Categorisation and polysemy in grammar: the network conception
Having outlined the conceptual bases of grammar according to Talmy and Langacker, in this final section we revisit the related issues of categorisation and polysemy from a grammatical perspective. Recall from Chapter 8 the idea that conceptual categories display prototype structure, and from Chapter 10 Lakoff’s proposal that words, like concepts, are represented in the mind as radial categories that typically exhibit polysemy. As we saw, this approach was held to account not only for open-class words like nouns but also for closed class words like prepositions.
Langacker develops a network model that represents the structure of categories. In this model, members of a category are viewed as nodes in a complex network. This can be seen as analogous to Lakoff’s radial category model. In Langacker’s model, the links between nodes in a network arise from a number of different kinds of categorising relationships that hold between the symbolic units stored in the grammatical inventory. One categorising relationship is extension from a prototype, represented as [A] [B], where A is the prototype and B shares some but not all attributes of A and is thus categorised as an instance of that category. A second type of categorising relationship is the relationship between schema and instance, represented as [A] → [B]. Recall from Chapter 14 that entrenched units that share a structural pattern give rise to a schematic representation of that structure, reflecting the usage-based nature of the model. The schema structures those related units as a category within the network, and novel expressions can be compared against such categories. According to Langacker, the network can grow ‘upwards’ via schematisation, ‘outwards’ via extension and ‘downwards’ as more detailed instances are added. Figure 15.18 captures the basis of the network model.
According to Langacker, the network model characterises not only polysemous open-class elements, but also underlies other kinds of linguistic categories, including those relating to sound as well as meaning and grammar. This means that morphemes, word classes and grammatical constructions are also envisaged as nodes in a network. It follows that while some nodes (like morphemes) are structurally simplex, other nodes (like the phonological poles of symbolic units or phrase- or sentence-level grammatical constructions) themselves have complex internal structure. For example, the English past tense morpheme is represented by the (partial) model in Figure 15.19.
At the semantic pole, PROCESS represents the verb and PAST represents the past tense morpheme. The next level in the network represents the various phonological instantiations of this schema, which are themselves schematic representations of the next level in the hierarchy, where specific instances of each schema are represented.
Polysemy and prototype structure in grammatical categories
If grammatical categories like closed-class words or bound grammatical morphemes are represented as discrete nodes in a complex network, and if such categories are by nature organised with respect to a prototype, we might expect closed-class categories to display polysemy in the same way that open-class categories do. Langacker argues that this is the case. For example, as we saw in Chapter 14, the modal auxiliaries represent a closed-class and show a fixed and predictable participation in the grammatical behaviour of the verb string. In terms of their semantic contribution to the clause, however, the modal auxiliaries present a striking case of polysemy. Compare the examples in (34), which illustrate the polysemous nature of the modal verb can.
In example (34a), the modal has a capacity reading: it conveys the speaker’s judgement concerning Lily’s capabilities. This is a kind of epistemic modality, which means it relates to the speaker’s knowledge. The interpretations in the other two examples are quite different: these reflect deontic modality, which relates to obligation and permission. While (34a) encodes permission, (34b) encodes obligation.
Langacker’s (1991: 185) discussion of gender morphology in Spanish illustrates prototype structure in grammatical categories. Spanish has two nominal gender suffixes,-o and -a. According to Langacker, the prototypical values of these morphemes are MALE (-o) and FEMALE (-a), as the examples in (35) illustrate.
However, these affixes also attach to nouns that have no inherent sex, such as the inanimate objects mesa ‘table’ and cerro ‘hill’. Langacker argues that these affixes are still meaningful, but have a highly schematic meaning in the latter type of example where they mean THING. Recall that this corresponds to Langacker’s representation of the noun category. This schematic semantic rep resentation is therefore consistent with the behaviour of these gender morphemes as noun-forming affixes. At the ‘centre’ of this category are the semantically rich instances which encode MALE/FEMALE and ANIMACY as well as having the status of entity or THING. At the periphery of the category are the schematic instances which share the semantic characterisation THING but lack the richer semantic characterisation.
