LINGUIST List 15.686

Tue Feb 24 2004

Review: Lexicography/Semantics: Murphy (2003)

Editor for this issue: Naomi Ogasawara <naomilinguistlist.org>


What follows is a review or discussion note contributed to our Book Discussion Forum. We expect discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for review." Then contact Sheila Dooley Collberg at collberglinguistlist.org.

Directory

  1. Ludwig Fesenmeier, Semantic Relations and the Lexicon

Message 1: Semantic Relations and the Lexicon

Date: Tue, 24 Feb 2004 18:09:21 -0500 (EST)
From: Ludwig Fesenmeier <ludwig.fesenmeieruni-koeln.de>
Subject: Semantic Relations and the Lexicon

AUTHOR: Murphy, M. Lynne
TITLE: Semantic relations and the Lexicon
SUBTITLE: Antonymy, Synonymy, and Other Paradigms
PUBLISHER: Cambridge University Press
YEAR: 2003

Announced at http://linguistlist.org/issues/14/14-2155.html


Ludwig Fesenmeier, Department of Romance Languages, 
University of Cologne 

PURPOSE AND CONTENTS OF THE BOOK

The purpose of the monograph is to offer a new, cross- disciplinary
approach to the traditional semantic relations discussion (henceforth:
SR's), especially concerning antonymy and synonymy. The author claims
that the relations themselves form part of the speaker's concepts of
the words they address and that they thus share a role in
instantiating meaning. She further claims that it is possible to
identify a single principle underlying these different relations.

The book is divided into two parts. Part I, consisting of three
chapters (pp. 1-129) deals with SR's in general. Part II (pp. 131-242)
discusses in detail the different SR's: three of its four chapters are
dedicated to the relations of synonymy, antonymy, and, as a catch-all
category of their own, hyponymy, meronymy and others respectively; the
last chapter returns briefly to the assumptions proposed at the
beginning of the book. At the end there is an appendix of ''Relation
elements'' (pp. 243f), the notes (pp. 245-253), the references
(pp. 254-273), and an index (pp. 274-292) of topics and authors.

Chapter 1 provides an overview of the book's topics and goals, and of
the concepts and terminology relevant for the discussion. The author
takes a pragmatic and psycholinguistic perspective, where
''pragmatic'' implies ''that the linguistic phenomena described [...]
are considered with reference to their use and their status in a human
mind within a human culture'' (p. 5), while the label
''psycholinguistic'' refers to ''a psychologically plausible model of
the knowledge and processes involved in semantic relations phenomena
in human language behaviour'' (p. 4f). As far as terminology is
concerned, the term ''intralexical'' is introduced to indicate ''that
a structure or piece of lexical information is contained within the
[mental] lexicon'', while ''metalexical'' refers to ''information that
is not contained in the [mental] lexicon, even though it may be
information about words'' (p. 9). Furthermore, various approaches to,
and models of, the concept of 'mental lexicon' are evaluated, in
particular those based on the ''dictionary metaphor'' and the
''thesaurus metaphor''. Finally the author proposes a distinction
between two types of knowledge related to words: ''lexical and
conceptual representation of words'' (p. 23).

Chapter 2 presents in detail the metalexical approach. Murphy firstly
presents nine properties of SR's which a theory must account for:
productivity, binarity, variability, prototypicality/canonicity,
semi-semanticity, uncountability, predictability, and universality. In
the following section the ''Principle of Relation by Contrast''
(henceforth RC-P) is introduced which is defined as follows: ''The
contrast relation holds among the members of a set iff: they have all
the same contextually relevant properties but one.'' (p. 44). It is
assumed that all types of SR's can be derived from this
principle. Discussion in the chapter goes on to reconsider six kinds
of phenomena - which can be observed in everyday language use and
which are related to paradigmatic relations - in the light of the
metalexical approach (among others metaphorical use, language
acquisition, stylistic competence).

Chapter 3 provides a survey of approaches to SR's from a number of
different disciplines - this in constant comparison with the author's
own metalexical approach. After a historical overview of work done in
philosophy, linguistics, anthropology, psychology, and computer
science, the sections that follow discuss approaches which Murphy
divides in three categories: the two approaches following the
dictionary metaphor and the thesaurus metaphor respectively, the third
category falling between these two models. The chapter concludes by
discussing ''approaches to the conceptual status of the semantic
relations themselves'' (p. 61).

