Even if unity is something to strive for, and not something given, it does not automatically follow that all struggle for unity is equally attractive: there is a significant difference in the kind of unity one tries to achieve. Generally, Gramsci prefers progressive struggles to conservative ones. But what 'practical' criteria are there to distinguish progressivism from conservatism?
Gramsci conceives of 'grammar' as a historically developing system of rules (Rosiello (1970: 358); cf. (PN 399-402); Gramsci (1967: 253)). Error presupposes a rule: viz., the rule of speaking in accordance with an immanent grammar. But within the conceptual model of Gramsci, immanence can only be defined by 'rules' we make up (and change!) as we go along. Like the unity of language, also the unity of its rules is not given a priori. The distinction between continuing in the same way and continuing differently is a historical one. Gramsci's concept of rule thus parallels that of Ludwig Wittgenstein ((1968); henceforth: PI; esp. 79-85, 197-242, cf. also Wittgenstein (1969)). Since adaptations of a rule, as new forms of use, are not given until they are realized, we have to form what Wittgenstein calls a 'tradition'. Such a tradition is exclusively based on a continuous replacement of instances. To follow a rule is to use a technique of doing the 'same' in different circumstances. Anything can count as the 'same' as long as people agree in their forms of life, i.e. mutually accept it as the same. The practical reactions of the community constituting the identities of itself, its subjects, and its objects decide whether or not a certain act follows a rule.
Gramsci's concept of language as social practice can be seen as a politicized version of Wittgenstein's 'forms of life'. It is concerned with grammar as 'history' or a 'historical document', as the 'photography' of a certain stage in the development of language (which "shows how people think and live"; PI 325); a fixation of something that is not fixed. Giving up all hope of finding 'the essence of language' now allows to pose a more practical question: what is the purpose of such a fixation? Is it just to fix, or to change, an aspect of civilisation? (CW 180); cf. Gramsci (1967: 268). For Wittgenstein, change and [p. 557:] conflict were easily overcome. Within a tradition, such epistemological problems do not create practical problems (cf. Wittgenstein (1969, e.g. 63, 347, 512, 611-612)). But, for Gramsci, the involvements of grammar are practical before becoming epistemological.
Immanent grammar, therefore, cannot reduce the multiplicity of language to a coherent whole. (As such it is almost the opposite of Chomskyan concepts of competence.) It does not 'found' the difference between grammaticality and error. Language is self-evident only within spontaneous and uncritical everyday thought. Only because of the contradictions between verbal and practical knowledge, language can be seen as "a totality of determined notions and concepts and not just of words grammatically devoid of content" (PN 323). Therefore, for the unification of language, immanent grammar is less important, too, than the way it is expressed, the kind of coherence that can be achieved at the level of common sense. At this point, Gramsci distinguishes a normative grammar, which supplements the immanent grammar of everyday language (like lingua supplements linguaggio). Gramsci's opposition to the idea of fixed norms (cf. (PN 430, 438, 462) on sociology) does not exclude all normativity. Norms are not given by nature, but they are "in a sense, and up to a certain point, (...) 'creations', which are closely linked to the interests of society" (PN 466). Gramsci defines normative grammar by reciprocal control, teaching, and 'censorship' (manifested in language by questions like 'What do you mean?', or, for example, by mimicry, or teasing) which combine to establish a grammatical conformism, i.e. 'norms' and judgments of correctness and incorrectness (CW 180-182); cf. Rosiello (1970: 356-357). These 'spontaneous' expressions of grammatical conformism are necessarily incoherent, interrupted, and limited to social strata, local centres, and so on. Normative grammar always presupposes a historical 'choice', and thus, norms are always acts of cultural-national policy: they are not predictions, but declare intentions. In this way, every grammar is normative, because reflection on language is always born out of political needs (CW 184-185).
Gramsci is in line with Wittgenstein; normativity is not a matter of opinions, but of a form of life (PI 241). Norms and rules develop within a community, parallel to the political aims the community produces, in other words, to its self-definition. Normative grammar, and the efforts spent in patiently learning it, discipline people (in addition to enabling them to know their way around in the world); they define a form of life, and do not have any success a priori. The definition of progressive development, therefore, depends for Gramsci, as for Wittgenstein, on historical development itself. But for Gramsci this development is more fundamentally social, and in other ways than for Wittgenstein. Whereas the latter claims that the lack of immanent foundation (which was what Croce searched for) is not a problem within 'our' everyday paradigms, Gramsci considers these paradigms in [p. 558:] themselves problematic, not metaphysically, but in practice: as part of a hegemonial struggle.