Contrastive Analysis in

Language

Identifying Linguistic Units of Comparison

Edited by

Dominique Willems

Bart Defrancq

Timothy Colleman

Dirk Noël

v

Contents

Acknowledgements vii

Notes on Contributors viii

Introduction 1

Part I Semantics: the Metalanguage of Comparison

1 Semantic Primes within and across Languages 13

Cliff Goddard

2 A Semantic Map for Imperative-Hortatives 44

Johan van der Auwera, Nina Dobrushina and Valentin Goussev

Part II Syntax: Constituent Order in Comparison

3 'Basic Word Order' in Formal and Functional

Linguistics and the Typological Status of

'Canonical' Sentence Types 69

Frederick J. Newmeyer

4 Division of Labour: the Role-semantic Function

of Basic Order and Case 89

Beatrice Primus

Part III Morphology: Agents in Comparison

5 Action and Agent Nouns in French and Polysemy 137

Petra Sleeman and Els Verheugd

6 Deverbal Nouns and the Agentive Dimension

across Languages 155

Filip Devos and Johan Taeldeman

7 Deverbal Nouns in Russian: in Search of a Dividing Line 172

Katia Paykin

Part IV Discourse and Beyond: Text in Comparison

8 Contrastive Analysis across Time: Issues in

Historical Dialogue Analysis 197

Andreas H. Jucker

9 Contrastive Textlinguistics and Translation Universals 213

Andrew Chesterman

10 Genre and Multimodality: Expanding the Context for

Comparison across Languages 230

John Bateman and Judy Delin

Language Index 267

Subject Index 269

vi Contents

vii

Acknowledgements

The ten chapters of this volume are all based on papers read at the

'Contrastive Analysis and Linguistic Theory' symposium that took place

at Ghent University on 21 and 22 September 2001. The symposium was

organized by the 'Contrastive Linguistics and Language Typology in

Europe' research network (CoLLaTE), sponsored by the Flemish National

Science Foundation (FWO).

The editors wish to thank Palgrave's commissioning editor for language

and linguistics, Jill Lake, for guiding the book so smoothly

through the publication process, and the administrative assistant of

Ghent's French Linguistics unit, Laurence De Wilde, for preparing the

camera-ready copy for the volume.

viii

Notes on Contributors

John Bateman is Professor of Applied Linguistics at the University of

Bremen, Germany. His main research focuses are multilingual NLG,

multimodal document design, discourse structure, and the application

of all areas of systemic-functional linguistics.

Andrew Chesterman is currently Professor of Multilingual

Communication in the Department of General Linguistics, University of

Helsinki. He has published widely in translation theory and contrastive

analysis.

Timothy Colleman is a researcher attached to the University of Ghent's

contrastive grammar research group (Contragram) where he is involved

in the design and compilation of a Dutch-French-English verb valency

dictionary. His main interest lies in (Dutch) syntax and lexicogrammar,

especially valency-alternating phenomena.

Bart Defrancq is a senior researcher attached to the University of

Ghent's contrastive grammar research group (Contragram) where he is

involved in the design and compilation of a Dutch-French-English verb

valency dictionary. His major publications are in the area of French

interrogative structures and subordination.

Judy Delin is head of research at Enterprise IDU and Reader at

Nottingham Trent University, where she focuses on issues of multimodality,

document design and multilinguality. Her main research

interests are in syntax, discourse analysis, the structure and usability of

information and the layout of illustrated documents.

Filip Devos is an assistant lecturer in the Department of Dutch

Linguistics at Ghent University, Belgium. His research interests include

word formation, contrastive linguistics and lexical semantics.

Nina Dobrushina obtained her PhD at the University of Moscow in

1995. Her main research interests are in grammatical typology, verbal

morphology, mood and modality, with specific attention on Caucasian

languages.

Cliff Goddard is Professor in Linguistics at University of New England,

Armidale, Australia. His research interests lie at the intersection of language,

meaning and culture. With Anna Wierzbicka, he is one of the

principal proponents of the natural semantic metalanguage approach.

Valentin Goussev is a researcher at the Institute of Linguistics of the

Russian Academy of Sciences in Moscow. His research interests focus on

typology of grammar and the Uralic languages.

Andreas H. Jucker is Professor of English Linguistics at the University of

Zurich, having taught previously at the Justus Liebig University,

Giessen. His current research interests focus on historical pragmatics,

cognitive pragmatics and hypertextlinguistics.

Frederick J. Newmeyer specializes in syntax and the history of linguistics

and has as his current research programme the attempt to synthesize

the results of formal and functional linguistics. He was President of the

Linguistic Society of America in 2002.

Dirk Noël is a senior researcher attached to the University of Ghent's

contrastive grammar research group (Contragram) where he is involved

in the design and compilation of a Dutch-French-English verb valency

dictionary. His major publications are in the area of English clausal

complementation.

Katia Paykin, a native speaker of Russian, obtained her BA at Columbia

University (NYC, USA). She is currently finishing her PhD thesis on the

syntactic behaviour of meteorological nouns at the University of Lille III,

France (CNRS-UMR 8528 SILEX).

Beatrice Primus is Professor of German Linguistics, University of

Cologne. She held previous positions at the universities of Munich,

Heidelberg and Stuttgart. She is co-editor of the monograph series

Linguistische Arbeiten (Niemeyer, Tübingen) and Vice President of the

Scientific Board of the Institute of German Language (IDS), Mannheim.

Her major areas of research are semantic roles, case selection, relational

typology, word order, writing systems and pragmatics.

Petra Sleeman has been working in the Department of Romance

Linguistics of the University of Amsterdam since 1987, teaching French

linguistics. Her main field of research is French syntax, more specifically

the syntax of the DP, often in comparison to other languages, especially

Dutch. She has also published on the interface between syntax and morphology

and the interface between syntax and information structure.

Johan Taeldeman is Professor of Dutch Linguistics and Dialectology at

Ghent University, Belgium. His research interests include phonology,

word formation, language variation and language change.

Johan van der Auwera teaches English and general linguistics at the

University of Antwerp. His current interests include mood and modality,

grammaticalization and negation, from a typological point of view.

Notes on Contributors ix

Els Verheugd has been teaching French linguistics in the Department of

Romance Linguistics of the University of Amsterdam since 1984. Her

main field of interest is French syntax. She has been working on the

structure of copula sentences, on the internal and external structure of

APs, on the argument structure of deverbal nouns, and on the structural

representation of information structure in French.

