Word classes
An important difference between formal and cognitive approaches relates to the characterisation of word classes. As we have seen, the cognitive approach favours a semantic characterisation. In Cognitive Grammar, for example, symbolic units vary in terms of specificity versus schematicity at the semantic pole. While content words are maximally specific, grammatical categories like NOUN are maximally schematic but both specific and schematic units belong within the same inventory. The major word classes receive a semantic characterisation: for example, the category NOUN is characterised at the semantic pole by the schema [THING] and the category VERB by the schema [PROCESS]. We also saw that closed classes like determiners and auxiliary verbs received a semantic account in terms of grounding predications.
In contrast, the formal approach argues against a semantic characterisation and defines word classes on the basis of morphological and distributional properties. As we saw in Chapter 14, this represents the traditional distributional approach to word classes. This type of approach is either explicitly adopted or taken for granted by most formal theories of language which reject a semantic characterisation of word classes on the basis that such an approach inevitably results in a description so vague as to be meaningless. In addition, the formal approach takes the position that a semantic characterisation cannot adequately distinguish word classes because members of two different categories can have the ‘same’ meaning. Consider the following examples:
According to the formal approach, the verbs in (5a) and (6a) are not semantic ally distinct from the nouns in (5b) and (6b), respectively. Love describes the same emotion in both (5a) and (5b), and destroy and destruction in (6a) and (6b) both describe the same kind of act. For this reason, a distributional approach is widely favoured because the structural characteristics of word classes are readily identifiable and, although not without exception, are also more or less predictable. Of course, most linguists would agree that there is some semantic basis to word classes. Speakers recognise that nouns typically describe things, verbs typically describe actions, adjectives typically describe properties and prepositions typically describe relations. According to the formal model, however, these rather ‘vague’ semantic characterisations are insufficient grounds upon which to base a model of language. Given that the aim of most modern theories of language is to describe a speaker’s psychological representation of language, the structural features of word classes are generally thought to lend themselves more readily to a model of this psychological representation of language, particularly in a modular system where morphology and syntax operate independently of meaning. According to the position adopted in cognitive linguistics, however, these distributional properties are epiphenomenal.
