Studies In Linguistic Motivation (Cognitive Linguistic Research)

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