In recent years there have been two still different approaches taken to the problems of CA，both resulting from the current enthusiasm for generative transformational theory.One of these approaches dismisses the hypothesis from any consideration at all.This dismissal stems from a strong negative reaction to CA，as，for example，in recent articles by Ritchie（1967）and Wolfe（1967）in Language Learning.The second approach attempts to use the generative-transformational model in order to provide some of the necessary overriding theory to meet either the demands of prediction in the strong version or of explanation in the weak version.

All natural languages have a great deal in common so that anyone who has learned one language already knows a great deal about any other language he must learn.Not only does he know a great deal about that other language even before he begins to learn it，but the deep structures of both languages are very much alike，so that the actual differences between the two languages are really quite superficial.However，to learn the second language，one must learn the precise way in which that second language relates the deep structures to its surface structures and their phonetic representations.Since this way is unique for each language，CA can be of little or no help at all in the learning task because the rules to be internalized are of course，unique.Even though the form and some of the content of the rules to be acquired might be identical for both languages，the combinations of these for individual languages are quite idiosyncratic so that superficial contrastive statements can in no way help the learner in his task.

Now there is obviously some merit in the above argument.If the underlying vowel system of French is something like the one Schane outlines in French Phonology and Morphology（1968），and the underlying vowel system of English is something like the one Chomsky and Halle outline in The Sound Pattern of English（1968），and if the speaker of English must somehow internalize the underlying vowel system of French and the fifty or so phonetic realization rules which Schane gives in order to speak acceptable French，then one may easily be tempted to reject the whole notion of CA，claiming that it has nothing at all to contribute to an understanding of the learning task that is involved.

Uncertainty is obviously piled upon uncertainty in making CA.Such uncertainties arise from inadequacies in existing linguistic theories.As an example of theoretical inadequacy，one may observe that the notion of deep structure itself is extremely uncertain.However，for the purposes of CA any claim that all languages are very much the same at the level of deep structure seems to be little more than a claim that it is possible to talk about the same things in all languages，which is surely not a very interesting claim，except perhaps in that it seems to contradict the one made by Sapir and Whorf.The preceding state-ment is not meant to be a criticism of generative transformational theory；it is meant to show how acceptance of that theory can fairly easily lead one to reject the idea that it is possible to make contrastive analyses，or，put less strongly，to reject the idea that generative transformational theory has something to contribute to a theory of CA，given the present state of the art.

Many experienced teachers find themselves unable to accept such reasons for rejection of the hypothesis.Their experience tells them that a Frenchman is likely to pronounce English think as sink and a Russian likely to pronounce it as tink，that a Spaniard will almost certainly fail to differentiate English bit from beat，and that an Englishman learning French will tend to pronounce the French word plume as pleem or ploom.They admit that in each case they must be prepared to teach the whole of the second language to a learner，but also insist that some parts of that second language are easier to learn than others，for no one ever must learn everything about the second language.However，many also admit that they do not know in what order learners should try to overcome the various difficulties they are observed to have.Should a Spaniard learning English learn to differentiate bit from beat and bet from bat because of the important surface contrasts that he does not make in Spanish？Or should he learn to associate the vowels in such pairs of words as weep and wept，pal e and pallid，type and typical，tone and tonic，deduce and deduction so that he can somehow internalize the underlying phonological system of English？The mind boggles at this last possibility！But it is one that descriptions of Spanish and English based on generative transformational theory would seem to hold out for teachers.

Some recent suggestions for using generative transformational theory in CA have actually been attempts to bring powerful theoretical insights to bear within the weaker version of the hypothesis in order to explain observed interference phenomena，for example some interesting work by Ritchie（1968）and by Carter（unpublished）.In their work，Ritchie and Carter have used distinctive feature hierarchies in attempts to explain such problems as why a Russian is likely to say link and a Frenchman sink for English think.Such work，using the notions of feature hierarchy，rule cycling，and morpheme and word structure rules，has considerable possibilities.Certainly this kind of work seems more promising than some being done by others in an attempt to show gross similarities between deep structures in an assortment of languages.