Grammatical category and discourse function
In a well-known article, Hopper and Thompson (1984) suggested an alternative to semantically or grammatically oriented approaches to the parts of speech like those discussed above. Hopper and Thompson claim that nounhood and verbhood should be defined in terms of discourse functional factors: factors which are not primarily about meaning or grammatical rules, but about the informational roles which words play in actual discourse, the particular function of different words in organizing the information conveyed by the utterance.
We’ll mainly concentrate on Hopper and Thompson’s proposals about nouns. They begin by observing that the grammatical properties associated with nouns are not constant across all the contexts in which nouns appear. They compare three uses of the noun fox:
In all three cases fox is, of course, a noun. But Hopper and Thompson note that the grammatical ‘nouniness’ of fox is manifested in rather different ways in the three cases. In English, nouns are often distinguished grammatically by the ability to take plural suffixes, and appear in NPs introduced by determiners. Ability to host number markers and compatibility with determiners are therefore among the cardinal grammatical possibilities that an English noun has. Most common nouns have these possibilities. But even though fox is an entirely typical English noun in terms of these abstract grammatical possibilities, whether or not the possibilities are available depends on the particular grammatical context in which the noun occurs. In (19) above, the full possibilities of number-hosting and determiner acceptance are, in fact, only both available in (19a). Case (19a) already contains the determiner an, and we could pluralize it to foxes (an alteration which would necessitate substituting the determiner some for an). In order to express the generic meaning of (19b), we either use the plural foxes, or we can use a sin gular noun with determiner: the fox is cunning. But we don’t have both options. We can’t say Fox is cunning (singular noun; no determiner) or the foxes are cunning (plural noun; determiner) and preserve the same meaning: the foxes are cunning doesn’t mean that foxes in general are cunning, just that these particular foxes, which we’re already talking about, are cunning. In (19c) fox can have neither plural marking nor a preceding determiner. In (19a) fox iden tifies a concrete, perceptible entity which is introduced as a participant into the discourse. In (19b) and (19c) fox(es) does not refer to any single, concrete set of foxes, but to the class of foxes in general.
What accounts for these differences? Hopper and Thompson suggest that nouns can be understood as prototype categories defined by their discourse functions. The difference in the grammatical options available to a given occurrence of a noun correlates with the discourse function that that noun is playing in any given context – the closer the noun is to playing its prototypical discourse role, the closer it comes to exhibiting the full range of grammatical possibilities of its class. Nouns’ prototypical function, Hopper and Thompson claim, is to introduce ‘participants’ and ‘props’ into the discourse, and to deploy them, as in (19a). We use nouns to bring participants into the discourse, and then to manipulate them through the course of the text. It is only in grammatical contexts which can fulfil this discourse role that nouns display their full range of grammatical options. That is the reason that the possibilities of pluralization and determiner selection are greatest in (19a). In (19b) and (19c), the grammatical context is not one which allows fox to be deployed as a participant/ prop in the discourse; as a result, fewer of the grammatical options associated with nounhood are available (see Table 9.2).
Hopper and Thomspon detail a large amount of cross-linguistic evidence designed to ‘show that the extent to which prototypical nounhood is achieved is a function of the degree to which the form in question serves to introduce a participant into the discourse’ (1984: 708). Not all occur rences of a given noun, in other words, are equally ‘nouny’. Even though fox is a typical English noun in having the full range of grammatical characteristics we associate with nouns – compatibility with number markers and determiners – these characteristics can only be manifested in some contexts. The context where the full range of these grammatical characteristics can be manifested is when the noun is fulfilling its prototypical discourse function: introducing or deploying a participant or prop, as in (19a). When a noun isn’t doing this, the range of grammatical possibilities shrinks (19b–c). ‘From the discourse viewpoint’, Hopper and Thompson explain,
nouns function to introduce participants and ‘props’ and to deploy them. To the extent that a linguistic form is carrying out this prototypical function, it will be coded as N, and will manifest the full possible range of nominal trappings conventional in the language. Forms which fail in some way to refer to concrete, deployable entities will typically lack some or all of these trappings (1984: 710–711).
