Within generative grammar, generative semantics is a variation on transformation grammar developed in the mid-1960s by some of Chomsky’s pupils at the time: John R. Ross, Paul Postal, James McCawley and George Lakoff.
As in Noam Chomsky’s grammar, a deep syntactic structure was assumed, but the idea was added that these deep structures also directly underlie the semantic interpretation of language utterances. In response, the later proponents of generative semantics regarding depth structure developed new theories that were more abstract and complex. For example, in the transition from the depth structure to the surface structure, transformations took place that were much more abstract and complicated than assumed in Chomsky’s grammar. Generative semantics, for example, assumes that the relationships between phonological representations (such as phonemes) and semantic representations can be made visible by means of transformations.
Differences from other models
Generative semantics is completely at odds with what is claimed by Chomsky and other linguists such as Ray Jackendoff, namely that generative syntax and interpretive semantics are two completely separate circuits. Particularly in the late 1960s and 1970s, this contradiction gave rise to much fierce debate between followers of generative semantics and the more “orthodox” followers of Chomsky. In the end, generative semantics lost out to traditional Chomsky grammar, but actually mainly because interest in it had declined in general. In 1980, a research project carried out within the framework of generative semantics was discontinued. Certain ideas from generative semantics have been transferred to contemporary disciplines, in particular pragmatics and cognitive linguistics.