Dominique Willems is Professor of French Linguistics at Ghent

University, Belgium. She is a member of the contrastive grammar

research group (Contragram). Her main research interests lie in the

intersection of syntax, lexicon and semantics.

x Notes on Contributors

1

Introduction

The reader must not be confused by the title of this volume. This book

is not about 'contrastive analysis' as a distinctive branch of linguistics,

either in its narrow sense of contrastive research in the context of second

and foreign language teaching, or in the slightly wider sense in which

contrastive analysis becomes synonymous with 'contrastive linguistics'

and refers to research about differences and similarities between a limited

number of languages carried out 'for its own sake'. This is a book

about linguistics, not about any of its subdisciplines in particular. It is a

book about comparison in linguistics, but linguistic comparison is not

the prerogative of contrastive linguistics, it is central to linguistics as a

whole. Even the linguist who is interested in just the one language will

have to make use of an analytical apparatus whose distinctions are more

generally applicable, for fear of producing a completely idiosyncratic,

and therefore futile, description. Every linguistic analysis, therefore,

presupposes, or should presuppose, some form of comparison. This

book is about the analytical apparatus we use as linguists and contains

reflections on whether it can stand the comparison test: does it allow

comparisons to be made across languages?

The ten chapters of the book cover a broad spectrum of linguistic disciplines,

ranging from contrastive linguistics and linguistic typology to

translation studies and historical linguistics. We have nevertheless given

pride of place to contrastive analysis in the title of the collection to

emphasize the fact that all of its contributions are either explicitly contrastive

in nature or very relevant to the comparison question. Also, of

all linguistic disciplines, contrastive analysis, or contrastive linguistics,

is undoubtedly the one that most consciously reflects on the methodology

of making comparisons (cf. Krzeszowski 1990; Chesterman 1998).

This preoccupation with methodology is very much present in all the

chapters of this volume. Some of the chapters are also arranged in such

a way as to juxtapose different methodologies, and therefore the book

can also be said to be contrastive analytical in a meta-theoretical sense.

The contributions to the book are grouped in four parts, each of which

focuses on a different area of linguistics: derivational morphology, syntax,

semantics and pragmatics, and discourse studies. Of course, any linguistic

comparison always involves elements from different linguistic

areas (cf. Chesterman 1998), and assigning chapters to only one of two

(or more) of the areas involved may be thought to be an unjustifiable

reduction. As we will point out below, contributions placed in different

parts of the book sometimes bear a close relationship to each other.

However, this area-based presentation has the advantage of showing

that not all fields of linguistics have developed an equal concern for

comparability. In general, the contributions on derivational morphology

and discourse are far more programmatic in nature and descriptively

less elaborate than the contributions on syntax and on semantics and

pragmatics, areas that can boast a much longer tradition of contrastive

research. We have placed the more programmatic sections at the end of

the volume so as to provide it with an open ending that can spark off

further investigations.

In the remainder of this introduction we will clarify what unites the

chapters grouped in each of the volume's four parts and indicate points

of connection across the various section boundaries.

1 Semantics: the metalanguage of comparison

The semantics section of the volume essentially deals with the methodological

question of how to develop and implement a framework that

allows a uniform semantic description of linguistic phenomena. Such a

description is to serve as the necessary tertium comparationis in the comparison

of formal features in different languages: a cross-linguistic

comparison of forms is only meaningful if the forms compared have a

comparable function, and to be able to establish comparability we need

a uniform metalanguage. In the first chapter, Cliff Goddard examines

the semantic primes theory developed by Wierzbicka (1972) and others,

to see whether it is theoretically appropriate to form the basis of such a

universal semantic description or, as Wierzbicka and her associates call

it, a natural semantic metalanguage. Goddard assumes that in order to

make language comparison on the basis of semantic primes possible

the inventories of basic primes for each of the languages under examination

should coincide. Every difference between inventories makes the

comparison considerably more difficult and risky. After a careful

examination of examples of such differences, Goddard arrives at the

conclusion that they are often only apparent differences and mostly the

result of analyses that remain too close to the surface. Since semantic

primes are lexical units of the languages themselves, they are also

assumed to allow a semantic description that is intuitively understandable.

Goddard believes, for instance, that a natural semantic metalanguage

can easily rephrase notions such as 'imperative' and 'agentivity',

used in other chapters of this volume, in a much more transparent way.

It can indeed not be denied that the notion 'imperative' is a problematic

one. In the second chapter, Johan van der Auwera, Nina Dobrushina

and Valentin Goussev point to the fact that the term can cover very

different realities in the grammatical descriptions of different languages.

Their own solution to the problem is not a semantic primes approach,

but a methodology that makes use of semantic maps, inspired by

Anderson (1982). Normally, the metalanguage of such an approach consists

of a number of semantically related concepts that are brought

together in a structured set. The projection of the formal structures of

different languages on such a 'conceptual map' makes it possible to visualize

the degree of formal correspondence between languages. Van der

Auwera et al., however, introduce a pragmatic metalanguage in order to

define the illocutionary force associated with the imperative. On the

basis of a considerable number of languages, they succeed in showing

that the results of the function-form mapping fully obey the principles

of semantic maps, in particular the principle that in every language

identical formal features only realize contiguous semantic concepts, that

is pragmatic uses.

However, though their methodology turns out to be successful, their

definition of these pragmatic uses as well as their function-form mappings

should perhaps be confronted with some of the claims made in

other chapters of this volume. Andreas Jucker, for example, observes

that 'it is not clear that the illocutionary force is necessarily the best

tertium comparationis even for contrastive speech act analyses, because

different speech communities may very well encode a different range of

speaker intentions'. John Bateman and Judy Delin, on the other hand,

illustrate that languages possess a much larger variety of formal realizations

of the directive illocutionary force than the description by van der

Auwera et al. suggests. All kinds of indirect requests are ignored, for

instance, even though they seem to be more frequent than the straight

imperative in some genres and in some languages. Because of the

competition they engage in with other forms, straight imperatives mayhave different values in different languages, in which case comparing

them cross-linguistically without describing the actual conditions of

their use is tricky. Of course, nothing really prevents the inclusion of all the

different parameters in the map. It remains to be seen, however,

whether the formal features still form contiguous areas on the map in

this case.

2 Syntax: constituent order in comparison

The syntax section centres on an area of linguistic comparison which

has proven to be very prolific in the last 40 years or so: constituent order.

Since the seminal work of Greenberg (1963), constituent order has been

a most inspirational research topic, both in terms of language coverage

and in terms of analytical depth. The original ideas of Greenberg have

been much criticized, however, especially for their reliance on the universality

and cross-linguistic comparability of notions like subject,

object and verb, which is called into question by numerous authors, and

for the privileged status accorded to sentences with two full lexical NPs,

which are not all that common in actual language use.

In Chapters 3 and 4, Frederick Newmeyer and Beatrice Primus concentrate

on the main aspects of this criticism. Newmeyer takes up the

issue of what he calls the 'canonical word order paradox'. With this he

refers to the discrepancy between the fact that the order of the constituents

of a declarative main clause with lexical arguments is typologically

very important because it correlates significantly with word order

on other syntactic levels (inside the noun phrase, for instance), and the

fact that this order is almost insignificant when we consider its frequency

in actual speech. After reviewing some of the possible solutions

to the paradox, Newmeyer concludes that the best way to deal with the

problem is to assume that the paradox results from the different requirements

imposed on structure in two different stages of speech production.

