Variable rules and literary style

Abstract

Since the time when Labov1 introduced the notion of the variable linguistic rule, several researchers into linguistic variability have suggested that the idea may be of use in the study of style. Cedergren and D. Sankoff claim that the idea of the variable rule "has wide ramifications for stylistics,"2 echoing a sentiment expressed by G. Sankoff.3 And Enkvist states that "Labov's method is of potential interest.. . to students of style precisely because it shows us an example of how probabilistic quantification can be built into a generative grammar."4 But as yet no effort has been made to determine whether stylistic variation can in fact be described in the framework of variable rules. This paper reports on a preliminary attempt to describe some aspects of the metrical styles of Robert Herrick and Ben Jonson according to the model of variable rules presented in Cedergren and D. Sankoff.s Metrical style was chosen as the area of investigation for two reasons. Cedergren points out that "Phonology has provided the first testing ground for theories of linguistic variability partly due to the possibility of constructing data sets large enough for simultaneous analysis along many dimensions of variation. With massive data, we are more likely to be bound by the patterning existing in the data, and less likely to resort to generating biased artificial data in order to. fill vacant 'cells' in tables, to be tempted into lengthy a posteriori speculations to explain away each piece of data which does not fit our models, or to dismiss systematically patterned exceptions as 'performance error.' "6 A phonological aspect of stylistic variability-specifically, the presence or absence of the phonological feature of stress-is desirable as a "first testing ground" for the theory of stylistic variability precisely because it enables us to generate massive data sets. A 1136 line sample of iambic pentameter verse from Herrick and Jonson provided an opportunity to measure the variable of stress 4544 times, a number sufficient for a consideration of many relevant dimensions of variation. A syntactic or lexical variable would not readily supply us with a large enough data set. A second reason for choosing metrical style is that there exists already a generative grammar of iambic pentameter which provides a ready-made framework for the investigation of metrical variation according to a variable rule. The Halle-Keyser theory7 regards the underlying structure of the iambic pentameter line as a series of positions: