LINGUIST List 10.935

Wed Jun 16 1999

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 carnielinguistlist.org

Directory

toby ayer, toby_ayeryahoo.com

Message 1: toby_ayeryahoo.com

Date: Tue, 15 Jun 1999 12:00:15 -0700 (PDT)
From: toby ayer <toby_ayeryahoo.com>
Subject: toby_ayeryahoo.com

Weigand, Edda, ed. 1998. Contrastive Lexical Semantics, John Benjamins, Amsterdam. 268 pages hardback.

Reviewed by Toby Ayer, Oxford University

This book is a collection of papers from a conference in Munster in May 1997, published as volume 171 of Current Issues in Linguistic Theory. Three of the papers are printed in German with English summaries; others appear to be translations or at least non-native writing, and occasionally the clarity suffers from it. Because there are thirteen papers, this will mainly be a summary, with a couple of comments mostly at the end. When I can, I will use the authors' words to represent their positions. In the Foreword, Weigand emphasizes that lexical semantics is to be seen as 'part of a theory of language use,' thus largely pragmatic in nature. In order to verify semantic conventions with 'hard, measurable evidence' rather than native intuitions, most of the authors adopt a corpus-oriented approach. The goal is 'a new methodology for the new object of vocabulary-in-use,' which will account for lexical idiosyncrasies, multi-word lexical units, and varieties of word use that a model-oriented approach cannot handle. Much of the work is directly oriented towards improving dictionaries, both of one language and between languages. A basic model of contrastive analysis is set out by John Sinclair in the first paper. He points out the inadequacy of traditional theories, which divide the lexicon from the grammar, for dealing with the way word meanings change when they are combined in phrases. One example is the phenomenon of 'reversal,' in which the meaning of a word 'arises predominantly from the textual environment, rather than the item choice.' For instance, in white wine the color-word white refers to a wine-specific color, not the usual color white; and in the expression borders on Adj., the adjective means something like 'abnormal,' regardless of what adjective it is. Further, considering meaning in multi-word units, ambiguity is seen to be rarer than usually thought (it is 'created by the method of observation, and not the structure of the text'). Form and meaning are the same thing--a lexical item is one or the other, when seen relative to other forms or meanings. There seems to be a conflict between the paradigmatic and the syntagmatic when considering the choice of a lexical item, but the two can be reconciled by a theory that takes cotext into account. Sinclair proposes five categories of item co-selection, which are the components of the lexical item. They are the core, semantic prosody, collocation, colligation, and semantic preference. His example is the verb budge. This word is defined in positive terms, but almost always used with a negative sense. Only the core and the semantic prosody are obligatory; here the core is 'NEG budge' (with the usual narrow definition) and the semantic prosody includes a notion of frustration with the 'budging.' Weigand follows with two papers, first elaborating on the priorities and assumptions behind the 'contrastive lexical semantics' approach, then starting to apply it to expressions of the emotion 'anger'. Again, the importance of pragmatics, of conventions rather than rules, is stressed. It is such conventions that 'determine how words are used.' And again, semantic conclusions must be quantifiable, checked by 'the criterion of frequency' in the corpus. Contrastive goals are finally laid out here. Weigand assumes a 'quasi-universal semantic structure' that is the basis for deciding on equivalence between languages. There is 'no independent reality,' however the universal structure 'contains the ways by which human beings perceive the world.' Accordingly, the theory is conceived in terms of human abilities, social/cultural conditions, etc.: 'our way of life determines our way of using language.' The six 'universal predicating fields' are AWARENESS, MOTION, ACTION, RATIONALITY, EMOTION, SOCIAL BEHAVIOR. These are divided subsequently into partial fields, predicating positions, and meaning positions, the minimal units of meaning. While she dissociates the theory from others 'based on independent semantic primitives,' she posits a basis of 'heuristic predicating positions' like BE, BECOME, LOSE, CAUSE. Multi-word units, or 'words-in-use,' are again taken as basic to linguistic expression: 'syntactically defined phrases are the lexical units of our communicative competence.' The facts of word combination cannot be covered by standard rules. Semantically equivalent phrasal expressions in individual languages (i.e. collocations) vary by convention, not by rule. With words-in-use as the object of study, Weigand hopes to provide the basis for improved dictionaries and tools for foreign-language learning. Weigand's analysis of the emotion anger consists substantially of parallel lists of expressions in German, English, and Italian. They are arranged by the criteria of [moral], [intensification], [curbed], and [minimization]. Within these, the 'predicating positions' mentioned above are also used to divide the example expressions (e.g. CAUSE + [moral] anger). The lists show how equivalent senses are captured in the different languages. The conclusion here is that 'to learn a language is to know how words are used and what utterances are used in specific situations.' Ways of use vary between languages in complicated ways. Meaning cannot be explained completely, so any language cannot be explicated completely. Weigand reminds us again that we need corpora both to determine and to confirm the semantic conventions revealed in the contrastive analysis. The next three papers--by Christian Schmitt, Valerij Dem'jankov, and Henning Westhiede--continue with the same method of contrastive analysis. They too concern 'anger,' comparing respectively German/French/Spanish, German/Russian, and German/Dutch. Schmitt's and Dem'jankov's papers are in German. All give similar parallel lists to Weigand's, divided under the same headings and comparing two languages at a time. Schmitt emphasizes action-oriented description. 'The function of words does not consist in making objective and ontological statements about the world but in applying words in meaningful propositions,' according to the English summary of the paper. 