He argues that, in an early stage, lexical information is retrieved

from memory in the form of 'lemmas', which, in the case of verbal

items, include a complete argument structure. It is this particular structure

that is subject to general processing constraints, which accounts for

the correlations with other 'early' structural constraints. However, once

we have reached the stage of the utterance, discourse requirements often

prevent lexical argument structures from being used. Since the stage of

retrieval is closest to the level of competence and the stage of utterance

closest to the level of performance, the order of the fully lexical argument

structure has to be considered to be the canonical orderFor Beatrice Primus, on the other hand, the basic constituent order of

a language is the result of a competition between various constraints

that languages seek to comply with in the structural set-up of the

sentence. Her contribution shows the importance of an approach whose

categorial inventory is not limited to a minimal set consisting of S, V

and O, but which takes into account all possible forms of case marking.

Primus considers case marking principally against the background of

thematic proto-roles. These are analysed following Dowty's (1991)

proto-role framework as the proportion of agent and patient relations

entailed by the predicate with respect to a specific argument. They are

presented by Primus as the basis for case assignment in many languages

and seem also to determine, via a cognitively motivated ordering

constraint that competes with other kinds of constraint, the order

in which constituents appear in the sentence: the arguments with

more proto-agent properties are placed before the arguments with more

proto-patient properties and this thematic constraint can be explained

as an iconic representation of a causal relation between the two

proto-roles.

Most typical of both Primus's and Newmeyer's contributions - and, in

fact, of the theoretical currents they represent - is that the analytical

tools used for the comparison are thought of as having at least some cognitive

relevance: Newmeyer argues that comparing constituent order on

the basis of lexical arguments is legitimate because he assumes argument

structure to be retrieved from the mental lexicon in a lexical form;

Primus refers to the cognitively prominent notion of causality in support

of her interpretation of proto-roles and the way these tend to be

ordered in the sentence, and, most important of all, presents neurolinguistic

evidence for the distinction she draws between different case

marking constraints.

It is interesting to compare Primus's view on proto-roles and agentivity

to the views represented in other contributions which also use agentivity

as a tertium comparationis. All the chapters in the morphological

section of the volume use some interpretation of agentivity in the comparison

of morphological processes, but the interpretations to be found

there differ substantially from the way it is viewed by Primus.

3 Morphology: agents in comparison

The morphology section focuses on the question of how the mapping

between the meaning and the suffixal form of derived nouns in different

languages can be compared. Special attention is given to nounsinvolving some degree of agentivity. This is a rewarding research topic,

because in many languages there is a derivational process that is

specially dedicated to the formation of agentive nouns on the basis of

verbal lexemes. The chapters in this section examine what it means

to be agentive or not agentive for a derived noun and if this distinction

is morphologically significant in the sense that it can be shown to

correlate with different word formation rules.

In Chapter 5, Petra Sleeman and Els Verheugd argue that for a derived

noun to be agentive, it must satisfy two conditions. First, it must be

derived from a verb which allows for an agent in its argument structure

(in which case the agent is not understood as analysable into protoproperties

but in a more traditional, atomic, sense) and it must refer to

that agent. Second, the noun must at least in some form inherit the argument

structure of the verb it is derived from. This seems to be the case for

both eventive and non-eventive -er nouns in the sense of Levin and

Rappaport (1988), but not for a third type of -er noun that Sleeman and

Verheugd propose to add to Levin and Rappaport's typology. In French,

derived -eur nouns referring to substances involved in a certain process

are said to belong to this third type, as well as nouns ending in -ant.

Starting from different premises, Filip Devos and Johan Taeldeman

arrive at similar conclusions in Chapter 6, which focuses on the

distribution of suffixes over derived nouns that represent the main

semantic roles of the verbs they are derived from. It appears from a

detailed study of word formation rules in French and Dutch that there

exists a dividing line between suffixes that are used to derive nouns

that can be interpreted as agentive (the agent and the instrument) and

suffixes used for non-agentive nouns. Devos and Taeldeman also

observe that derived nouns for substances behave like non-agentive

nouns, even though they are semantically close to instrument nouns.

Like Sleeman and Verheugd, they conclude that genuine instrument

nouns and nouns referring to substances result from different kinds of

word formation.

Contrary to Devos and Taeldeman's findings for French and Dutch,

Katia Paykin in Chapter 7 cannot confirm the existence of an agentive

dividing line in the word formation rules of Russian. If there is any

tendency at all in the extremely complex field of Russian derivation, it

seems to be motivated by an aspectual feature rather than a rolesemantic

one. These contrasting results clearly illustrate the possibilities

and limitations of a particular tertium comparationis in language comparison:

although agentivity is indeed a viable concept in the description of

the derivational properties of certain languages, it cannot be used of allMore fundamentally, what is striking about the use of agentivity in

language comparison is that different interpretations of the notion

appear to be relevant for the analysis of different formal features.

Whereas a traditional Fillmorean view on semantic roles yields interesting

results in the area of derivational morphology, research on case morphology

seems to fare better with a smaller inventory of roles enriched

with prototypicity criteria (cf. Primus in the section on syntax). Future

research will tell whether this is a motivated difference or simply

the result of a different stage in the advancement of comparative linguistic

research. After all, in the early days of typological research into

case morphology, Fillmore's approach to role-semantics yielded very

interesting results as well.

4 Discourse and beyond: text in comparison

The final part of the volume is dedicated to various tertia comparationis in

contrastive textlinguistics or contrastive discourse studies. Traditionally,

researchers working in this field investigate texts in differing languages,

focusing on their syntactic, semantic and discourse-functional characteristics.

The three contributions to this section take us far beyond this

point. All three explore new areas of contrastive research in which special

kinds of discourse are confronted.

In Chapter 8, Andreas Jucker considers some of the things that need

to be taken into account to be able to successfully implement the principles

and methodology of contrastive discourse analysis in the field of

historical discourse analysis. Comparing two (or more) diachronic stages

in one and the same language involves a great deal more than a synchronic

comparison of two (or more) separate languages, especially with

respect to the available data, the tertium comparationis and the historical

relationship between the stages under analysis. With regard to the

tertium comparationis, the main obstacle to comparison is that neither

form nor function remain constant in a linguistic change and that neither

of the two can thus be selected as a tertium comparationis for the analysis

of the other. Jucker therefore suggests an alternative in the shape of a

pragmatic map, which bears much resemblance to the kind of semantic

map used by van der Auwera et al. in Chapter 2.

