In linguistics, transformational grammar (TG) or transformational-generative grammar (TGG) was the earliest model of grammar proposed within the research tradition of generative grammar. Like current generative theories, it treated grammar as a system of formal rules that generate all and only grammatical sentences of a given language. What was distinctive about transformational grammar was that it posited transformation rules that mapped a sentence's deep structure to its pronounced form. For example, in many variants of transformational grammar, the English active voice sentence "Emma saw Daisy" and its passive counterpart "Daisy was seen by Emma" share a common deep structure generated by phrase structure rules, differing only in that the latter's structure is modified by a passivization transformation rule.

Transformational grammar was a species of generative grammar and shared many of its goals and postulations, including the notion of linguistics as a cognitive science, the need for formal explicitness, and the competence-performance distinction. Transformational grammar included two kinds of rules: phrase-structure rules and transformational rules.

In transformational grammar, each sentence in a language has two levels of representation: a deep structure and a surface structure. The deep structure represents a sentence's core semantic relations and is mapped onto the surface structure, which follows the sentence's phonological system very closely, via transformations.

Deep structures are generated by phrase structure grammars using rewrite rules.

Transformations are rules that map a deep structure to a surface structure. For example, a typical transformation in TG is subject-auxiliary inversion (SAI). That rule takes as its input a declarative sentence with an auxiliary, such as "John has eaten all the heirloom tomatoes", and transforms it into "Has John eaten all the heirloom tomatoes?" In the original formulation (Chomsky 1957), those rules held over strings of terminals, constituent symbols or both.

(NP = Noun Phrase and AUX = Auxiliary)

In the 1970s, by the time of the Extended Standard Theory, following Joseph Emonds's work on structure preservation, transformations came to be viewed as holding over trees. By the end of government and binding theory, in the late 1980s, transformations were no longer structure-changing operations at all; instead, they added information to already existing trees by copying constituents.

The earliest conceptions of transformations were that they were construction-specific devices. For example, there was a transformation that turned active sentences into passive ones. A different transformation raised embedded subjects into main clause subject position in sentences such as "John seems to have gone", and a third reordered arguments in the dative alternation. With the shift from rules to principles and constraints in the 1970s, those construction-specific transformations morphed into general rules (all the examples just mentioned are instances of NP movement), which eventually changed into the single general rule move alpha or Move.