The concept of deep structure plays an important role in transformational grammar. In early transformational syntax, deep structures are derivation trees of a context free language. These trees are then transformed by a sequence of tree rewriting operations ("transformations") into surface structures. The terminal yield of a surface structure tree, the surface form, is then predicted to be a grammatical sentence of the language being studied. The role and significance of deep structure changed a great deal as Chomsky developed his theories, and since the mid 1990s deep structure no longer features at all (see Transformational grammar).
It is tempting to regard deep structures as representing meanings and surface structures as representing sentences expressing those meanings, but this is not the concept of deep structure favoured by Chomsky. Rather, a sentence more closely corresponds to a deep structure paired with the surface structure derived from it, with an additional phonetic form obtained from processing of the surface structure. It has been variously suggested that the interpretation of a sentence is determined by its deep structure alone, by a combination of its deep and surface structures, or by some other level of representation altogether (logical form), as argued in 1977 by Chomsky's student Robert May. Chomsky may have tentatively entertained the first of these ideas in the early 1960s, but quickly moved away from it to the second, and finally the third. Throughout the 1960s and 1970s, the generative semantics movement put up a vigorous defence of the first option, sparking an acrimonious debate, the "Linguistics Wars".
The "surface" appeal of the deep structure concept soon led people from unrelated fields (architecture, music, politics, and even ritual studies) to use the term to express various concepts in their own work. In common usage, the term is often used as a synonym for universal grammar — the constraints which Chomsky claims govern the overall forms of linguistic expression available to the human species. This is probably due to the importance of deep structure in Chomsky's earlier work on universal grammar, though his concept of universal grammar is logically independent of any particular theoretical construct, including deep structure.
According to Middleton (1990), Schenkerian analysis of music corresponds to the Chomskyan notion of deep structure, applying to a two-level generative structure for melody, harmony, and rhythm, of which the analysis by Lee (1985) of rhythmical structure is an instance. See also Chord progression#Rewrite rules.