In Chapter 9, Andrew Chesterman explores the extent to which, and

how, students of translation concern themselves with comparison. He

points out that translation studies, even in the very early, pre-theoretical,

stages of the discipline, are contrastive in nature, with (implicit) universals

and comparative practices of their own. The difference between

pre-theoretical and modern translation studies is that the former are

mainly concerned with what translations should be, whereas the latter

concentrate on what they really are, that is a specific text genre. Even

though the latter view has undoubtedly made an enormous contribution

to a realistic assessment of translation, Chesterman highlights some of its

blind spots and suggests new lines of research to deal with them.

John Bateman and Judy Delin, finally, take us beyond the purely linguistic

aspects of discourse. They argue that the locus of the comparison

is not necessarily restricted to the text as traditionally conceived, that is

its linguistic matter. Meanings traditionally expressed linguistically in

certain text types in certain cultures may very well be expressed graphically

in the corresponding or similar texts in other cultures. Bateman

and Delin therefore plead for the adoption of a multimodal approach to

contrastive analysis, which takes into account all the semiotic resources

that can be put to use.

The three chapters of the discourse section are highly programmatic

and this is hardly surprising: a strong emphasis on performance rather

than competence forces the researcher to take into account a vast range

of linguistic and extralinguistic parameters which can only be integrated

into a model of language comparison with extreme caution. Competenceoriented

studies, on the other hand, can restrict themselves to a limited

number of parameters and can therefore more rapidly arrive at an

exhaustive description of the data. The contrast between the chapters by

van der Auwera et al. and Jucker illustrates this difference most eloquently.

However, even though performance-oriented research may be

less effective in terms of research results, we should not ignore the fact

that it plays an important part as a forerunner in the progress of linguistic

research and that the new parameters it comes up with will eventually

find their way into a model of competence, for the one cannot exist

without the other.

Part I

Semantics: the Metalanguage of

Comparison

13

1

Semantic Primes within and

across Languages

Cliff Goddard

1.1 Introduction1

The chapter adopts the standpoint of the Natural Semantic Metalanguage

(NSM) theory, originated by Anna Wierzbicka (1972, 1980, 1988, 1992,

1996). It outlines how semantic primes have been identified within and

across languages, over several decades of empirical research, and how

they can be used as a tool in lexical and grammatical typology and

contrastive linguistics.

The semantic primes of any language L can be defined as the terminal

elements of reductive paraphrase analysis conducted within that language,

that is the set of lexical meanings which cannot be defined

language-internally without circularity. For present purposes, rather

than argue about whether or not semantic primes exist, I would prefer

to assume the 'metasemantic adequacy' of ordinary languages, that is to

assume that ordinary languages are adequate to represent their own

semantics via language-internal paraphrase. On this assumption, it follows

that all languages have an irreducible 'semantic core' which would

be left after all the decomposable expressions had been dealt with. This

semantic core must have a language-like structure, with a lexicon of

semantic primes and associated grammar. What needs to be established

by empirical-analytical work are the identities and properties of semantic

primes in individual languages, the extent to which semantic primes

of different languages coincide and the extent to which semantic primes

are relevant to understanding the workings of language generally.

How can one establish the identities and properties of semantic

primes? Essentially, by experimentation; that is by an extensive programme

of trial-and-error attempts to explicate diverse meanings, aiming

always to reduce the terms of the explications to the smallest and mostversatile set. This is exactly what Anna Wierzbicka has done over

a period of 30 years (and continues to do). When Wierzbicka and colleagues

claim (for example Wierzbicka 1996, Goddard and Wierzbicka

2002) that GOOD, SAY and BECAUSE, for example, are semantic primes, the

claim is that these words are essential for explicating the meanings of

numerous other words and grammatical constructions, and that they cannot

themselves be explicated in a non-circular fashion. The same applies

to other examples of semantic primes such as: I, YOU, SOMEONE, SOMETHING,

THIS, HAPPEN, MOVE, KNOW, THINK, WANT, DO, WHERE, WHEN, NOT, MAYBE, LIKE,

KIND OF, PART OF.

Perhaps the key theoretical difference between NSM primes and, say,

Jackendoff's (1990) conceptual primitives or Pustejovsky's (1995) qualia

is the 'natural language principle' (cf. Goddard 1994a: 10). In the NSM

system, both the prime meanings and their syntax are taken from

within natural languages. The mini-language of semantic primes is literally

a subset of natural language, not an external system of representation.

The natural language principle also means that NSM primes are

'non-abstract'. In this way, the NSM system avoids an infinite regress of

interpretation. As John Lyons once remarked (1977: 12): 'any formalism

is parasitic upon the ordinary everyday use of language, in that it must

be understood intuitively on the basis of ordinary language'.

For many linguists and logicians working in other frameworks, nothing

is more mysterious and intangible than meaning, but adopting reductive

paraphase as a way of grasping and stating meanings makes meanings

concrete, tangible. Above all, it makes statements about meanings

testable - because explications couched in natural language can be

directly or indirectly substituted in place of the expressions they are

intended to represent, and so can be submitted to the test of substitution

salvo sensu.

The NSM approach has proved to be an extremely productive one

(see below). Using the approach it has repeatedly proved possible to

defy the sceptics and to 'define the indefinable', that is to explicate

semantic nuances which have been claimed to be either impossible or

excruciatingly difficult to describe. A couple of examples will help.

Chomsky (1987: 21) remarks: 'Anyone who has attempted to define a

word precisely knows that this is an extremely difficult matter, involving

intricate and complex properties. Ordinary dictionary definitions do not

come close to characterizing the meaning of words.' To illustrate this

point, one of Chomsky's examples is the visual vocabulary: words such as

watch, glare, gaze, scrutinize and so on. The meanings of language-specific

visual words are not difficult to explicate, however, within a reductive

paraphrase framework. To give a single example (cf. Wierzbicka 1996:

251-3, Goddard 1998):2

(1) X was watching Y

for some time, X was doing something

because of this, X could see Y for all this time

X was doing this because X thought:

when something happens in this place

I want to see it

For Chomsky's equally anti-semantic predecessor, Leonard Bloomfield,

the standard examples of words whose meanings could not be defined

precisely were emotion terms. There is, however, an extensive body of

descriptive semantics of emotion terminology (cf. Wierzbicka 1999).

Again, to give just a single example:

(2) X felt envious

X felt something bad

because X thought about someone else:

something good happened to this person

it didn't happen to me

this is bad

I want things like this to happen to me

As a final example, consider a word from the family of causative verbs.

Break (trans.) is often defined simply (and simplistically) as 'cause to

break (intr.)'. The NSM explication below (for one meaning, perhaps the

most central meaning, of break) is more elaborate, but considerably

more explanatory (cf. Goddard 1998):

(3) Person X broke Y (for example Pete broke the window)

X did something to Y

because of this, something happened to Y at this time

because of this, after this Y was not one thing any more

These explications should underscore the point that although the NSM

approach can be seen as a classical approach to semantics in some

respects, especially in its commitment to semantic description in

discrete propositional terms, it is quite unlike other so-called classical

approaches to semantics. NSM explications are not lists of necessary and

sufficient conditions, or bundles of semantic features. Further, as can be

seen from the examples above, it is entirely possible to incorporate

conceptual prototypes, scenarios and so on, within NSM explications.

