The basic sociolinguistic model for relationships between idiolects anddialects is the speech community, defined by shared linguistic features andattitudes and relatively high internal density of communication. Since thesedefinitions are relative, speech communities can nest and overlap, so that localsubdialects, sociolects, ethnic lects, and personal networks can form smallerspeech communities that share local traits and are locally high in communicationdensity, while still belonging to a broader speech community which shares widertraits, and whose internal communication density is high, relative to, say, othergeographic regions. In variation studies, it has long been assumed that some of theshared linguistic traits that define a community are certain constraints on variableprocesses. The conceptual problem that arises with this model is, how far areconstraints shared, and how much can they differ? If some constraint was due, forexample, to a universal process, then it would be expected to be shared by allspeakers of all speech communities, while at the other extreme, the existence ofidiosyncratic differences in language usage raises the possibility that at least someconstraints may differ for each individual. This paper takes a cross-dialectalcomparative approach. Two variable processes are studied in four communitiesdrawn from the VARSUL corpus, each with distinctive ethnic and sociolinguisticcharacteristics. A socially diversified sample of 8-12 speakers is investigated ineach community. The variables investigated include one syntactic process (nounphrase Agreement, NPA) and one phonological process (final -s deletion, SDEL).The constraints on NPA are mainly morphosyntactic in nature while theconstraints on SDEL are straightforwardly phonological. In each case, theconstraint effects are broadly similar across communities and speakers. Betweenspeakerdifferences within communities are mainly the result of either statisticalnoise (smaller sample sizes lead to larger differences), or of predictable socialdifferentiation (e.g. speakers with less formal education use more of the nonstandard variants.) And, strikingly, the main constraint effects are highlyconsistent across the different communities. The results generally lend support tothe model of Cedergren & Sankoff (1974), that "performance is a statisticalreflection of competence", and competence, here dealing with the knowledge ofwhat varies where, is powerfully shared across a language community.
INSERTION AS A SYNTACTIC CONSTRUCT AND SYNTACTIC PROCESS
