Restricting the Theory of Transformations: A Case Study

  • Howard Lasnik
Part of the Studies in Natural Language and Linguistic Theory book series (SNLT, volume 20)


The purpose of this chapter will be to explore the expressive power of transformations. In particular, I will be primarily concerned with restrictions leading to a decrease in the class of possible transformations. The importance of such a step is that it potentially carries us towards a narrower class of possible grammars, and hence towards an answer to the fundamental question of how language can be acquired. The central arguments will be directed towards showing that various devices and rule-particular conditions often postulated in transformations are either not required to describe the kinds of facts that inspired them, or cannot provide the correct description of those facts.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    V“ in (1) consists essentially of terminal symbols, non-terminal symbols, and variables. The precise definition will not be of concern here.Google Scholar
  2. 2.
    In fact Peters and Ritchie do not make explicit use of and for such structural conditions, but their definitions implicitly have this effect. Their explicit use of and is for simple concatenation.Google Scholar
  3. 3.
    This assumes that do is inserted prior to Affix Hopping. When the reverse is assumed, it is the term immediately precedingthe affix that must not be a verb.Google Scholar
  4. 4.
    That is, a single non-terminal symbol. Possibly, this will have to be extended in certain limited cases to allow a single terminal as well.Google Scholar
  5. 5.
    Many of the details of the base are irrelevant to the concerns of this paper. In the exposition, the crucial aspects will become clear — any alternative set of PS rules preserving these latter features could be substituted. The feature system of Chomsky (1970) will be assumed, augmented in minor respects. Phrasal category names should be regarded as abbreviations for representations in X format.Google Scholar
  6. 7.
    Permutation is not available in that framework. If Subject Aux Inversion is a root transformation, as argued in a number of works, the formalism for cyclic transformations should not be enriched solely to accommodate that operation but rather, this power should be restricted to root T’s. Alternatively, one could possibly analyse SAI as substitution for (a portion of) COMP.Google Scholar
  7. 8.
    I follow the terminology of Bresnan (1971b) here. In the structure syntactic dependent of A. Note that I use the term `surface structure’ loosely here. The relevant level is presumably the output of the cliticization rules. Whether this level coincides with either the traditional level of surface structures, or the more abstract level of Chomsky and Lasnik (1977) is not certain.Google Scholar
  8. 9.
    The category labels here and elsewhere in the paper are to be taken as abbreviations for bundles of features. The most relevant features are as presented in (16). Irrelevant details of phrase structure are omitted.Google Scholar
  9. 10.
    Note that since (17) e (like all of the transformations) is optional, (23) would also underlie (a): This conclusion seems completely unobjectionable.Google Scholar
  10. 11.
    Kupin has suggested, as an alternative, that Affix Hopping be regarded not as a transformation but rather as a morphological re-analysis rule. As such, the process would automatically require adjacency — see Anderson (1979) for relevant discussion.Google Scholar
  11. 12.
    Reduction of object pronouns in English seems similar in most respects.Google Scholar
  12. 13.
    This treatment also extends to the copula; all that is required is that copula be be [+ aux].Google Scholar
  13. 14.
    At one time, I thought that faulty derivations of the type discussed here provided evidence for a precedence requirement in the structural description function, as given in Lasnik and Kupin ( 1977, Ex. 27). However, in the discussion during a presentation of this material at McGill University, it became clear that if (17) d leaves behind an empty category in accord with `trace theory’, then the independently motivated (25) would suffice, as John Goldsmith pointed out.Google Scholar
  14. 15.
    It might also be the case that this latter sort of incorrect derivation is blocked by the A/A condition, since the position into which movement takes place is crucially a verb within another verb. Attempting to replace the entire Aux verb position kke]Pres] would violate recoverability.Google Scholar
  15. 16.
    Alternatively, beginning again with (39), (17) a might be applied first, followed by (17) d. Now an attempted re-application of (17) d could raise be EN into the vacancy under Perfect creating no structure in violation of (25), but also creating no ungrammatical sentence. Whether there is any problem in this is not clear. A related derivation is considered in note 20 below.Google Scholar
  16. 17.
    Of course, under the alternative analysis of Affix Hopping mentioned in note 11, iteration would not be possible at all, since no movement would be involved.Google Scholar
  17. 18.
    See Chomsky (1957: 64) for a completely parallel argument.Google Scholar
  18. 19.
    An alternative account could be based on the claim that modals are morphologically defective, occurring only as finite forms, as suggested in Langendoen (1970). But note that selectional restrictions are independently required in this position, for example to preclude the selection of be with a [— aspect] affix, ie in the Aux.Google Scholar
  19. 20.
    ) would also prevent a derivation beginning with structure (39) in which do is inserted into the position vacated by have ultimately giving (a): though it is somewhat less clear that once have is raised the structure satisfies the SD of (45) in a relevant way.Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 1990

Authors and Affiliations

  • Howard Lasnik

There are no affiliations available

Personalised recommendations