Returning now to the topic of semantic primes themselves, it is to be

expected that each semantic prime will have characteristic syntactic

properties - combinatorics, valency and complementation options - as

a consequence of its meaning. By hypothesis, these syntactic properties,

as well as the identities of the primes themselves, are universal - and this

hypothesis seems to be borne out by a growing body of empirical

research. To give some impression of the kind of syntactic properties

which may be involved, Table 1.1 summarizes proposed valency and

complementation options for two predicate primes - SAY and THINK. The

prediction of NSM researchers is that in all languages it will possible to

express meanings equivalent to those expressed by SAY and THINK in these

specific syntactic contexts.

16 Contrastive Analysis in Language

Table 1.1 Proposed syntactic frames (valency and complementation options)

for SAY and THINK

SAY: X says something [substantive complement]

X says something to someone [addressee]

X says something about something [locutionary topic]

X says: '---' [direct speech]

THINK: X thinks about Y [topic of cognition]

X thinks something good/bad about Y [a 'compound valency' with

GOOD/BAD]

X thinks: '--' [quasi-quotational complement]

at this time, X thinks that [--]S [propositional complement]

A considerable amount is known about language-specific manifestations

of semantic primes, after some 30 years of research, initially by

Wierzbicka and since the early 1990s by an increasing number of colleagues.

Some of this work is tabulated in Table 1.2. Unless otherwise

indicated, language data adduced later in this chapter originates with

the sources listed in this table.

In section 1.3, I will address the question: What are semantic primes

good for? with particular reference to the concerns of contrastive linguistics

and typology. There I will argue that semantic primes enable

improved precision in contrastive studies, essentially because they

provide a stable tertium comparationis in terms of which one can investigate

lexical and grammatical typology. It should also be mentioned that

semantic primes provide a valuable tool for exploring polysemy. As

the term is currently used in linguistics, polysemy is best characterized

Cliff Goddard 17

Table 1.2 Languages other than English studied in NSM framework

Language Primes and syntax: Descriptive semantic studies

comprehensive study

Lao (Tai) Enfield (2002) Enfield (1999, 2001)

Mangaaba-Mbula Bugenhagen Bugenhagen (1990, 2001)

(Austro) (1994, 2002)

Malay (Austro) Goddard (2002a) Goddard (1994b, 1996,

1997a, 2001a, b)

Mandarin Chappell (1994, 2002) Chappell (1986, 1991),

Chinese (Sinitic) Ye (2001, in press),

Kornacki (1995, 2001)

Polish (IE) Wierzbicka (2002) Wierzbicka (1997, 2001)

Spanish (IE) Travis (2002) Travis (1998a), Curnow

(1993)

Hawaiian Stanwood (1997, 1999)

Creole English

Primes and syntax:

partial study

Acehnese (Austro) Durie et al. (1994),

Harkins (1995)

Amharic Amberber (2001a) Amberber (2001b)

(Ethiosemitic)

Arrernte (PN) Harkins and Wilkins (1994) Van Valin and

Wilkins (1993),

Harkins (2001),

Wilkins (1986, 2000)

Bunuba (non-PN) Knight (forthcoming)

Cantonese (Sinitic) Tong et al. (1997)

Cree (Algonquian) Junker (2001)

Ewe (Niger-Congo) Ameka (1994a) Ameka (1990a, b, 1994b,

1996)

French (IE) Peeters (1994, 1997a) Peeters (1993, 1997b, 2000)

German (IE) Wierzbicka (1997, 1998a),

Durst (1996, 2001)

Italian (IE) Maher (2000)

Longgu (Austro) Hill (1994), Hill

and Goddard (1997)

Japanese Onishi (1994, 1997), Hasada (1996, 1998, 2001),

Hasada (1997) Wierzbicka (1997),

Travis (1998b)

Kalam (Papuan) Pawley (1994)

Kayardild (Tangkic) Evans (1994),

Harkins (1995)

as a family resemblance concept, subsuming phenomena as diverse

as compositional ( classical lexical), non-compositional polysemy

(cf. Goddard 2002b: 26-30), regular polysemy, generative polysemy,

some kinds of metaphor/metonymy phenomena, and pragmatic enrichment

(via extralinguistic contextual inference). The core concept of polysemy,

however, that is classical lexical polysemy, crucially rests on the

'definitional test' - that is an item is polysemous if it is necessary to posit

two or more distinct but related definitions (explications) in order to

account for its range of use, entailments and so on (Geeraerts 1994,

Dunbar 2001). In my view, much of the confusion surrounding discussions

of polysemy stems from inadequate semantic methodology

(cf. Goddard 2000). (If one cannot define even a single meaning clearly,

how can one ever hope to decide whether a word has two or more

related meanings?) Insisting that lexical explications be formulated

within a simple and standardized vocabulary, that is the metalanguage

of semantic primes, imposes a degree of rigour and precision such that

the traditional definitional test of polysemy can be operationalized.

1.2 Identifying semantic primes within and

across languages

In this section I will survey various problems which arise in identifying

and matching exponents of semantic primes across languages. Before

18 Contrastive Analysis in Language

Table 1.2 (Continued)

Language Primes and syntax: Descriptive semantic studies

comprehensive study

Russian (IE) Wierzbicka (1992, 1997,

1999, in press), Zalizniak

and Levontina (1996),

Mostovaja (1997, 1998)

Samoan (Austro) Mosel (1994)

Sm'algyax Stebbins (forthcoming)

(Pacific-NW)

Thai (Tai) Diller (1994)

Ulwa (Misumalpan) Hale (1994)

Yankunytjatjara (PN) Goddard (1991a, 1994b) Goddard (1990, 1991b,

1992)

IE: Indo-European, PN: Pama-Nyungan (Australia), Austro: Austronesian.

that, however, a few more words may be useful on how the current

inventory of semantic primes was arrived at in the first place.

The definition of the term 'semantic prime' hinges on indefinability.

A semantic prime is a linguistic expression whose meaning cannot be

paraphrased in any simpler terms. A secondary criterion (on the hypothesis

of universality) is that a semantic prime should have a lexical equivalent

(or a set of equivalents) in all languages. These twin criteria mean

that the number of expressions which can be entertained as candidates

is rather small - because the vast majority of linguistic expressions can

readily be shown to be either semantically complex and/or languagespecific.

There is also a third consideration: taken as a whole, the metalanguage

of semantic primes is intended to enable reductive paraphrase

of the entire vocabulary and grammar of the language at large, that is it

is intended to be comprehensive.