'We have to determine the purpose of expressions and the role linguistic entities play in concrete speech acts.' Dem'jankov extends the metaphor of the 'word family' to that of a 'semantic village,' fancifully using concepts from Soviet history/society. In such a village, words have pragmatic 'jobs.' They can be 'rooted' in one village (morphologically), but 'dwell' in another (semantically). 'Both evolution and revolutions are possible.' While semantics is where a word lives, and pragmatics is both travelling (change in the word's use) and 'relations between an actual lexical meaning of a word and its 'inner form.'' The difficulties of determining the equivalence of expressions between languages is again stressed by Westheide. The only way to succeed is through knowledge of the cultures/societies whose languages are under investigation. Pragmatic rules determine the use of words. When considering Dutch and German, it is important to remember the general principle of 'avoid conflict' at play in Dutch. Likewise, foreign language students need to know that a language is 'a communicative system bound to a cultural environment.' Westheide includes information on his learner's German dictionary for Dutch students, which uses the contrastive method described above. Eckhard Hauenherm's paper, in German, compares a lexical field approach with a pragmatic approach employing the notion 'meaning position' and applies them to the verbs stehen (German) and stare (Italian). The lexical field approach covers only the 'regular uses' of the verbs, though they are used in ways that deviate from their core meanings. The same general approach as above is advocated again here. Espousing a principle of homology, Claude Gruaz's short paper extends the above methods to the level of morphemes within words; both situational and linguistic context is relevant to lexical choices. Gruaz distinguishes between different aspects of content: referent, sense, meaning, and signification. He defines motivation, the relation between root words and derived forms, and describes synchronic word families based on 'close' or 'loose' motivation. The difference between composition at the morphological and phrasal levels is that there is more choice in the latter case. Gruaz's emphasis, though, is that pragmatics shows that on both levels the one unit-one meaning relation is clearly wrong--the relation is many-many. J\252rgen Esser argues more radically for a notion of 'medium-independent word-form.' Given the noun show, the verb show, the orthographic form 'show', and the phonological form //, he posits a single form <show> which is not bound to phonology, graphology, or grammatical category. This abstraction is analogous, says Esser, to positing a single root that underlies words such as find and found. This medium-independent form is important for analyzing homonyms and homophones, comparing written and spoken corpora (especially when accents/dialects/spelling conventions play a part), and for composite corpora. Statistical comparison of word-forms, indeed, is only possible using this notion. Focusing on Italian, Christoph Schwarze explores two instances of lexical variation: numerals used as approximate quantifiers (e.g. quattro salti 'four jumps' used to mean 'a little dance') and polysemous spatial prepositions (e.g. sopra/su 'on' used in various metaphorical expressions). He points out that 'variation of interpretations which only depend on situations are not to accounted for by the grammar.' He proposes that there are basic meanings that comprise the semantic structures of the various words, and the other readings are derived via a few principles (referential tolerance, transfer to non-spatial orientations, etc.). The goal is to conceive of a lexicon that can be both limited and variable, and the key is to recognize that the variation itself is constrained. We find a different kind of focus in Anita Steube and Andreas Spath's paper on Russian partitive constructions. Assuming a 'Chomskian grammar enriched with a semantic component' (i.e. Semantic Form), they argue for certain lexical entries that have no phonology. In the case of partitives and pseudo-partitives, they propose that the quantifier (e.g. 'much') is always present, but is non-overt if a measure phrase (e.g. 'three cups') is also used. The partitive PP modifies a head noun, which itself is non-overt. Thus the pseudo-partitive and partitive constructions are respectively 'three pots MUCH of strong tea' and '[three pots MUCH] [STRONG TEA of the strong tea],' where the capitalized words are non-overt. Full SFs of the elements in these constructions are given, and the amalgamation process is explained. Dmitrij Dobrovol'skij contrasts a Russian and a German idiom that are usually thought equivalent. The fact that they do not really match in interpretation supports the position that 'absolute cross-linguistic equivalents are, as a rule, lexicographic fictions which result from inaccurate semantic analysis.' To continue, he examines the Russian idiom 'cover one's self with a copper basin' and gives about a dozen rough equivalents in German, exploring various choices that affect the exact meaning of the idiom. He concludes that many contextual variations are crucial for choosing an appropriate equivalent idiom in another language. The final paper, by Wolf Paprotte, reports on an experiment in 'sense tagging' a text. The method of 'word sense disambiguation' attempts to use only the local context of each word to assign it the correct sense. Assuming that each local context will correspond to a single sense, this experiment first identified consistently co-occurring words ('relevant collocators') in a training set. By then identifying a second layer of collocators, 95% coverage of an experimental text was achieved. Using such 'signatures' for word senses is a new approach to defining 'similar contexts.'

Many of the papers here seem to overlap considerably with each other, covering different data though not advancing distinct positions. Clearly much of the work is actively directed towards practical improvements in dictionaries and learning tools, which is commendable. I would question some instances of practical methods being abstracted to theoretical assumptions--e.g. a word's meaning being derived from the surrounding text, or the theoretical need for a medium-independent word-form. Clearly this book would be of interest for those who work in corpus-based contrastive analysis or foreign-language dictionaries, but probably not for those accustomed to generative approaches.

Background: I graduated from MIT in 1996 and am now a doctoral student at Oxford University. My thesis concerns lexical semantics and syntax, specifically denominal and de-adjectival verbs.