Chapter 4 addresses ''synonymy and similarity'' (title) which is
firstly described from the author's metalexical perspective, thus
being considered ''a relation between our conceptualizations of words,
rather than between their lexical entries [in the mental lexicon]''
(p. 134). Adopting the RC-P, it follows that a synonym ensemble
''includes only word-concepts that have all the same contextually
relevant properties, but differ in form''. This she calls ''Relation
by Contrast - Synonymy'' (henceforth RC-S; p. 134). Murphy continues
by discussing the aspects of identity, similarity, and difference,
showing how RC-S allows for deriving ''context-dependent synonyms,
rather than logical synonyms'' (p. 143). Another section treats the
question of defining which specific properties of the words concerned,
can reliably be said to reify a synonymy relation (denotation,
connotation, etc.). The following section is dedicated to the logical
properties of synonymy (reflexivity, symmetry, and so on), showing
that such properties ''do not hold in natural languages instances of
synonymy'' (p. 157). Finally, attention is paid to the effects
synonymy might have on the vocabulary of a language.

Chapter 5 deals with ''antonymy and contrast'' (title), paying special
attention to ''why contrast involving semantic incompatibility is so
central a semantic relation'' (p. 169). In the metalexical approach,
opposition and contrast are accounted for by the principle of
''Relation by Contrast - Lexical Contrast'' (henceforth RC-LC),
discussed in the first section: ''A lexical contrast set includes only
word-concepts that have all the same contextually relevant properties
but one'' (p. 170). In what follows, some properties of antonymy that
are traditionally discussed in literature (binarity, symmetry, and
markedness) are reconsidered in the light of this principle and it is
shown how antonym subtypes can be accounted for from this point of
view. Furthermore the author discusses antonymy and contrast as far
as, among others, discourse functions, semantic change, and language
acquisition are concerned.
 
Chapter 6 discusses in particular the relations of hyponymy/hyperonymy
and meronymy/holonymy, and a case is made for considering these less
as ''relations among word- concepts rather than relations among the
things (the concepts) that those words denote'' (p. 216).

Chapter 7 gives a brief summary of the topics addressed in the
preceding chapters, considering possible counter- arguments to the
metalexical approach and questioning its compatibility with different
models of the lexicon. The lexicon is examined both as being more or
less of modular nature as well as analysed as part of a gradual
continuum towards ''grammar'' at the other end of the scale
(Construction Grammar, Functional Grammar, among others).

CRITICAL EVALUATION 

One of the merits of Lynne Murphy's book is that it clearly lays out
the problems intrinsically connected to the study of semantic
relations: in fact, much of the work on this topic seems to be
grounded on what she calls ''knowledge about words'', clearly
distinguished from what the author terms ''knowledge of words''. This
distinction is sometimes not so straightforwardly drawn in
literature. The lack of awareness of this difference might be one of
the reasons for the frequent co-occurrence of ''problem'' with
''synonymy'', ''antonymy'' and the like. Nevertheless, the distinction
between various types of knowledge might have been further refined by
considering the work done by Hans- Martin Gauger (see Gauger 1970,
1972 and in particular 1976) and Eugenio Coseriu (e.g., Coseriu 1973,
1988) who discuss in detail various aspects of ''types of linguistic
knowledge''; furthermore, Andreas Blank proposed a rather detailed
model of the relationships between levels of knowledge and levels of
linguistic meaning (see Blank 1997, in particular pp. 54-96).

Another great advantage of the study under review resides in its
taking into account the results of a broad range of disciplines often
not discussed in more ''traditional'' (especially linguistic) accounts
of SR's.

What ought to be highlighted also as an important point is that
Murphy's metalexical treatment of SR's includes the speakers
themselves in the discussion, since SR's are sometimes considered
either a phenomenon of language itself or a mere instrument for
description (e.g., Berejan 1971, 129 and Harris 1973, 1 respectively
for synonymy). In fact, some of the properties of SR's discussed by
Murphy are intimately bound to speakers' judgements (e.g.,
productivity, variability, prototypicality, semi- semanticity). As far
as these properties are concerned, it must be noted, however, that
they are of different status: some bear on properties of the relations
themselves (uncountability, universality), some concern rather the
entities involved (variability, semi-semanticity), some others still
seem to be of interest for antonymy only (binarity, prototypicality).

The metalexical treatment further demands reconsidering how far the
role of denotation has to be treated in the study of SR's (see, e.g.,
Casas G�mez 1999 who discusses mainly this point).

