LINGUIST List 11.1213
Tue May 30 2000
Review: Weigand: Contrastive Lexical Semantics
Editor for this issue: Andrew Carnie <carnielinguistlist.org>
What follows is another discussion note contributed to our Book Discussion
Forum. We expect these 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 discussion." (This means that
the publisher has sent us a review copy.) Then contact Andrew Carnie at
carnie
linguistlist.org
Directory
Lynne Murphy, review of Weigand (ed.) - Contrastive Lexical Semantics
Message 1: review of Weigand (ed.) - Contrastive Lexical Semantics
Date: Mon, 29 May 2000 18:12:16 +0100
From: Lynne Murphy <lynnemcogs.susx.ac.uk>
Subject: review of Weigand (ed.) - Contrastive Lexical Semantics
Weigand, Edda (ed.) 1998. Contrastive Lexical Semantics. (Current
Issues in Linguistic Theory, 171.) Amsterdam: John Benjamins. 268pp.
Reviewed by M. Lynne Murphy, University of Sussex
Toby Ayer already reviewed this book in Linguist 10.935, and I do not
disagree with him in his description and evaluation of it. Thus, I do
not here thoroughly describe the contents of all the chapters; the
reader is urged to turn to Ayer's review for such detail. Instead, I
briefly consider the editorial approach to the book and then discuss the
claims made in some of the stand-out chapters.
The book consists mostly of papers from a conference on '(Contrastive)
Lexical Semantics' at the University of M�nster in May 1997 (according
to the editor, some additional papers have been added for the book, but
it's not clear which). All of the 13 papers address lexical meaning in
European languages, with German and English being most discussed. Not
surprisingly, the book has a very European theoretical flavour. As
noted by Ayer, this book is 'probably not for those accustomed to
generative approaches.' Its prevalent assumption is that lexical items
must be studied in use. This is to say that the approaches are from a
pragmatic perspective and often based on corpus-derived data. For the
most part, meaning is approached here as interactional convention and
contextual construct, rather than as mental representation in an ideal
speaker.
Like many books derived from conference papers, the progression of
chapters is not always coherent. For instance, the lone computational
linguistics chapter seems quite out-of-place, while the group of papers
on emotion terms is often repetitive. The book as a whole might have
benefited from more editorial effort at presenting the collection as an
anthology, rather than as a proceedings--it seems to be half way between
the two. For example, some groups of papers might appropriately be
considered a 'section' of the book (e.g., the three articles on emotion
terms), and it would have been suitable to have some forward to such
sections. As Ayer also mentioned, the need for copy-editing in some
places (especially of non-native English) is quite apparent. (For
example, on page 166, 'and' is spelt 'et'!)
Three of the chapters stood out for me in their arguments for particular
approaches to lexis. The first of these is John Sinclair's 'The
Lexical Item' (1-24), in which he demonstrates the inadequacy of the
traditional assumption that a word is an autonomous, meaningful unit and
that an autonomous lexicon is composed of such words. Sinclair is
correct when he points out that lexical semantics has traditionally been
over-interested in paradigmatic relations to the neglect of syntagmatic
patterns. Certainly, the latter that is more relevant to any theory
that purports to model sentence production. His demonstrations include
examples of words whose meaning is different in certain contexts than in
isolation (e.g., 'white' in 'white wine'), and cases of semantic prosody
(with a thorough look at the negative prosody of 'budge'). Sinclair
does not provide a new model of the lexicon, but suggests that thinking
of words as lexical items unto themselves may not be the right approach.
He concludes that the way to better understanding of words' syntagmatic
patterns is through computational analysis of large text corpora.
Sinclair, here and elsewhere, has proved the strength of corpus
investigations as discovery tools. While corpora have certainly
demonstrated their worth in debunking assumptions about the lexicon and
its contents, it's less clear that they serve as hatching grounds for
new models.
The next article is Edda Weigand's 'Contrastive Lexical Semantics'
(25-44). Weigand starts with two assumptions: that 'we have left
structural linguistics with its autonomous areas' and that 'our
reference point can no longer be the competence of an ideal speaker' (p.
