A blind spot of variation studies:

Structural dependence between linguistic variables

Frans Hinskens

Sun. 9-10:40 C

Dialect use can consist of more than simply using dialect features, i.e. the non-prestigious variants of linguistic variables. Especially in the case of 'divergent dialects', dialects which display a "considerable amount of linguistic distance from the national standard" (Trudgill 1986:91), in the realization of words or phrases, several dialect features can play role at the same time. The dialect features involved can then be either structurally independent or structurally dependent on one another. In the first case, the presence or absence of the dialect feature in one linguistic variable does not in any way structurally influence the presence or absence of the dialect feature in the other linguistic variable(s) - although the linguistic variables concerned can covary. In the second case, there are two possible relationships between the linguistic variables involved. One can be informally labelled disjunction (EITHER x OR y, in logic called 'exclusive or'), the other conditionality (IF x THEN y, in logic called 'strong implication') - cf. the types of relationships between rules of phonology or morpho-phonology which are referred to as 'bleeding' and 'feeding', respectively. These structural relationships between linguistic variables will form the topic of this talk, in which I will present what is intended as an exploration of a new approach.

It should be stressed that I will not be dealing with covariation. Covariation is a statistical relation which can vary between -1 and +1, except 0. In the case of covariation (Weinreich, Labov & Herzog 1968; Labov 1971), the chances that, say, dialect features of different linguistic variables are used are mutually dependent to some extent. Covariation can be a stylistic device. Neither will I be dealing with the use of statistical techniques to the inference of rule order (Sankoff & Rousseau 1989).

In certain branches of sociolinguistics and creolistics scalogram analysis has been used to investigate whether and to what extent tendential, statistical implicational relationships (i.e. not 'strong' ones) exist between linguistic variables.

I will be concerned with relationships of a structural rather than a statistical nature. The relevant types of variation are concealed within a blind spot of the standard sociolinguistic practice of analyzing isolated linguistic variables in the speech material.

For the two types of structural dependence, the number of possible combinations of two or more dialect features within one word or phrase, i.e. the number of possible realization types, is logically limited. In each case of structural dependence between dialect features, the realization types can on objective grouns be ordered according to their degree of dialectality. This approach has been developed on the basis of data from a specific divergent dialect of Dutch. A part of the speech material I collected for this dialect makes it possible to systematically analyse both types of structural dependencies between certain dialect features in the phonological, morpho-phonological and morphological components. The subject of the research was dialect levelling, the process of the reduction of structural variation. Therefore, the several realization types that are allowed by each of the two dependence relations, along with occasional other realization types, were studied in apparent time.

References

Labov, William, 1971, "The notion of 'system' in creole languages", in: Dell Hymes (ed.), Pidginization and creolization of languages. Cambridge. 447-472

Sankoff, David & Pascale Rousseau, 1989, "Statistical evidence for rule ordering", Language variation and change 1, 1, 1-18

Trudgill, Peter, 1986, Dialects in contact. Oxford

Weinreich, Uriel, William Labov & Marvin Herzog, 1968, "Empirical foundations for a theory of language change", in: Winfred Lehmann & Yakov Malkiel (eds), . Austin etc. 95-189Directions for historical linguistics; a symposium