Volume 29, Issue 3 - The Natural Semantic Metalanguage approach to linguistic meaning

After thirty years of language-internal, as well as cross-linguistic research, the Natural Semantic Metalanguage (NSM) model developed by Anna Wierzbicka and her colleagues has turned out to be a most useful theoretical and methodological framework for semantic analysis in various linguistic, and even non-linguistic, domains. This paper argues that the NSM approach to semantics constitutes a new paradigm in linguistic research which is free from various shortcomings of other semantic frameworks. The first section of this article provides a brief survey of the historical development of NSM theory from the early seventies up to the present stage. Its theoretical and methodological principles are outlined in sections two and three, before its applications in various domains are illustrated in section four by means of examples from a number of languages.

In 1972, the publication of Wierzbicka's Semantic Primitives initiated the Natural Semantic Metalanguage approach to linguistic semantics. NSM sets out to analyze the meaning of natural language utterances using only paraphrases built from a set of concepts called semantic primes. The main hypothesis is that a definite small set of unanalyzable primes that occur in every natural language suffices to characterize all natural meaning.

I would like to begin by briefly reviewing the assumptions and hypotheses the NSM approach is based on. The goal of this exercise is to understand how these are interrelated, what hinges on every one of them, and what consequences arise from discarding an assumption or disproving a hypothesis. This process defines a number of projects which all agree up to some point and then depart. Prima facie, all of these constitute legitimate avenues of inquiry. My aim is to evaluate the decisions made within the NSM program vis-à-vis these alternatives, and to do so from a particular perspective – that of a field worker dedicated to the study of semantics in Non-Indo-European languages and of a semantic typologist interested in variation and universals of semantic representations across languages. It should be clear from the outset that NSM has made contributions to the crosslinguistic perspective in semantics unsurpassed by those of any other framework (in particular, Goddard & Wierzbicka (eds.) 1994, 2002; Goddard (ed.) 1997) – so my interest should not come as a surprise.

Over the past few decades, the interpretation of language has gradually come to be accepted as a subject deserving its own academic regalia, such as conferences, journals, and university chairs. The definition of the emerging field of inquiry is, as yet, very moot. It is not even decided that it is a single field, rather than several related ones, and whether it (or they) should be subsumed under linguistics, psychology, or somewhere else. But there is a measure of consensus about the kind of topics the fledgling discipline is supposed to address. Here are a few examples:

Overall Durst has given a valuable and accurate synopsis of the NSM approach to linguistic meaning. I will attempt here to augment and clarify certain points.

Uwe Durst (D) is to be commended for a clear exposition of the Natural Semantic Metalanguage (NSM) theory of Anna Wierzbicka (W) and her associates. In addition to laying out clearly the assumptions of the NSM theory, D states that The …NSM model… has turned out to be a most useful theoretical and methodological framework for semantic analysis in various linguistic… domains.

0. Introduction Durst's paper is a well-written and clear survey of the Natural Semantic Metalanguage (NSM) model. However, it strikes us as somewhat over-enthusiastic and slightly simplistic: alternative positions and voices are not present, there is often no deep argumentation for or against a certain position, problems and complications are very seldom mentioned.

1. On the status of the primitives It is interesting and surely non-coincidental that the semantic primitives proposed by NSM researchers include some of the most hotly-debated topics in the formal semantics literature. There is a large body of formal semantic research (too large to be cited here) on each of the following NSM primitives: indexical pronouns such as I and you , demonstratives like this , quantifiers such as something, all, many , and one , modals like can , propositional attitude verbs like know and think , adjectives such as good and bad , the predicates have and (there) is , the connectives because, when , and if . Other proposed primitives such as before, after, the same, like, and kind (of) have also been the subject of discussion and debate. Indeed, there may not be a single proposed semantic primitive which fails to strike formal semanticists as extremely complex. Thus, it is difficult for us to accept the NSM claim that primitives such as i , you , someone , this , think , and want are ‘simple words’ and that they are ‘intuitively comprehensible and self-explanatory’ (Durst, p. 2).

The Natural Semantic Metalanguage (NSM) approach articulated by Uwe Durst is a componential theory of meaning, and it inherits many of the strengths of such theories. This is especially evident when we compare NSM with componential models that share its view of linguistic cognition as a reflex of the human meaning-making capacity in general. One such strength is the model's ability to account for prototype effects in categorization judgments without assuming scalar category membership or fuzzy category boundaries. Durst argues (section 3.3) that “[s]ince meaning is more than reference, one cannot conclude from referential fuzziness or vagueness that the meanings of words are fuzzy or vague as well”. The view is reminiscent of Lakoff's (1987) radial model of category structure, in which prototypicality ratings reflect not category structure but divergence of cognitive submodels that jointly define the best exemplars. Another strength of NSM that can likewise be traced to its decompositional base is its ability to capture cross-linguistic differences in lexical conflation patterns, as exemplified by Durst's comparison of words denoting anger in a variety of languages (section 3.3). Similarities and differences among the cognate words are captured by partial overlaps in their propositional representations, and what emerges is a relatively constrained picture of the range of typological variation. This is a strength that NSM shares with Talmy's (1985) model of motion-verb lexicalization patterns: these models allow otherwise ineffable translation problems to be described in rigorous ways. Just as Talmy's model enables us to talk about rhetorical-style differences among languages (or language families) by reference to fundamental features of event schematization (Slobin 1996), so the NSM approach captures ‘connotational’ differences among cognate lexical items that have been neglected in denotation-based lexicography.

1. Introduction Uwe Durst opens his helpful précis with the claims that NSM ‘has turned out to be a most useful theoretical and methodological framework for semantic analysis’, and that it ‘is free from various shortcomings of other semantic frameworks.’ Since the insights of NSM scholars into the details of many semantic domains have frequently been acknowledged, the theory's usefulness for semantic description need scarcely be doubted. I would like to suggest, however, that Durst's second claim is more questionable. Not only is NSM subject to shortcomings of its own distinctive stripe, but it inherits others from the approach to semantic analysis characteristic of linguistics generally, of which, in the last analysis, it is only a particularly forthright exemplar. NSM's most significant problems, then, are just the problems of semantics in general, seen as part of the would-be scientific project of linguistics. These problems are not always obvious from Durst's survey, and in this commentary I will indicate some of the most interesting. Given the space available, I will not discuss the universality of NSM's proposed primitives, perhaps the issue which has most often stimulated discussion. I will also only address the lexical semantics aspects of NSM. These are, in any case, at the theory's core, and since analogous issues arise in its treatment of other topics, the omission from this commentary of any discussion of the NSM approach to morphology, cultural scripts, pragmatics, and the like is, I trust, appropriate.

1. Is NSM ‘scientific’? The absence of technical terms and formal devices in NSM explications seems to invite some critics to dispute the scientific nature of the theory. In one example, Riemer 2 states that “[the] commitment to naturalness seems contradictory because in other areas of investigation, the development of a ‘scientific’ (i.e. empirical and testable) theory necessitates the very type of technical, artificial vocabulary which NSM explicitly repudiates”. To support this criticism, Riemer refers to evolutionary theory and generative phonology. But evolutionary theory is about the development of natural species, and generative phonology investigates (the rules of) phonological representations on the basis of phonological features. Both disciplines can rely on physical data, and hypotheses can be tested on this physical basis. Even abstract phonological features such as [±vocalic] or [±grave] can be described in physical terms.