25). I have to admit here that I hadn't realised 'we' had done this--I
guess that W wasn't expecting people like me as an audience. Like
Cognitive Linguists (although not identifying herself as a part of that
program), W holds that lexical fields must be rooted in the physical and
cognitive abilities of humans. She identifies six universal predicating
fields (AWARENESS, MOTION, ACTION, RATIONALITY, EMOTION, SOCIAL
BEHAVIOUR), which are 'intended to comprehend all the types of human
ability which are the basis for predicating' (p. 32). These in turn are
subdivided by particular realms of activity, in which semantic
structures are built up from predicating positions such as BE, BECOME,
LOSE, and CAUSE. According to W semantic primitives are vacuous, but
these 'atomic predicates' are not vacuous like primitives because they
are composed through the observation of human abilities, rather than
decomposed through analysis of words. This is where W and I must part
company. While I too am interested in a pragmatic perspective on word
meaning, I believe that as linguists the only reliable evidence we have
for the cognitive underpinnings of word meanings is words and their use.
To start with words and work backward toward the cognitive intangibles
is a process that I find more empirically driven. To start with the
realms of cognitive activity and move toward the observable (words)
seems a backward process, involving a number of untested assumptions--any
one of which could lead the theory astray. For example, this approach
seems to assume that the six semantic fields are separate and equal.
(But why should we assume this?) In her second paper in this collection
('The vocabulary of emotion,' 45-66), W puts the approach to use within
the realm of emotion terms, but immediately sorts out what belongs in
the EMOTION realm (e.g., JOY, SADNESS, FEAR, COURAGE, ANGER, ADMIRATION,
DISGUST) from cognitive attitudes that belong to other domains (GUILT
and TRUST belong to RATIONALITY, FRIENDLINESS and BROTHERLINESS belong
to SOCIAL BEHAVIOUR). An odd decision, from my perspective, is to
consider PAIN a part of EMOTION. Why not part of AWARENESS? Aren't
TRUST and ADMIRATION sometimes SOCIAL BEHAVIOUR? W's analysis seems to
entail that the application of TRUST to social interactions involves a
cross-field metaphor, or that there are two TRUSTS, one that is based in
rationality ('I can't put a lot of trust in the current market') and one
that is based in social interaction ('Getting along with your co-workers
is a matter of trust.'). While the selection of these six cognitive
fields was meant to ground meaning in experience, they instead act as
arbitrary constraints on the composition of meanings. While we might be
forced to decide in a decompositional process that there are two
meanings of 'trust', the six semantic fields of W's compositional
analysis make the decision for us. Thus, starting with the cognitive
and moving toward the linguistic does not free us from the arbitrary
categories of language^�it simply changes the identities of the
categories in keeping with a hypothetical (i.e., artificial) cognitive
structure.
Grounding meaning in experience is not W's only reason for choosing
composition' over 'decomposition'. The other is the well-motivated
desire for a semantic theory that is universally applicable. Since
decompositional approaches start with words and work back to meaning
representations, the results of these approaches represent meanings
particular to the language investigated. W's compositional approach is
meant to avoid this problem by starting with universals of human
cognition and building up to language-specific structures, thus noting
the universals and particulars along the way. However, this requires
independent motivation for the underlying cognitive architecture, and it
s both difficult to see how this can be achieved without using
linguistic behaviour as insights into cognition. W dismisses
out-of-hand the notion that a general theory of lexical semantics can be
achieved through lexical decomposition, but offers no reason why a
decompositional contrastive (or comparative) approach would not work.
This would involve attempting an analysis of more than one language with
a single metalanguage derived through observation of the languages in
question. This is, of course, the approach taken by Anna Wierzbicka
(1996 and elsewhere), and whether or not one agrees with her particular
method and results, Weigand has not provided a convincing argument
against such attempts.