To get a sense of what semantic primes look and feel like, we may

briefly consider two of them: GOOD and SAY. How could one decompose

or explain the meaning of good in terms which are simpler and not

language-specific? It would be no use appealing to terms such as approve,

value, positive and please, as these are both demonstrably more complex

than good and highly language-specific. The only plausible route would

be to try to decompose good in terms of actual or potential 'desirability';

for example, by saying that 'this is good' means 'I want this' or 'people

want this'; but such proposals founder for several reasons. Perhaps most

importantly, to label something as good is to present the evaluation in an

objective mode, not as the desire of any specific person, or even of

people in general. Explications of good in terms of 'wanting' yield very

peculiar results in cases where good is used in contexts such as 'X said

something good about Y', or about generic or hypothetical situations,

such as 'If someone does something good for you, it is good if you do

something good for this person.'

The difficulty of finding a satisfactory reductive paraphrase for GOOD

makes it a candidate for the status of semantic prime. Furthermore,

GOOD will clearly be required for the explication of innumerable lexical

items which imply positive evaluation (such as, to name a handful, nice,

tasty, kind, happy, pretty) and for grammatical constructions such as

benefactives. Upon checking in a range of languages, one finds that all

languages appear to have a word with the same meaning as English good.

For example: Malay baik, Yankunytjatjara palya, Ewe nyó, Japanese ii.

(Obviously, this does not mean that different cultures share the same

views about what kind of things are GOOD.)

Cliff Goddard 19

Now for the second example: SAY. Consider an exchange such as the

following:

(4) A: X said something to me.

B: What did X say?

A: X said 'I don't want to do it.'

How could we paraphrase away the term SAY in these contexts? It just seems

impossible. It would be no good to say 'verbally express', since using terms

like 'express' and 'verbally' would be moving in the wrong direction: in the

direction of increased complexity, rather than the other way around. The

only plausible line of explication appears to be via DO, WANT and KNOW; for

example, 'X said something to Y' 'X did something, because X wanted Y

to know something'. But this equation fails because the right-hand side

could be satisfied by many actions which were non-verbal (and not symbolic).

As in the case of GOOD, there are numerous lexical items whose

20 Contrastive Analysis in Language

Table 1.3 Semantic primes (after Goddard and Wierzbicka 2002)

Substantives I, YOU, SOMEONE/PERSON, PEOPLE;

SOMETHING/THING, BODY

Relational substantives KIND, PART

Determiners THIS, THE SAME, OTHER

Quantifiers ONE, TWO, SOME, ALL, MUCH/MANY

Evaluators GOOD, BAD

Descriptors BIG, SMALL

Mental/experiential THINK, KNOW, WANT, FEEL, SEE,

predicates HEAR

Speech SAY, WORDS, TRUE

Actions and events DO, HAPPEN, MOVE

Existence and possession THERE IS, HAVE

Life and death LIVE, DIE

Time WHEN/TIME, NOW, BEFORE, AFTER,

A LONG TIME, A SHORT TIME, FOR

SOME TIME, MOMENT

Space WHERE/PLACE, HERE, ABOVE, BELOW;

FAR, NEAR; SIDE, INSIDE, TOUCHING

Logical concepts NOT, MAYBE, CAN, BECAUSE, IF

Intensifier, augmentor VERY, MORE

Similarity LIKE

Cautionary notes

Exponents of primes can have other polysemic meanings which differ

from language to language.

They can have combinatorial variants (allolexes).

They can have different morphosyntactic properties (including wordclass)

in different languages.

meanings seem to be based on SAY - most notably, the class of speech-act

verbs. And there are grammaticalized meanings which involve SAY; for

example, evidential particles of the so-called quotative or hearsay variety.

The full current inventory of semantic primes is given in Table 1.3.

1.2.1 Matching primes across languages

It is useful at the onset to be clear on exactly what is involved in matching

exponents of semantic primes across languages. The key analytical construct

is the 'lexical unit' (Cruse 1986: 77-8, cf. Mel'ˇcuk 1989), that is the

pairing of a single specifiable sense with a lexical form. The concept of

lexical unit is not to be identified either with 'lexeme' (a family of lexical

units) or with 'lexical form' (a family of word forms which differ only

in respect of inflection). A polysemous word is a lexeme which consists of

more than one lexical unit. A lexical form need not be formally monomorphemic,

but may be a compound or derived word, or a phraseme. When

we seek to match exponents of semantic primes across languages, what

we are seeking to do is to align lexical units (across languages) which

share a given putatively primitive meaning.

One source of confusion is due to the fact that exponents of the same

prime can have different polysemic extensions in different languages; in

this situation there can be a match-up between lexical units but not

between whole lexemes. Common polysemies in which semantic primes

are involved include: WANT with 'like', 'love' or 'seek' (Spanish, Ewe, Ulwa),

HAPPEN with 'arrive' or 'appear' (French, Ewe, Mangaaba-Mbula), DO with

'make' (Malay, Arrernte, Samoan, Kalam), SAY with 'speak' or 'make

sounds' (Thai, Mandarin, Yankunytjatjara, Kalam), BEFORE with 'first', 'go

ahead' or 'front' (Lao, Samoan, Kayardild, Ewe), FEEL with 'taste', 'smell'

or 'hold an opinion' (Acehnese, Ewe, French, Mandarin, English), and

BECAUSE with 'from' (Yankunytjatjara, Arrernte).

These polysemies are usually not difficult to understand in the light of

language-internal semantic analysis. For example, 'appearing' and

'arriving' both involve something HAPPENING IN A PLACE, after which something

or someone is in the place in question. In the case of 'appearing',

there is presumably an additional component involving being 'able to

see' something, and in the case of 'arriving' there is an additional component

involving prior motion. Hence it is not surprising that a single

form can express both the prime meaning HAPPEN, and some additional

meanings based on HAPPEN.

More difficult cases fall into three kinds: (i) apparent gaps, for example

languages apparently lacking any word for a proposed prime, such as

PART, (ii) apparent underdifferentiation, that is a word which seems to be

general across two or more primes, for example a single form used for

both SAY and DO and THINK, (iii) apparent overdifferentiation, that is

when two or more words apparently divide up the meaning of a supposed

unitary prime, for example two words for different kinds of FEELING.

We will look at examples of these problems in turn.

1.2.2 Apparent 'gaps'

Example [1]: No exponent for TIME in Hopi?

In some cases a gap is reported, but in reality there is a perfectly serviceable

exponent present. For example, contrary to Whorf's (1956) assertions

that Hopi is a 'timeless' language, there is good evidence, in Malotki's

(1983) extensive study, that Hopi has exponents of all the proposed NSM

temporal primes. I will here illustrate simply with the prime WHEN/TIME,

but exponents of other temporal primes such as BEFORE, AFTER, FOR SOME

TIME, A LONG TIME, are also present (cf. Goddard in press). Of course, one

would not expect to find in Hopi any lexeme which is fully equivalent to

English time, with an identical range of language-specific polysemic

meanings and uses, for example its various uses as an abstract noun (for

example we didn't have time, time flies, times have changed), and its role in

phrasemes such as a long time, in compounds such as lunchtime, and so on.

