In the 1950s, Noam Chomsky introduced into linguistics the notion of a generative grammar, which has proved to be very influential. Now there are very many different types of generative grammar which can be conceived of, and Chomsky himself defined and discussed several quite different types in his early work. But, from the beginning, he himself favoured a particular type, to which he gave the name transformational grammar, or TG; TG has sometimes also been called transformational generative grammar.
Most types of generative grammar in which anybody has ever been interested can be usefully viewed as working like this: starting with nothing, the rules of the grammar build up the structure of a sentence piece by piece, adding something at each step, until the sentence structure is complete. Crucially, once something has been added to a sentence structure, it must remain: it cannot be changed, deleted or moved to a different location.
TG is hugely different. In TG, the structure of a sentence is first built up in the manner just described, using only context-free rules which are a simple type of rule widely used in other types of generative grammar. The structure which results is called the deep structure of the sentence. But, after this, some further rules apply. These rules are called transformations, and they are different in nature. Transformations have the power to change the structure which is already present in a number of ways: not only can they add new material to the structure (though only in the early versions), but they can also change material which is already present in various ways, they can move material to a different location, and they can even delete material from the structure altogether. When all the relevant transformations have finished applying, the resulting structure is the surface structure of the sentence. Because of the vast power of transformations, the surface structure may look extremely different from the deep structure.
TG is thus a theory of grammar which holds that a sentence typically has more than one level of structure. Apart from the structure which it obviously has on the surface, it also has an abstract underlying structure (the deep structure) which may be substantially different. The point of all this, in Chomsky’s view, is that certain important generalizations about the structures of the sentences in a language may be stated far more easily in terms of abstract deep structures than otherwise; in addition, the meaning of a sentence can often be determined much more straightforwardly from its deep structure.
TG has developed through a number of versions, each succeeding the other. In his 1957 book Syntactic Structures, Chomsky provided only a partial sketch of a very simple type of transformational grammar. This proved to be inadequate, and, in his 1965 book Aspects of the Theory of Syntax, Chomsky proposed a very different, and much more complete, version. This version is variously known as the Aspects model or as the Standard Theory. All text-books of TG published before 1980 present what is essentially the Standard Theory, sometimes with a few additions from later work.
Around 1968 the Standard Theory came under attack from a group of younger linguists who hoped to equate deep structure, previously a purely syntactic level of representation, with the semantic structure of a sentence (its meaning). This programme, called Generative Semantics, led to the positing of ever more abstract underlying structures for sentences; it proved unworkable, and it finally collapsed. Around the same time, two mathematical linguists demonstrated that standard TG was so enormously powerful that it could, in principle, describe anything which could be described at all – a potentially catastrophic result, since the whole point of a theory of grammar is to tell us what is possible in languages and what is not possible. Yet these Peters-Ritchie results suggested that TG was placing no constraints at all on what the grammar of a human language could be like.
Chomsky responded to all this in the early 1970s by introducing a number of changes to his framework; the result became known as the Extended Standard Theory, or EST. By the late 1970s further changes had led to a radically different version dubbed the Revised Extended Standard Theory, or REST. Among the major innovations of the REST were the introduction of traces, invisible flags marking the former positions of elements which had been moved, a reduction in the number of distinct transformations from dozens to just two, and a switch of attention away from the transformations themselves to the constraints which applied to them.
But Chomsky continued to develop his ideas, and in 1981 he published Lectures on Government and Binding; this book swept away much of the apparatus of the earlier transformational theories in favour of a dramatically different, and far more complex, approach called Government-and-Binding Theory, or GB. GB retains exactly one transformation, and, in spite of the obvious continuity between the new framework and its predecessors, the name ‘transformational grammar’ is not usually applied to GB or to its even more recent successor, the minimalist programme. Hence, for purposes of linguistic research, transformational grammar may now be regarded as dead, though its influence has been enormous, and its successors are maximally prominent.