One obvious type of noun which does not introduce a participant into the discourse are those nouns which are complements of verbs of being. This type of nominal, Hopper and Thompson show, often loses many of the grammatical characteristics associated with full nounhood. In Mokilese (Austronesian, Micronesia) and Hungarian (Finno-Ugric, Hungary) (optionally), and in French and Ancient Greek (obligatorily), nouns which are the complements of verbs of being do not take determiners, even though appearance with determiners is one of the grammatical hallmarks of nounhood in these languages:
In all these languages, however, when the noun phrases serve their proto typical function of introducing or deploying a participant or prop, they regain the possibility of taking a determiner. The loss of this grammatical possibility in the above examples
correlates with an absence of intention to refer to an extant entity: the thing named is not used as a participant or prop in the discourse. In other words, there are no discourse contexts in which such a predicate nominal serves to introduce a participant into the discourse for further deployability (Hopper and Thompson 1984: 716).
Another context in which participants are clearly not introduced or deployed is that of negation. Nouns under the scope of negation (e.g. nouns which are the objects of verbs in negative clauses) often cannot take the same range of determiners as nouns which are not under the scope of negation:
Since a noun under the scope of negation cannot have the function of deploying a participant or prop, its grammatical possibilities are much more restricted.
We will not go into the details of Hopper and Thompson’s discussion of verbhood, which sees assertion of the occurrence of an event as the proto typical discourse function of a verb. When verbs assert the occurrence of events, they characteristically can take the full range of tense–mood–aspect markers available in the language, as well as the other hallmarks of verb hood, like syntactic agreement with associated noun phrases. In contexts where verbs do not assert the occurrence of an event, many of these possibilities disappear. This accounts for the difference between (22a) and (22b):
The verb in (22a) asserts the occurrence of an event and, as a result, can occur with a wide range of tense, aspect and modality inflections (McTavish threw/is throwing/will throw/might throw, etc.), and may, depending on the language, show agreement with its associated NPs. In (22b), on the other hand, no event is asserted to have occurred and these possibilities are lack ing (to throw/*will throw/*may throw/*can throw/*should throw, etc.).
Hopper and Thompson’s conclusion is that ‘linguistic forms are in principle to be considered as LACKING CATEGORIALITY completely unless nounhood or verbhood is forced on them by their discourse functions . . . In other words, far from being given aprioristically for us to build sentences out of, the categories of N and V actually manifest themselves only when the discourse requires it’ (1984: 747; emphasis original). This conclusion is similar to the one we reached at the end of the previous section, where we argued that nounhood and verbhood need to be considered not as proper ties of given roots, but of grammatical ‘slots’ or ‘contexts’ in the sentence. Hopper and Thompson’s analysis suggests that this is an oversimplification: there is no such thing as a ‘noun’ slot or a ‘verb’ slot, each of which is defined by a fixed range of grammatical properties. Rather, the slots into which nouns and verbs can be inserted are structured around a central, most prototypical member, which most completely manifests the available grammatical properties for the category in question. Hopper and Thompson’s innovation was to explain the origin of this prototype structure: the most grammatically prototypical nouns or verbs are those which fulfil the most prototypical discourse function – introducing/naming a participant in the case of nouns, and asserting the occurrence of an event in the case of verbs. When nouns introduce props and participants, they can appear in the most elaborated versions of their own slots – versions where the full range of grammatical possibilities is available. When they do not, the grammatical slots show a correspondingly reduced set of grammatical options.
Hopper and Thompson’s analysis has certainly not gone unchallenged (see Francis 1998). In particular, critics might well wonder whether the notions of introducing a participant and asserting an event are any less slippery than the semantic notions of reference and predication they replace. But it is extremely suggestive, and in its refusal to ground categoriality in semantics alone, it opens up the interesting question of how many grammatical facts can be explained by discourse-functional principles rather than semantic ones.
