The structure of Chomsky's argument, then, is that not only does Skinner add a mere obscure pseudo-scientific terminology, but that slot grammars themselves are not revealing. (I thank Drs Anthony Woodbury and Willem de Reuse for patient explanations of the relation between structruralism and behaviorism - none - and for telling me who structuralists were - pers. com. ) Examples of the slot/frame approach on a sentential level are the sentences (1) and (2) - one slot is for adjectives in this pair, eager and easy. The eager/easy case, Chomsky likes to point out in lectures, is an example of a frame/slot approach not being sufficient to account for language data.
(1) John is easy to please
(2) John is eager to please
This descriptive method, producing two adjectives for inclusion in a lexicon, was called 'search and discovery procedure'. Chomsky argued these procedures were inadequate by reasoning that (3) and (4) are not accounted for by 'search and discovery'.
(3) * It is eager to please John
(4) It is easy to please John
After all, (3) is ungrammatical. Instead of using frames/slots, Chomsky argued, we need to account for (3) and (4) and their asymmetry with (1) and (2). To that end, we must make the underlying grammatical relations visible.
One way of achieving this is the idea that constituents of sentences can be 'moved' during derivation. Underlying (1) and (2) would be the following 'deep structures':
(5) [empty subject position] is easy to please John
(6) John is eager [ John to please]
(5), but not (6), would then undergo a transformation that placed John into the subject position of the infinitival complement. Again, I am simplifying both the then theory and its justification tremendously here. (6) has a deep structure subject that is equal to the main clause subject, and the subject of the embedded clause (an infinitival complement in traditional grammar) is deleted.
Perhaps the clearest argument against behaviorism Chomsky put forward in the review is the fact - disputed by some as a fact - that grammars of natural languages cannot be described without recursion. To use my own phrasing, human linguistic creativity can be in part described by recursion, which Skinner's stimulus-response approach does not account for at all.
During my graduate school period, Chomsky's Transformational Generative Grammar, meant to address the short-comings of Behaviorism, fell apart. The theory exhibited many internal contradictions. This is a normal scientific life-cycle for scientific theories, not some devastating judgment on my part. Dilemma's appeared that, no matter what the choice, uncovered seemingly insoluble problems - (see Michael Brame; I'll see if I can find the reference).
Chomsky put forward a new version of his theory, denying that it was a new theory altogether (lecture at UT). It was christened the (Revised) Extended Standard Theory. The new theory was also called Government Binding, and the main foci seemed pretty much restricted to what in Transformational Generative Grammar was called 'Raising Verbs' (John seems [t to go home]), 'Equi-NP deletion' (John wants [PRO to go home]), and 'Forward Pronominalization' (John was thought [ t to be believed [ PRO to despise himself]]. Government-Binding, by way of an astute observation of Luigi Rizzi, also made good on its promise of discovering parameters by discussing a bounding difference between English and Italian. Apart from the transormations-of-old named above, wh- extraction was also well-studied.
However, other interesting phenomena, including various forms of verb deletion, and arguably even principles (cyclicity!), were conveniently relegated to the area of non 'core grammar', and the theory started exhibit a strange complexity. It would make an interesting Master's thesis to see which significant transformations simply disappeared in the transition from TGG to GB
At the time, a faculty member at the University of Texas, Prof C. L. Baker, was working on the quintessential linguistic problem of first language acquisition in Chomsky's Aspects framework. His introductory book is likely still in print for its remarkable clarity of writing and, hence, reflected thought. Even in the present day it is one of the few places where the central research goal of Chomsky's program, viz. to characterize Universal Grammar and thus account for rapid first language acquisition, is clearly stated rather than sketched in an oblique, almost cult-like manner.
In a fashion that matched his introductory book for clarity and precision, Baker worked on various forms of verbal deletion in Irish in comparison to verbal deletion in English. With this work, he intended to illustrate Universal Grammar (UG) principles at play in first language acquisition. The technical phrase for this line of research is solving the projection problem. One seeks to solve the question of how, out of a range of possible grammars, a language learner - known as child to non-linguists - selects one particular grammar out of a range of possible grammars. Sadly, with the advent of (the new version of) Chomsky's theory, the stock of Prof Baker's work went down on the academic market of ideas overnight. It seems to me an entire range of interesting grammatical facts, sluicing, gapping, VP deletion and verb-movement in Coninential Germanic were suddenly outside the range of study - this while sluicing obeys Ross' constraints on wh-movement.
So what's a guy to do to solve problems brought on by an arcane theory, arcane to him, at least? A guy is to go into applied linguistics. In my case, that meant computational linguistics. After a few minor publications in Linguistic Analysis I decided it was time for a change in directions. (One squib [LA 15:2, pp. 177-86] was guided by Dr Richard Kayne, and claimed the coreference of Dutch anaphora was directional. This eventually grew into my Master's Thesis. Working on this page, I unexpectedly found this paper cited. The other publication was about parasitic gaps. This one is cited, too!) A new adventure, computational linguistics had just begun, although it took me a few months to legitimize working performance issues, viz computational linguistics, rather than competence issues to myself. To me, those who are able to come up with sluicing data by themselves, and then fit the data neatly into Generative framework theory will always be the True Linguists (tm), whereas I merely struggle with electronic language data, a craft at best - insert Mozart-Salieri type reflections here. I am slowly coming around to the view, though, that Shannon's information theory is also linguistics. After all, science and mathematics are brothers of old, and there are a lot of impreciuse data in the Generative framework. Ah, that Calvinist pseudo-integrity.