Abstract
Much of the recent discussion of language learnability has centered around the absence for the learner of negative evidence and the implications of that absence. The basic argument has been reiterated many times: If the child does not have access to negative evidence — the information that certain structures are not part of the language — then Universal Grammar presumably does not make available choices that can only by resolved by such evidence. (See Chomsky and Lasnik (1977) for early discussion.) In principle, the concern is exclusively with the situation schematized in (1).
I am indebted to Stephen Crain for extensive discussion of the issues in the paper.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Notes
Much of this material was presented at the society for philosophy and psychology workshop on language learnability, held in London Ontario, July, 1982. The first draft of the written version was circulated in 1983.
See Davis (1983) for discussion of Dell’s proposal, and some possibly undesirable consequences.
This is not strictly speaking correct, as in null-subject languages, overt pronouns are more heavily restricted in their distribution than is the case in obligatory overt subject languages. For example, overt pronouns cannot function as bound variables. English sentence (i) contrasts with Spanish sentence (ii) in this regard. In Spanish, (i) would be expressed with a null subject for the embedded clause.(i) Everyone, thinks he, is intelligent.(ii) *Todo el mundo, piensa que el, es intellegente.It is also relevant in this connection that null-subject languages lack overt expletive pronouns. Thus, diagram (2b) more accurately reflects this situation than does (1). For the purposes of the present discussion, however, these properties will be ignored.
It might be thought than Bresnan’s proposal is immediately falsified by a phonological property of the contracted copula: it assimilates in voicing to the preceding segment, rather than to the following one, as shown in the following examples.(i) Bill’s crazy [z] *[s](ii) Pete’s nice[s] *[z] In this respect, it patterns in just the same way as the regular plural suffix and the possessive marker. If such assimilation requires syntactic dependence, then, contrary to the proposal, ’s must be an enclitic. However, it is not clear that the voicing assimilation at work here requires syntactic dependence rather than, say, mere adjacency. Edwin Williams (personal communication) provides a bit of evidence for this point of view. He observes that the possessive (with its usual assimilatory properties) can appear not merely on a noun, but on a full NP, as in (iii).(iii) The man next door’s car Williams further observes that when the possessive is syntactically associated with a pronoun, suppletion results: (iv) he’s his I’s my etc. Fianlly, he concludes that the association in e.g. (iii) between door and ’s is not syntactic. If it were, then in (v), the association between him and ’s would also have to be syntactic.(v) [A picture of him]’s frame (= the frame of a picture of him). But if the association were syntactic, we would expect not (v), but rather (vi).(vi) [A picture of his] frame.Thus we have voicing assimilation with the preceding word even where ’s is clearly not a clitic on that word. This substantially weakens the potential counter-argument to the Bresnan analysis.
See, for example, Chomsky and Lasnik (1977), Postal and Pullum (1982), Bouchard (1982).
There is one additional assumption in this discussion, namely, that the learner is forced to pick one or the other analysis of contraction. A particular contraction process could not freely allow encliticization or procliticization in a particular grammar. In the absence of this assumption, evidence for e.g., encliticization, as in (4b) would not be evidence against procliticization. Then, positive evidence would not in general suffice for the choice among the three possibilities: procliticization, encliticization, free cliticization.
Note that it will be of little help to posit that the English dative alternation does not belong to core grammar but rather to the periphery. This is so since the acquisition problem is no less severe for “peripheral" processes, and direct negative evidence is presumably no more available.
This latter problem can he resolved if we assume, with Mazurkewich and White, that the phonological and semantic properties in the constraints are not salient to the child in the early stage. As soon as they become salient, (14) becomes the unmarked case. Of course, it remains an open question why these properties are not salient all along.
See Lasnik (1981) for further examples.
See, for example, Wexler and Culicover (1980).
An apparent typographical error has been corrected here. As presented in Grimshaw (1981), the lexical category immediately dominating books is N. This was clearly not Grimshaw’s intention, since she claims that the phrase marker is “entirely consistent with X theory,” yet the VP is not heated by a V.
One might speculate about expected errors in acquisition under such a learning theory. A non-canonical word, say a noun such as destruction which names an action, should be mistakenly treated as a verb under the boot-strapping theory, it would appear. Neither Grimshaw nor Pinker gives evidence that this does happen.
Pesetsky (1982), based on work of Grimshaw (1979, 1981) goes so far as to argue that once the meaning (the 0 property) of a word is learned, its syntax, in particular, its subcategorization frame, is known automatically. Pesetsky indicates that acquisition must proceed in this fashion, on grounds of epistemological priority. The primitives of 0-assignment — agent, patient, etc. — are plausible primitives, while those of subcategorization — NP, S — are not. That is, the former can be used by the child in initial analysis of the data, while the latter cannot. Certainly, as Pesetsky argues, subcategorization is largely predicatable from semantics (’S — selection’) and the redundancy in the theory should somehow be reduced along the line he suggests. But the acquisition scenario deserves further consideration. First, while the notion `agent of an action’ might well be available in advance of any syntactic knowledge, is `agent of a sentence’ similarly available? That is, in advance of any syntactic knowledge, can the child determine what portion of a sentences constitute the agent? Second, there appears to be a tacit assumption that the meaning of, e.g., a verb, can be presented and apprehended in isolation. But this seems implausible. Rather, verbs are presented in grammatical sentences which, therefore, explicitly display subcategorization properties. In fact, one might consider reversing the whole story: subcategorization is explicitly presented, and the child uses that information to deduce central aspects of the meaning of verbs.
Another markedness substitute for indirect negative evidence — Wexler’s “Uniqueness Principle” — is discussed in Lasnik (1981) and Wexler (1981). By this principle, in the unmarked case there is at most one grammatical surface structure for each base structure. Where alternative derivations from the same deep structure would yield more than one sentence, the presence in the data of one of them is taken as “indirect” evidence that the absent alternatives are ungrammatical.
Rights and permissions
Copyright information
© 1990 Springer Science+Business Media Dordrecht
About this chapter
Cite this chapter
Lasnik, H. (1990). On Certain Substitutes for Negative Data. In: Essays on Restrictiveness and Learnability. Studies in Natural Language and Linguistic Theory, vol 20. Springer, Dordrecht. https://doi.org/10.1007/978-94-015-6859-3_9
Download citation
DOI: https://doi.org/10.1007/978-94-015-6859-3_9
Publisher Name: Springer, Dordrecht
Print ISBN: 978-0-7923-0629-0
Online ISBN: 978-94-015-6859-3
eBook Packages: Springer Book Archive