Of course, the most basic problem in any cross-linguistic analyses of
meaning is the limited linguistic range of the analyst. Even more so
than other areas of linguistic investigation, semantic analysis has been
particularly dependent on the linguist's intuitions, and thus semantic
analysis is often limited to the languages in which the analyst is
(nearly) native. While ethnolinguistic field studies have provided some
methods for semantic investigation in unfamiliar languages, such work
has mostly explored limited semantic fields (e.g., kinship systems,
colour, disease). So the studies in this book are exciting because
they demonstrate how efficiently computer-driven corpus investigations
can highlight semantic contrasts among near-equivalents in different
languages.
Lastly, I turn to Claude Gruaz's 'Composition principles within the word
and within ways of use of words' (163-71). G starts with the pragmatic
principle that there is no one-to-one correspondence between form and
meaning for words, and wishes to extend this to (bound) morphemes, which
he does with a look at French diminutive suffixes. What interests me
here is G's interest in 'synchronic word families'--networks of words
that are formally and semantically related (usually due to etymological
relation). Lexical items can be closely related (or 'motivated' in G's
terms) through derivation, or loosely related, such as 'possible' and
impotent.' G does not discuss what role such families have in the
mental representation of language, but the discussion of word-families
brings me back to the question of what a word is and what a lexical
entry is. If a lexeme is an abstract category, which encompasses the
morphologically variant forms of a word (e.g., 'bring', 'brought'), are
word families mentally represented as further abstract categories
consisting of these abstract word categories? There would seem to be
evidence for this in antonym relations. Canonical antonym pairs
co-occur at far greater rates than chance or than synonymous
non-canonical oppositions (e.g., 'rise'-'fall', '(a)live'-'dead', as
opposed to 'ascend'-'fall' or '(a)live'-'deceased'). But Christian
Fellbaum (1995) has shown that canonical antonym co-occurrence patterns
extend beyond the two members of the canonical pair. Derivational
relatives of the members of canonical antonym pairs act like antonyms
even though they belong to different syntactic categories . So, for
example, the verb 'begin' co-occurs more regularly with the noun 'ending
than the verb 'start' does. It seems then that the antonym relation
holds here between word families rather than among lexical entries, and
thus we have evidence of some psychological reality for word family
organisation. I argue elsewhere that such word-family effects indicate
that at the metalinguistic (conceptual) level, words are conceptualised
as prototype categories (Murphy in prep., but see also Murphy forth.).
While G's article raised some interesting questions for me, it
unfortunately is just a very brief exploration of the issues and a
cursory introduction to his approach.
In conclusion, Contrastive Lexical Semantics raises several issues that
deserve further exploration, and it demonstrates the value of corpus
investigation for lexical semantics across languages. The articles
included are of varying interest and originality, and since the book
does not cohere as a whole, most readers will be interested in picking
and choosing among its offerings. The book doesn't quite have something
for everyone, though. I expect that it will prove particularly valuable
to those interested in machine translation for European languages,
because of its comparative and descriptive bent.
References
Fellbaum, Christiane. 1995. Co-occurrence and antonymy. International
Journal of Lexicography 8, 281-303.
Murphy, M. Lynne. forth. Knowledge of words versus knowledge about
words: The conceptual basis of lexical relations. In Bert Peeters
(ed.), The Lexicon/Encyclopaedia Interface. Amsterdam: Elsevier.
Murphy, M. Lynne. in prep. Semantic Relations and the Lexicon: An
extralexical approach. Cambridge: CUP.
Wierzbicka, Anna. 1996. Semantics: Primes and universals. Oxford:
OUP.
About the reviewer:
I am a Lecturer in Linguistics in the School of Cognitive and Computer
Sciences at the University of Sussex (Ph.D., University of Illinois,
1995). My current work concerns the relevance of paradigmatic
lexical-semantic relations to lexical organisation and meaning.
M Lynne Murphy
School of Cognitive and Computing Sciences
University of Sussex
Brighton BN1 9QH UK
lynnemcogs.susx.ac.uk