A level of representation in generative phonology which sets up a single underlying form capable of accounting for the phonological variations which relate grammatical structures (e.g. words). In such pairs of words as divine ~ divinity, obscene ~ obscenity, there is plainly a regular relationship of some kind, but it is not an easy relationship to state explicitly. Chonsky and Halle, in their approach to this problem, argue that the root morpheme in each pair of words can be given a single underlying representation, and that this, along with the rules which relate such representations to the surface alternants, accounts for the native-speaker’s awareness of the ‘systematic’ relationship which exist between grammar and phonology. (Such rules also often reflect sound changes which have taken place in the history of the language.) The units in these representations are referred to as systematic phonemes, as opposed to the ‘autonomous’ phonemes of traditional phonemic phonology, which are established without reference to grammatical structure. Some generative phonologists (such as Chomsky and Halle) prefer the term ‘phonological’ to refer to this level representation, because of the undesirable associations of the term ‘phonemic’ with traditional phonemic theory.