What matters from an NSM point of view is only whether there are

Hopi lexical units with the meaning WHEN/TIME in a narrow range of basic,

and putatively universal, combinations such as (I DON'T KNOW) WHEN IT

HAPPENED, IT HAPPENED AT THIS TIME, AND THEY DID IT AT THE SAME TIME.

On this understanding, the Hopi equivalent to WHEN, both as an indefinite

and as an interrogative, is hisat, as shown in (5). Morphologically

hisat is analysable as a question formative hi- (much like English wh-)

and -sat TIME. In particular, -sat TIME can combine with the demonstrative

yàa- THIS to form the expression yàa-sat AT THIS TIME as shown in (6).

An allomorph of -sat, namely -saq, combines with the Hopi exponent of

THE SAME suu-/sú-, to form expressions meaning AT THE SAME TIME, as in (7).

Examples from Malotki (1983):

(5) Pam hisat nima?

that when go home

When did he go home? (Malotki 1983: 305)

(6) Taavok yàa-sat haqam ay nu' tsöng-moki.

yesterday this-time APPROX ASSR I hunger-die

Yesterday at about this time I got really hungry. (Malotki

1983: 146)

(7) Pam sú-'inùu-saq nakwsu.

that the same-I-time start out

He started out at the same time as I. (Malotki 1983: 144)

Example [2]: Languages without a distinct form for PART

Linguists seem to agree that the part-whole relationship is fundamental

to the vocabulary structure of all languages, but there certainly are

languages which do not have a unique lexical form for the postulated

semantic prime PART (OF). This does not necessarily mean, however, that

these languages lack a lexical unit with the meaning PART. In three unrelated

languages in which such an apparent gap has been investigated

(Acehnese, Mangaaba-Mbula, Yankunytjatjara) it appears that PART exists

as the meaning of a lexical unit of the same lexeme which can also mean

SOMETHING, THING or WHAT.

In these languages the meaning PART is expressed when the relevant

lexical form is used in a grammatical construction associated with possession.

(It is as if instead of saying, for example, 'the nose is a part of the

face' one says 'the nose is a thing of the face'.) Examples follow.

[Yankunytjatjara, Australia]

(8) Puntu kutju, palu kutjupa-kutjupa tjuta-tjara.

body one but something many-having

(It is) one body, but with many parts.

[Acehnese, Indonesia]

(9) Bak geuritan angèn na lè peue.

at vehicle wind there is many what/something

A bicycle (lit. wind-vehicle) has many parts.

[Mangaaba-Mbula, Papua New Guinea]

(10) Iti tomtom na koroN-Na-nda boozo, kumbu-ndu, nama-nda

we person GIV thing-NMZ-our many leg-our head-our

We people, our parts are many: our legs, our heads,...

The fact that the notion of PART is expressed using the same lexical form

as for SOMETHING is obviously no coincidence, since a PART of something is

itself a something (the notion PART can be termed a 'relational

substantive'). The absence of a unique term for PART is certainly notable,

and it may indicate, in an indexical sense, that the notion of PART is less

culturally salient than it is in languages which have a unique exponent

for PART.

Even so, the fact remains that there are Acehnese, Mbula and

Yankunytjatjara expressions (lexical units) with the meaning PART.

1.2.3 Apparent underdifferentiation: that is where a word

appears to be general across two or more primes

Example [1]: DO and SAY in Samoan

There are languages in which the same lexical form can express both SAY

and DO, or SAY and WANT. That is, there are languages in which there is

polysemy between SAY and DO, or between SAY and WANT. It goes without

saying, of course, that polysemy should never be postulated without

language-internal evidence and analysis. As an example of such evidence,

consider the situation with Samoan (Mosel 1994), in which the

verb fai can express two meanings - SAY and DO. The two meanings are

associated with different morphosyntactic properties. Fai SAY is a

non-ergative verb, selecting an absolutive subject, as in (11a) and (11b):

(11a) Ona toe fai atu lea 'o le fafine, 'Se...

then again say DIR then ABS the woman friend

Then the woman said again, 'Friend,...' (Mosel 1987: 459)

(11b) Na e fai mai au oti?

PAST you say hither PERF die

You said he has died?

Fai DO, on the other hand, selects an ergative subject, as in (12a) and (12b)

below. As well, fai DO often occurs in the so-called long (suffixed) form

fai a, which is usual when an ergative verb is preceded by a pronoun,

even when fai DO is used in a non-transitive frame, as in sentence (12b).

(12a) ...'ua fa'apênâ lava ona fai e le tama.

PERF like this EMPH that do ERG the youth

...the youth did it like this. (Mosel 1987: 122)

(12b) 'O ai na faia?

PERF who PAST do?

Who did it

Example [2]: Hyperpolysemy in Bunuba

For a more extreme example, we can take Bunuba, a language from the

Kimberley region of north-west Australia (Knight forthcoming; for

a contrary analysis, see Rumsey 1990, 2000). Polysemies involving SAY,

DO and sometimes other elements, are common in the languages of this

area. In Bunuba, a simple use of the root ma, with a 3rd singular subject,

is no less than five ways ambiguous - between 'do', 'say', 'think', 'feel'

and 'happen'.

(13) Ngaanyi ma ø-miy?

what? I/I 3SGS-MA: PAST

What did she do/say/think? or What happened?

The meaning HAPPEN has very distinctive syntactic restrictions, which

make it easy to separate it from the others. Specifically, it only ever takes

3SG or 3NSG subject (indicating 'something', as in 'something [3SG]

happened', or 'things', as in 'some things [3NSG] happened'). When an

extra argument (say, X) is added by way of OBLique pronominal crossreferencing,

the meaning corresponds to 'something happened to X'.

For reasons of space, I will report here only Knight's (forthcoming)

arguments regarding the status of the SAY-THINK-FEEL polysemy. To begin

with, she observes that 'speakers of Bunuba have no difficulty in discerning

one meaning from another'. She then notes that the meaning

ma SAY is associated with certain unambiguous syntactic contexts, as

when it introduces a direct quote, as in (14). The meaning SAY also

appears unequivocably when an extra argument is added to ma via the

OBLique cross-referencing strategy, in which case the new argument

assumes the role of addressee (or, if context permits, topic3):

(14) Yaninja wau wurr-ma-iy-nhingi.

alright whoa 3NSGS-MA-PAST-3SGOBL

Alright, 'Whoa!' they said to him.

The existence of a real language-internal contrast between SAY and THINK

is dramatized when the question is raised: How, in Bunuba, could one

express a meaning such as 'I know what you said but what are you

thinking?' When Knight explored this question with consultants, it

emerged that the expression ma thangani 'mouth/words' means

unambiguously SAY and that ma gun.gulu 'head' means unambiguously

THINK.

