This paper was presented at the 34th Annual Meeting of the Chicago Linguistic Society. A published version appears in the conference proceedings for CLS 34, pp. 159-170.
Langacker (1987, 1991) proposes a theory in which all conventional elements of a language (lexical items, grammatical morphemes, lexical classes, syntactic constructions) are 'symbolic units' -- i.e., pairings of form and meaning. For example, a category like NP is defined not by a formal syntactic rule but rather by the pairing of formal and conceptual content common to its members. Thus, with no autonomous level of syntax, Langacker must account in some other way for the fact that a single form can have different meanings. For certain lexical and phrasal categories, the claim is that all members share a schematic conceptual core: nouns and NPs are always 'things', while PPs, adjectives, adverbs and quantifiers are always 'atemporal relations'. This approach acknowledges that category members may differ greatly in non-core meaning (see Lakoff 1987), while in general accounting well for the core meanings of these categories.
There are, however, problematic examples in which it appears that a noun or NP designates an atemporal relation rather than a thing. In (1), for example, the noun 'lot' serves a relational function, and in (3) the NP 'a fool' is predicative and relational. Likewise, there are examples in which a relational category such as PP designates a thing, as in (2), where the PP subject 'under the bed' designates a particular location. Langacker accounts for these examples as follows. In (1), 'lot', formerly as noun, has been reanalyzed as a type of atemporal relation (1991: 88-9). Similarly in (3), 'a fool' is an atemporal relation zero-derived from a NP (1991: 66). And in (2), 'under the bed' is a nominal zero-derived from a PP (1991: 66). Thus, Langacker preserves the claim that only nouns and NPs designate things and only other categories designate relations.
However, this solution is problematic even within Langacker's own theory, because it renders virtually meaningless the thesis that syntactic constructions are conventionalized form-meaning units. Conventionally, the units P+det+N in 'under the bed' (example 2) constitute the category PP, not NP. There is no morphological evidence of any derivation, unlike in usual cases of nominalization; and as shown in (5), 'under the bed' is grammatically unlike NPs with respect to determiners and plural marking. The only apparent evidence for a nominalization analysis of 'under the bed' is its sentential function as grammatical subject. The zero-derivation "solution" in principle allows any string of symbols to be called a zero-derived nominal, thus bypassing the formal aspect of symbolic units altogether.
I propose an account of locative subjects, quantificational nouns, and predicate nominals in the framework of Autolexical Grammar (e.g., Sadock 1991, 1990). I avoid Langacker's problem by accepting a weaker version of the symbolic thesis: the meaning associated with a syntactic category can vary according to context, within the limits imposed by a general constraint. Such variability is easily represented in the Autolexical framework, which posits separate but interacting levels of syntax and semantics. I account for the properties of locative subjects in (2) and (5) as follows: there is a tension between the demands of syntax, which prefers a NP subject, and the demands of semantics, which allows any semantically appropriate external argument. Here, the semantics wins out: a PP subject (yes, it's really a PP!) is allowed because of its NP-like semantics. The non-occurrence of PP subjects with determiners and plural marking in (5), on the other hand, is explained by formal constraints which override semantic considerations. Similar results will be shown for quantificational nouns as in (1) and (4) and predicate nominals as in (3) and (6).
(1) (a) Sam's got a lot of friends.
compare: Sam's got many friends.
(b) A lot of Sam's friends were angry.
compare: Many of Sam's friends were angry.
(2) (a) Under the bed is dusty.
compare: The area under the bed is dusty. / The top of the bookshelf is dusty.
(b) Under the bed and on top of the fridge make good places to store things.
compare: The floor of the closet and the top of the fridge make good places to store things.
(3) She was a fool to leave.
compare: She was foolish to leave. / She was out of her mind to leave.
(4) (a) Sam's got a whole lot of friends.
compare: *Sam's got a whole many of friends. / Sam's got very many friends.
*Sam's got a very lot of friends.
(b) *Sam's got lot of friends.
(c) Sam's got lots of friends.
(5) (a) *The under the bed is dusty.
compare: The top of the bookshelf is dusty.
(b) *The unders the beds are dusty.
compare: The tops of the bookshelves are dusty.
(c) *Whenever the under a bed is dusty, Joe gets out the broom.
compare: Whenever the top of a bookshelf is dusty, Joe gets out the dust rag.
(6) (a) He's more of a fool than I thought.
compare: *He's more of foolish than I thought. / *He's more of out of his mind than I thought.
(b) *He's a fool person.
compare: He's a foolish person.
Lakoff, G. (1987) Women, Fire and Dangerous Things: What Categories Reveal about the Mind.
Langacker, R. (1987) Nouns and verbs. Language 63: 53-94.
------- (1991) Foundations of Cognitive Grammar, vol II. Stanford:
------- (1993) Reference-point constructions. Cognitive Linguistics 4: 1-38.
Sadock, J. M. (1990) Parts of speech in autolexical syntax. BLS 16: 269-281.
------- (1991) Autolexical Syntax.