calasanz s/n Campus Universitario E-26004 Logroño Spain e-mail: isabel_velasco@hotmail. com
386 List 01 contributors Vyvyan Evans Department of Li nguistics and English Language Arts B136 University of Sussex Brighton, BNl 9QN Great Britain e-mail: [email protected]
Ad Foolen Department of Linguistics Erasmus Building Erasmusplein 1 P.O. Box 9103 NL-6500HDN材 megen
The Netherlands e-mail: a.foolen@le t. kun.nl
Bernd Heine Universität zu Köln Institut für A企ikanistik Albertus-Magnus-Platz D-50932 KδIn Germany e-mail: [email protected]
Christian Koops Department of Linguistics M523 Ri ce University 1600 Main St. Houston Texas , 77006 USA e-mail: ckoops@mai l. rice.edu
List 01 contributors 387
Teenie Matlock Oepartment ofPsychology Stanford University Building 420 Stanford Califomia, 94305-2130 USA e-mail: 恤[email protected]
JohnNewman Oepartment of Linguistics 4-32 Assiniboia Hall University of Alberta Edmonton Alberta, T6G 2E7 Canada 。mail: [email protected]
Kl aus-Uwe Panther (ed.) Institut 缸r Anglistik und Amerikanistik Universität Hamburg Von-Melle-Park 6 0-20146 Hamburg Germany e-mail: [email protected]
Günter Radden (ed.) Institut 阳r Anglistik und Amerikanistik Universität Hamburg Von-Melle-Park 6 0-20146 Hamburg Germany e-mail: [email protected]
388 List 01 contributors
AdaRohde Bahrenfelder Str. 332 D-22765 Hamburg Germany e-mail: [email protected]
Francisco José Ru Ïz de Mendoza Universidad de La Rioja Departamento de Filologías Modemas Edificio Filología CI San José de Calasanz s/n Campus Universitario E-26004 Logro负。 Spain e-mail: [email protected]
Anatol Stefanowitsch FB 10: Sprach-und Literaturwissenschaften Anglistik Universität Bremen Postfach33 0440 Bibliotheksstraße D-28334 Bremen Germany e-mail: [email protected]
John R. Taylor English Department. Li nguistics Programme Universi可 ofOtago
PO Box 56 Dunedin New Zealand e-mail: [email protected]
List 01 contributors 389
Andrea Tyler Dep町恤lent of Linguistics Box 571051 Georgetown University Washington D.C. , 20057-1036 USA e-mail: [email protected] Gerhard B. van Huyssteen School of Languages Potchefstroom University for CHE Private Bag X6001 2520 Potchefstroom SouthA企ica
e-mail: ntigbvh@puknet. puk.ac.za
