A B S T R A C T
Arnold Zwicky
Observation (1): The input is skewed; the stimulus is impoverished, as we say in the trade. At the very least, the elements contributing to a sign have background frequencies that skew the learner's data. But in addition, exemplars of a complex sign tend to cluster in a few subtypes, the seeds, which do not necessarily constitute a (whole) natural class. The stranded to phenomenon (Verb Phrase Ellipsis leaving infinitival to as remainder: Zwicky 1982) serves as an illustration.
Observation (2): The amount of variation in the grammars of different speakers is enormous; the details of the same complex sign are strikingly different for different speakers. Stranded to , Auxiliary Reduction (Kaisse 1985), and quasi-serial verbs (as in Go see who's at the door:Pullum 1990) illustrate the point.
Observation (3): Observation (2) follows, at least in part, from (1). Differences in grammars have their sources in differences in the input (even if the inputs are samplings from the same population, small differences can be crucial if they involve especially noticeable data), or in differences in the way learners generalize from the skewed sample they get, or of course both. Unless the way is absolutely barred, as with the asyntactic idiom by and large, everyone seems to extrapolate from the seeds of signs, but in many different directions and to greater or lesser degrees.
Observation (4): Differences in individual grammars are then windows on universal grammar, since every idiolect is a possible language. The individual grammars provide precious evidence about the relevant dimensions of language structure, including evidence about universal constraints; no one, for instance, seems to generalize from the seeds of stranded to on the basis of the phonological properties of these seeds, just as the Principle of Phonology-Free Syntax (Pullum & Zwicky 1988) would predict.
Observation (5), which follows from (2): The interpretation of a single example is inscrutable (even assuming that the example is not an error)—inscrutable not only to linguists, but also to other speakers.
Observation (6): Not surprisingly, given (2), many speakers have varieties with alternative versions of the same sign.
Observation (7): Not surprisingly, given (1)—though many linguists find this very surprising, indeed unacceptable—the conditions on many complex signs are disjunctive, a collection of disparate formal contexts. Riehemann (2001) observes that, though an idiom generally has a canonical form (which dominates the data statistically, is reflected in the way dictionaries and speakers refer to the idiom, and serves as the seed for learning), some have more than one. The same is true of constructions; a widespread variety of English, for example, permits Auxiliary Reduction with preceding subjects (of any category), fronted wh words (not necessarily subjects, but not multi-word phrases), and certain subordinators (like relativizer that), a collection of contexts that falls under no single formula.
Observation (8): The collection of seeds of a sign will tend to maintain itself through time; the seeds of stranded to have maintained themselves for at least a generation. However, the assortment of different grammars extrapolated from these seeds serves as a reservoir of variants that can then spread throughout the population of speakers in the usual ways—by chance, by spread on social grounds, or by spread on the basis of the communicative values of the variants; over a few centuries, the quasi-serial verb construction has spread from imperatives to bare infinitive forms in general and then, for almost all speakers, to some wider class of contexts.
References
Kaisse, Ellem M. 1985. Connected speech: The interaction of syntax and phonology. NY: Academic Press.
Newmeyer, Frederick J. (ed.). 1988. Linguistics: The Cambridge survey. Cambridge: Cambridge Univ. Press.
Pullum, Geoffrey K. 1990. Constraints on intransitive quasi-serial verb constructions in modern colloquial English. OSU WPL 39.218-39.
— & Arnold M. Zwicky. 1988. The syntax-phonology interface. Newmeyer 1988, I.255-80.
Riehemann, Susanne. 2001. A constructional approach to idioms and word formation. Stanford Univ. Ph.D. dissertation.
Zwicky, Arnold M. 1982. Stranded to and phonological phrasing in English. Linguistics 20.3-57.