Compared to phonological properties, it is more difficult to abstract a list of relevant distributional properties from the psycholinguistic literature, with some psycholinguists looking at n-grams, while others using (some subset of) more traditional morphological and/or syntactic properties, inspired by structuralist and generative theory. The position taken here is that, of all accounts of word classes available, Croft's (2001) theory of word classes is the most responsible to the facts of crosslinguistic variation (consider especially the predictions it makes with regard to what Croft labels "structural coding" and "behavioral potential", see fn. 4 and 5, above). For that reason, the distribution of nouns and verbs will be studied here in terms of occurrence in reference and predication constructions, respectively.

The question then emerges as to what are (the clearest examples of) reference and predication constructions in English. As I suggested above, this is where traditional generative criteria for nouns and verbs are useful.

One slot that is often used in the generativist literature to define nouns is that of the subject. However, objects also perform the role of keeping track of (or setting up a cognitive file for) a referent. Both these syntactic frames will therefore be analysed as reference constructions.

Co-occurrence with a determiner is also often cited as indicative of noun status, and since it establishes or maintains a referent this will be included as well (see also Croft 1991: 248). However, I take the position that co-occurrence with a determiner contributes less to referentiality than appearance in either the subject or object slot. Consider for example that determiner-noun combinations may function as predicate nominals (as in e.g. John is the culprit), which Croft (1991) suggests are crosslinguistically often more verb-like (see also Section 2.2.2, above). Similarly, oblique objects of prepositions may well contain determiners (see e.g. John hit Harry with a stick). The typological literature (e.g. Given 1976, Croft 1988) shows that agreement is used for very salient referents. Obliques frequently do not show agreement, which suggests that they are less salient and not among "the referents whose identity the hearer is most required to keep track of in discourse" (Croft 2001: 235). This is tantamount to suggesting that they are less referential than subjects and objects even if they contain a determiner.[1]

Other properties that are often taken as indicative of nounhood do not necessarily establish reference at all. This is because they are often merely morphological (e.g. the presence of a nominal or genitival suffix, see Aarts 2007: 102, 144), but anything less than a full noun phrase cannot ever be referential (see Searle 1969: 25). Thus, despite the fact that renumeration contains a nominalising suffix, if it occurs in the NP renumeration package it ends up representing a modifying rather than a referring propositional act. Other syntactic properties of nouns mentioned in the literature are not sufficient for acts of reference either. Consider for instance the presence of a PP complement (Aarts 2007: 144). A word string like paintings of his daughters does not yet set up a cognitive file, unless it is embedded in a larger NP construction such as the paintings of his daughters, which may in its turn occur in the subject or object position in a clause, e.g. The paintings of his daughters are very realistic or I admire the paintings of his daughters.

Generative criteria for verbhood do not all give rise to propositional acts of predication either. What we need for predication is a full grammatical predicate (Searle 1969: 25). Thus, an -ing ending, frequently analysed as indicative of verbal status (e.g. Aarts 2007: 138, 145) does not result in predication. In combination with a finite form of BE it does, but note that there it is really the presence of the auxiliary that is responsible for the effect. Langacker's earlier work on scanning (e.g. 1987a) could be interpreted as claiming that in auxiliary-main verb sequences, sequentiality is only invoked for the finite auxiliary, with non-finite forms being scanned summarily. However, Langacker (2008b: 576-577) makes it clear that it is the entire verb sequence in a finite clause that is subject to sequential scanning. His (2008b) revision of scanning modes also implies that predication becomes increasingly less salient to the extent that a predicate is syntactically more deranked

A related point could be made regarding partitive and generic subjects or objects, as in He spilt some of the beer and I hate spiders. Nandris (1945: 183-185, cited by Croft 2001: 232) provides evidence from Romanian, where such NPs are not marked by the locative preposition pe, whereas they are for human and definite objects. The questionnaire responses did not contain any examples of such subjects or objects. (see also Section 2.2.1, above). The continuum is hard to operationalise, not least since Langacker's (2008b) account is still, by his own admission, less than fully worked out (577). Here we will analyse the verb or verb sequence in finite clauses (including finite subordinate clauses)[2] as predicational in function. By contrast, non-finite novel forms in to-infinitive and -ing participial clauses feature much less sequentiality, and will be treated here as non-predicational.

The scoring scheme for nouns/reference is as follows. I analyse the examples of novel nouns that occur in subject or object position as fully referential. Forms that appear in other positions but co-occur with a determiner are interpreted as referential to some degree. Finally, forms in predicate nominal or oblique positions without a determiner are seen here as non-referential. This is of course an oversimplification, but it is an approximation of the conclusions we draw on the basis of cognitive-typological research.

For verbs/predication, the scoring scheme is simpler: Forms that occur as the main verb in a finite clause are rated 1, those that do not score 0.

In my analysis I will also address the question as to whether there is a difference between the properties of subject position, object position, or occurrence in a definite NP in terms of frequency in my participants' sentences, and therefore perhaps in the way these distributional facts are stored as part of speakers' knowledge of nouns. I will argue, though, that the nature of the task renders it difficult to draw any firm conclusions as to the relative importance of subject and object position in relation to the noun schema.

  • [1] Furthermore, the full NP such as your teeth in the verb phrase gnash your teeth is probably not highly referential. Evidence for this comes from languages where the NP is incorporated into the verb, such as German Zähneknirschen or Dutch tandenknarsen, lit. 'teeth-gnash. See also Croft (1991: 121), who cites Mithun's (1984: 859) claim that such nouns "simply ride along with their host V's". This point is more interesting from a theoretical than a practical perspective, as there were no such cases in my questionnaire data.
  • [2] See also the discussion in Langacker (2008a: 415-419), which suggests that finite subordinate clauses are often at least equal in prominence to the main clauses and sometimes even more so. This high degree of prominence implies that in mentally simulating the events described, speakers attend carefully to each configuration, i.e. they scan it in sequential fashion.
< Prev   CONTENTS   Next >