As Murphy rightly argues, it is beyond doubt that the common
denominator of SR's is similarity between the entities related. Who
says ''similarity'' necessarily also says ''difference''. The RC-P is
thus a quite appealing idea in that it offers a unified account of the
different SR's, defining them as ''relations on the basis of minimal
difference'' (p. 44) and thus as being based on a quantitative
criterion rather than on a qualitative criterion. Yet it seems to me
that one can likewise make a strong case for claiming that especially
with synonymy and antonymy (I leave out of discussion
hyponymy/hyperonymy and meronymy/holonymy for the reasons exposed in
chapter 6 of the monograph) there is a fundamental qualitative
difference which challenges notably the possibility of a unified
account. It might be formulated in the following way: (a judgement of)
synonymy is based on similarity IN meaning, while (a judgement of)
antonymy is based on similarity OF meaning (see also p. 43 (my
capitals): ''In the case of synonymy, words are expected to be similar
IN meaning. Antonymy also requires similarity OF meaning.''). That is
to say that the tertium comparationis in synonymy is, roughly
speaking, a single entity (a single ''meaning'' or ''concept'' or
whatsoever) which is at least partially ''contained'' in the related
words themselves, whereas in the case of antonymy it is a scale (e.g.,
''warmth'', ''length'') or some other complex (e.g., ''pets'' in the
case of ''dog - cat'') to which the related words are to be referred
to, and it is precisely the difference concerning the tertium that
RC-P, at least in its actual version, does not account for. This is
not to deny a priori the possibility that there might exist a common
principle underlying (judgements of) both synonymy and antonymy (and
other SR's), but it seems that further research is needed to prove the
case, Murphy's approach being an important step.

Another aspect that needs to be emphasised is the fact that Murphy is
right in insisting on the importance of contextual factors in
discussing SR's, showing how different contexts (in a rather broad
sense, including sentential contexts as well as somewhat artificial
ones like thesauri and word-association tests) allow for different
judgements on the synonymy or antonymy of the words under
discussion. One however must be careful not to blur the borders
(thereby compromising the operationality of the concepts) that
separate what one would reasonably still consider as an instance of a
'nym-relation from attributes no longer compatible with such a
judgement. If we consider for example what is discussed on pp. 148f:
although RC-S allows for treating ''to punish'' as a synonym of ''to
correct'', ''to chastise'', ''to discipline'', ''to castigate'', and
''to penalize'', ''since for many purposes it is a reasonable
substitute (and thus similar enough to) any of the others'', The
American Heritage Dictionary on the contrary suggests it as being
''the 'least specific' of the words''. One might in fact find it more
natural to consider the relation between ''to punish'' and its
neighbours as a relation of hyperonymy/hyponymy.

Still more counterintuitive are ex. 3 on p. 28 (''brown'' rather as an
antonym of ''blue'' than of ''red'') and ex. 12 on p. 142 (''doggy''
as a synonym of ''kitty''), where the relations are simply posited by
the speaker in the concrete situation, and where it is quite easy to
imagine the judgements as being the other way round, thereby depriving
the possibility of distinguishing between antonymy and synonymy. Thus,
as a reader of Murphy's theses, one has to bear in mind that a set of
words one finds for example in dictionaries of synonyms and antonyms,
will fulfil the requirements of RC-S or RC-LC, but that the contrary
will not necessarily hold. So, the principles as proposed by Murphy
seem too strong in their predictive power, and the idea that SR's are
to be accounted for as categories which show some kind of
prototypicality effects (linguists actually limiting their attention
normally to the more/most prototypical cases) must not lead to
assigning them shadowy status.

In conclusion, despite the problems discussed so far, Murphy's book
sheds light on the important challenges and interesting insights which
the matter of SR's are still able to offer, even if such relations
have been subject of interest for centuries.

REFERENCES

Berejan, Silviu (1971): ''� propos de la d�limitation des unit�s
synonymiques dans un champ conceptuel'', in: Revue roumaine de
linguistique, 129-134.

Blank, Andreas (1997): Prinzipien des lexikalischen Bedeutungswandels
am Beispiel der romanischen Sprachen, T�bingen: Niemeyer.

Casas G�mez, Miguel (1999): Las relaciones l�xicas, T�bingen:
Niemeyer.

Coseriu, Eugenio (1973): Sincron�a, diacron�a e historia: El
problema del cambio ling��stico, 2. ed., Madrid: Gredos.
 
Coseriu, Eugenio (1988): ''Die Ebenen des sprachlichen Wissens. Der
Ort des 'Korrekten' in der Bewertungsskala des Gesprochenen'', in:
Albrecht, J�rn/L�dtke, Jens/Thun, Harald (eds.) (1988): Energeia
und Ergon. Sprachliche Variation - Sprachgeschichte - Sprachtypologie,
Studia in honorem Eugenio Coseriu, T�bingen: Narr, vol. 1, 327-375.

Gauger, Hans-Martin (1970): Wort und Sprache, T�bingen: Niemeyer.

Gauger, Hans-Martin (1972): Zum Problem der Synonyme. Avec un
r�sum� en fran�ais, T�bingen: Narr.

Gauger, Hans-Martin (1976): Sprachbewu�tsein und Sprachwissenschaft,
M�nchen: Piper.

Harris, Roy (1973): Synonymy and linguistic analysis, Oxford:
Blackwell.

ABOUT THE REVIEWER 

Ludwig Fesenmeier teaches Romance linguistics at the Department of
Romance Languages, University of Cologne, currently working on his
post-doctoral thesis on lexical synonymy in the Romance languages.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue