Gisborne's chapter is concerned with three related questions: what is the proper analysis of a definite expression as a quantified expression or as an expression of discourse familiarity; how can the alternative approaches be modelled in a cognitive theory, especially as the debate has largely been conducted in different models of formal semantics; and which theory best accounts for the historical development of the English definite article? Gisborne argues that the best treatment of definite expressions is that they are in fact quantified, as originally proposed in Russell (1905), and he goes on to elaborate a set-based treatment of definite expressions modelled in one particular cognitive theory, Word Grammar (Hudson 1990, 2007).

The chapter argues that the diachrony of the English definite article is straightforwardly accounted for in this semantic approach because there is no need to assume that there are major semantic changes in the development from demonstrative to definite article. Instead, what is seen is a loss of the deictic element in the meaning of the demonstrative, with the result that the meaning of the definite article is somewhat more abstract.

This chapter shows that it is possible to model research findings from formal semantics within a cognitive theory, and it sets out to show that a full evaluation of different theoretical positions requires an exploration of both diachronic facts and synchronic distributions, which brings about certain similarities with Cristofaro's chapter. It is also argued that the cognitive embedding of the semantics of definite expressions makes a theory of change more straightforward than a model theoretic semantics would.


The chapter by Cristofaro is yet another chapter with a diachronic dimension. The general issue that concerns the author is how to weigh synchronic evidence against diachronic evidence in cognitive explanations of linguistic patterns. The data that the chapter is concerned with mostly involves prototype effects in dependent clauses and alignment phenomena, such as ergative-absolutive vs. nominative-accusative case marking, and the question of whether synchronic evidence and theories based on synchronic evidence lead to robust conclusions about underlying cognitive mechanisms.

Cristofaro observes that synchronic distributional patterns, such as can be found in alignment systems and the parts of speech are routinely taken as evidence for specific cognitive mechanisms. However, one of the issues that this approach does not address is that these cognitive mechanisms do not appear to play any role in the diachronic development of these constructions. For example, nominative, ergative, and active alignment systems often originate from processes which are independent of the cognitive mechanisms that are postulated to account for these systems on synchronic grounds. Likewise, some types of prototype effects, such as those found in dependent clauses, arise from processes that are arguably independent of the fact that the relevant grammatical categories have a prototype structure.

Cristofaro's conclusion is that synchronic distributional evidence cannot be taken as evidence for specific cognitive mechanisms, in the sense that it is not clear that these mechanisms actually lead speakers to create or use particular constructions.


The possibility of a rapprochement between (some versions of) generative and cognitive linguistics is also suggested by Hollmann, albeit in a very different context. His contribution addresses the representation of word classes. Traditionally, word classes have been one of the main areas of debate between generative and cognitive linguists. Generative grammarians (e.g. Baker 2003, Aarts 2007) tend to emphasise distributional properties, while cognitive linguists argue that categorisation is determined semantically (Langacker 1987) or semantically-pragmatically (Croft 2001). However, Hollmann suggests that the distinction between the two traditions is perhaps not as sharp as it might appear, as Croft's suggestions concerning the discourse propositional act functions of the different word classes may be translated into language-particular distributional properties.

Focusing on English word classes, Hollmann reports on a production experiment in which subjects were asked to devise novel nouns and verbs, and use these in a sentence. The phonological and distributional properties of these nonce forms were then analysed. With regard to phonology, the results revealed a high degree of overlap with work on lexical categorisation in psycholinguistics (e.g. Cassidy & Kelly 1991, Kelly 1992, 1996, Monaghan et al. 2005).

In relation to the theme of this volume, the psycholinguistic evidence suggests that both generative and cognitive linguists have failed to acknowledge certain highly relevant facts. Incorporating phonology into the models, for example, would lead to more convergence. The analysis of the syntax of the novel forms led to an interesting conclusion as well: at least some distributional facts emphasised in the generative model in particular, co-occurrence with a determiner may be more important than one might expect on the basis of cognitive linguistic work.

One important caveat Hollmann notes with regard to the potential convergence between generative and cognitive linguists is that one extreme incarnation of generative grammar, i.e. Distributed Morphology (Halle & Marantz 1993), would fare very poorly indeed in the face of the evidence presented in this chapter. Theorising in DM thus constitutes a paradigm case of the kinds of formal linguistic overabstraction lamented by the psycholinguist Ferreira (2005).

< Prev   CONTENTS   Next >