In the field of dialectology, a diasystem or polylectal grammar is a linguistic analysis set up to encode or represent a range of related varieties in a way that displays their structural differences.

The term diasystem was coined by linguist and dialectologist Uriel Weinreich in a 1954 paper as part of an initiative in exploring how to extend advances in structuralist linguistic theory to dialectology to explain linguistic variation across dialects. Weinreich's paper inspired research in the late 1950s to test the proposal. However, the investigations soon showed it to be generally untenable, at least under structuralist theory. With the advent of generative theory in the 1960s, researchers tried applying a generative approach in developing diasystemic explanations; this also fell short.

According to some leading sociolinguists, the diasystem idea for incorporating variation into linguistic theory has been superseded by William Labov's notion of the linguistic variable. As such, the concept has not been part of any substantial linguistic theory and the term has limited currency in linguistics.^{[citation needed]}

Trubetzkoy (1931) first suggested comparing accents by their synchronic states, rather than by comparing their different historical developments. He classified sound differences between dialects into three types:

Despite Trubetzkoy's proposal, linguists continued to consider variation between varieties outside of the scope of inquiry of grammar construction; each variety, in their thinking, should only be studied on its own terms.

Inspired by Trubetzkoy, Weinreich (1954) proposed a synthesis of linguistic geography and descriptive linguistics by applying the structuralist concept of grammar to the description of regular correspondences between different varieties; a resulting supergrammar, which he called a diasystem, would be consistent with the individual grammars of all the member dialects. A diasystem is a higher order system and its component units of analysis would accordingly be abstractions of a higher order than the units of analysis of the individual systems. That is, just as the phones present within an individual variety are grouped together into abstract phonemes, the phonemes present within a group of varieties could be grouped together into even more abstract diaphonemes. Weinreich exemplified the diasystemic approach by a formulaic arrangement of phoneme correspondences in three dialects of Yiddish, focusing on the vowels but arguing that the principle could work for other aspects of language.

Although Weinreich did not elaborate the diasystemic approach, he did consider some theoretical pitfalls to be avoided. He recognized that phonemic mergers and splits with dissimilar results across dialects would pose a difficult challenge for the construction of a diasystem; he cautioned against positing a diasystem when the work of creating all the member systems (e.g. the work of phonemicization) was yet incomplete; and, following the lead of Trubetzkoy (1931), he noted that the differences in phonological inventory and etymological distribution might prove problematic in the construction of diasystems.

A few linguists took up Weinreich's challenge and quickly found it to be inadequate. Some of the failures had been anticipated by Weinreich himself, as described above.