Explaining the paradoxes
The first paradox is the belief that “language is a homogeneous system”. Language, understood as the overall ability to acquire, store and cognitively use a set of abstract constructions of whatever kind—phonological, morphological, semantic, and so on—is certainly systematic. No one would deny this. But that ability must also enable speakers and listeners to manipulate the system as and when the need arises, which is what Weinreich et al. mean by “orderly differentiation”. Homogeneity, on the other hand, implies a regularity that would contradict differentiation, and herein lies the paradox. We need a system that we share with other members of the community to which we belong or within which we are communicating, but we also need the freedom to manipulate and change the system.
Human language enables us to use the variety of language we have acquired to mediate our physical, social and mental worlds and the worlds of others. It allows us to enlarge and expand our own individual mental worlds in infinite ways. In doing this, “the actual production of syntax is locally managed” (Cumming & Ono 1997: 132), and its “‘rules’ are the construction of particular speakers” (Bex 2008 ; 222). In this sense, as Bex maintains, “grammars are ‘emergent’ at the moment of utterance” (224).
For the moment, imagine an isolated community of speakers with no or minimal contact with anyone outside that community, a rare occurrence in the modern world but still possible (see Schreier 2003). The linguistic constructions that they learn to produce in prompting for and negotiating meanings with others are used automatically without those speakers having to think twice about them. There are no “mistakes” that can be made, only meanings that are not, or not fully, negotiated. For the purposes of coexistence, collaboration and occasional conflict in the isolated community, it is immaterial what the speakers themselves call the variety of language they are using, as long as the linguistic constructions they share with others can be put to use in social practice. Why, then, is it so important to insist on homogeneous language systems when it is more important for speakers to be able to use the potential heterogeneity of human language in use?
Thus the first paradox resides in the hypostasisation of individual languages—looking for homogeneous linguistic systems rather than studying how “grammars are ‘emergent’ at the moment of utterance”. Weinreich et al. take linguistic systems themselves to show heterogeneity as well as homogeneity, but they are aware of the paradox in this assumption. In addition, they characterise the approach to language that looks for linguistic “competence” without admitting to the creative variability of language in use as leading to the attempt to create total homogeneity where it does not exist. For this reason, it is hardly surprising that Weinreich et al. characterise the approach as “needlessly unrealistic” and as “a backward step” (1968: 100).
The second and the third paradoxes are directly derivable from the paradox of a homogeneous linguistic system. The second presupposes two homogeneous states of language, distinguished by a change and ignores the variability inherent in language use in which speakers may use both the prechange and the postchange structure. The third paradox is that this way of looking at language prevents us from thinking in terms of the heterogeneity inherent within it when used in instances of social practice. It completely forecloses any notion of heterogeneity and variability.
-  This is the major problem, of course, with all forms of theoretical linguistic model, not just withgenerative models of the Chomskyan kind.