(de)selecting Arguments for Transitive and Predicated Nominals

Transitive nominals, that is nouns or adjectives that syntactically govern object arguments, are a problematic phenomenon under many influential categorizations of word classes. I show that in some languages, syntactic context can restrict the transitivity of transitive nominal categories; specifically, there is a clear association between predication and nominal transitivity, such that nominals can only be transitive when predicated. I develop a formal model of this restriction, based on the assumption that it should be specified syntactically, rather than lexically; I also consider the selectional properties of copular clauses more generally. 1 Transitive Nominals Transitive nominals, by which I mean nouns or adjectives that subcategorize for object arguments, are a problematic concept for several influential categorizations of word classes (as noted, in relation to adjectives, by Vincent and Börjars 2010: 459). For example, Bresnan and Kanerva (1989: 25), defining the feature [+objective], explicitly deny that objects appear with nouns or adjectives. Similarly, for Bresnan and Moshi (1990: 166–167) this is part of the very definition of an object: “objects are hypothesized to have the primitive property of complementing transitive predicators such as verbs and adpositions, and not complementing intransitive predicators such as basic nouns and adjectives.” Bresnan (2001: 100, 120) assumes the following feature distinctions for the major lexical categories (cf. also Bresnan 1976: 19): (1) +predicative −predicative +transitive V P −transitive A N In this model, +transitive categories “may take an object or direct complement function.” Adjectives and nouns, then, are denied the ability to take objects. These features are essentially the same as those of Chomsky (1970); the categorization of Jackendoff (1977: 31–33) likewise defines adjectives and nouns as inherently non-transitive (having the feature [−OBJ]). Despite the seemingly widespread formal definition of adjectives and nouns as inherently non-transitive, transitive adjectives have been recognized to exist in some languages. Platzack (1982a) noted and formalized the existence of transitive adjectives in Mordern Swedish. Other works that discuss the possibility of transitive nominals, but that seem uneasy with the idea, include Maling (1983: English), I am very grateful to Mary Dalrymple and Louise Mycock for comments on earlier versions, and to the audience at LFG13 for their attention and insights, especially Joan Maling, Dag Haug and Oleg Belyaev. I am also grateful to the editors and the external reviewer for their comments and suggestions. All errors are my own. This work was supported by an Early Career Research Fellowship funded by the Leverhulme Trust. van Riemsdijk (1983: German), and van Kemenade (1987: Old English). Vincent and Börjars (2010) find evidence for transitive adjectives in Swedish, Danish, Norwegian and Dutch, but treat them as marked relics in Modern Germanic, rather than a synchronically systematic phenomenon. Beside this relatively widespread evidence for transitive adjectives, transitive nouns appear to be cross-linguistically considerably rarer. Some do exist, however, and will be discussed below. Note that we are not here considering non-finite verb forms like participles and infinitives that may be morphologically nominal and yet transitive. The transitivity of such categories is fundamentally dependent on their verbal status, and is in no way typologically remarkable. From an LFG perspective non-finite verb forms can be treated as inflectional forms of verbs, sharing the same PRED value and subcategorization frame as finite verbal forms inflected to the same stem, in contrast to verbally derived nominals, whose PRED values are not paradigmatically identified with a verbal PRED (Lowe 2012). Note also that we are not here considering derived event nominals that display nominal phrasal syntax but that are often analysed in LFG as selecting for subject and object arguments at f-structure, although these appear in c-structure as possessive or prepositional modifier phrases (e.g. Laczkó 2000). The assumption that derived event nominals like destruction, observation etc. select for syntactic subject and object arguments involves a conflation of semantic and syntactic argumenthood. The fact that in the noun phrase Napoleon’s invasion of Russia, Napoleon is the semantic agent of the invasion and Russia is the semantic patient does not mean that at f-structure Napoleon need necessarily be a SUBJ and Russia an OBJ. In a parallel architecture, no such direct isomorphism between syntax and semantics is required. Rather, it is simpler to assume that at f-structure ’s possessives are consistently POSSs, and optional postmodificatory PPs are consistently ADJs, but in some contexts mapped to semantic arguments rather than semantic adjuncts. For the purposes of this paper, I restrict the notion of nominal transitivity to refer only to syntactic subcategorization for an object (OBJ) argument. Subcategorization for other grammatical functions by nouns and adjectives is typologically less rare and will not be treated in detail here, but the formalism advanced is capable of accounting for nominals that subcategorize for any combination of non-subject arguments. As discussed by Asudeh (2005) there may be distinct differences between syntactic and semantic argument structures, e.g. arguments may be present in the semantics that are entirely absent from the syntax. Despite the apparent rarity and problematic status of transitive nominals as discussed in this section, there is of course no formal problem in licensing subcategorization frames, including subcategorization for objects, for nouns and adjectives in LFG; cf. e.g. Butt et al. (1999: 105). Recent works discussing adjectives that select for non-subject arguments include e.g. Mittendorf and Sadler (2008), Al Sharifi and Sadler (2009) and Raza and Ahmed (2011). It is the syntactic alternation discussed in the following section, and not nominal “transitivity” per se, that I seek to account for in this paper.