(15) Ngayini binarri nganggu thangani

1SGPRO know 2SGOBL mouth/words

nganggu gun.gulu nginjaga gi-nj-i-ma?

2SGOBL head what PRES-2SGS.NONFUTURE-INS-MA

I know what you said (your words) but what are you thinking

(lit. ma 'head')?

It further emerged that the meaning FEEL can be unambiguously

expressed by means of a similar compound term: ma guda 'stomach'

(and this term is not confined to physical sensations, let alone to visceral

ones). Expressions like the following are therefore unambiguous;

and the existence of these expressions testifies to the existence in

Bunuba of SAY, THINK and FEEL as meanings of lexical units.

(16a) Ngaanyi ma ø-ma-iy thangani?

what? I/I 3SGS-MA-PAST mouth

What did she say?

(16b) Ngaanyi ma ø-ma-iy gun.gulu?

what? I/I 3SGS-MA-PAST head

What did she think?

(16c) Ngaanyi ma ø-ma-iy guda?

what? I/I 3SGS-MA-PAST stomach

What did she feel?

1.2.4 Apparent 'overdifferentiation'

Example [1]: Two words for 'thing' in Japanese

Anna Wierzbicka has long insisted that SOMETHING is a semantic prime,

while allowing that in English (as in many languages) this meaning has

a combinatorial variant (namely thing), which occurs in combination

with specifiers and quantifiers. Some languages, however, appear to distinguish

'concrete things' from 'abstract things', thus drawing into question

the unitary nature of the proposed prime SOMETHING. For example,

Japanese has two words - mono and koto - which are normally used

about concrete and abstract things, respectively. Wierzbicka (1998b)

argues, however, that both terms can be regarded as allolexes, that is

partially conditioned lexical variants, of SOMETHING.

To begin with, she notes that SOMETHING does have a neutral exponent

in Japanese - the indefinite/interrogative nani - which covers both things

called mono and those called koto. Closer investigation shows although

mono is normally used for concrete (that is visible and tangible) things, in

some contexts it can be used for invisible things as well. For example, the

phrase seishin-teki na mono 'spiritual things' can refer to 'things' such as

kokoro 'heart', ki 'spirit', tomashii 'soul', kimochi 'feelings' and ai 'love'

(Yuko Asano, pers. comm.). Mono can also be used in sentences like the

following: 'I want only two things (mono): a job and a car.' The word for

'job' is shigoto (where -goto is a variant of koto), yet there is no problem with

including shigoto under mono in this context. Thus, while mono is normally

used with reference to 'things' that one can see, it could not possibly be

defined as 'something that one can see' (obviously, one cannot see a job).

Similarly, although koto is normally used with reference to 'things' that

one cannot see, it could not possibly be defined as 'something that one

cannot see'. For example, a sentence like Honto ni ii koto o shimashita ne

'You have really done something good' (Alfonso 1966 [1971]: 402) does

not mean 'you have really done something good that can't be seen' (that

is 'invisibility' is not part of the sentence's intended meaning).

Wierzbicka (1998b) concludes from this that mono and koto are in fact

two specialized allolexes of the same semantic prime, whose primary

exponent is the neutral indefinite/interrogative form nani SOMETHING.

Example [2]: Two words for FEEL in Chinese

As a second example, we can take the case of FEEL in Chinese. In Mandarin

Chinese (Chappell 1994: 118-20; 2002), there are two apparent equivalents

to FEEL - g˘andào and g˘anjué. Both can take an equi-clause pattern, in

the sense that they combine with a following same-subject verb: X

g˘andào /g˘anjué Verb. G˘andào typically combines with a second predicate

containing a stative verb designating a sensation of temperature (hot, cold,

warm and so on) or an emotion (delighted, ashamed, sad, angry and so on);

see example (17). The noun g˘anjué is restricted to combination with the two

evaluators - h˘ao GOOD and huài (bù h˘ao) BAD - to code statements about a

person's general well-being, physical and emotional; see example (18).

(17) D¯ang w˘o t¯ıngdào nàjiàn shì de sh¯ıhou w˘o g¯andào

when 1SG hear that: CL matter LIG time 1SG feel

hˇen g¯aoxìng.

very happy

'When I heard that, I felt very happy.'

(18) W˘o j¯ınti¯an g˘anjué tìng h˘ao de.

1SG today feeling very good ASST

'Today I'm feeling very well.' [for example a response to inquiry

about one's health]

The two words thus appear to divide up the labour between specific and

general feelings. On the basis of facts such as these, it has been claimed

(Shi-xu 2000) that Chinese lacks a precise exponent of FEEL.

As Chappell (2002) shows, however, it is possible to use ganjué both

about physical well-being, as in (18), and also about inner states of wellbeing,

as in (19). Note that given its status as a noun, g˘anjué FEEL tends to

occur in an S-V predicate forming a double subject construction.

(19) W˘o g˘anjué bù huì hˇen h˘ao.

1SG feeling NEG likely very good

'[if this happens] I won't feel anything very good.'

Notwithstanding the existence of the more specialized item g˘andào,

Chappell (2002) therefore concludes that g˘anjué is the Mandarin exponent

of the prime FEEL, which is neutral between sensation and emotion.

1.2.5 Lexico-semantic universals in perspective

These abbreviated descriptions of apparent problem cases have been

presented to illustrate some of the issues which arise in matching exponents

of semantic primes across languages. It would be unfortunate,

however, if in so doing I conveyed the impression that matching semantic

primes across languages is profoundly problematical. On the contrary,

the lesson of numerous studies is that such difficulties are

sporadic, and that they have solutions.

The robustness of the NSM primes, as candidates for lexico-semantic

universals, is dramatized when a comparison is made with other nonprime

meanings which have been advanced in anthropological, psychological

or philosophical literature as universals of human experience or

cognition. Goddard (2001c) surveyed the cross-linguistic status of a

selection of non-prime meanings - such as 'man', 'woman', 'mother',

'head', 'ear', 'hand', 'hot', 'sweet', 'tree', 'water', 'sun', 'wind', 'go', 'sit',

'eat', 'give', 'hit', 'angry' - comparing these with a comparable number

of proposed semantic primes. From even a small sample of languages it

is clear that many impressionistically 'basic' items of English vocabulary

(such as go, water and eat) lack exact equivalents in other languages. Of

48 non-prime meanings surveyed, only the following 12 seem to have

any real hope of universal status (not necessarily as separate words, but

as the meanings of lexical units): 'man', 'woman', 'child', 'mother',

'head', 'eye', 'ear', 'nose', 'hand', 'day', 'kill' and 'make'. In general, however,

doubts remain even about the universality of these, doubts

28 Contrastive Analysis in Language