Mapping Southern English

It all depends on how you look at it. Even in the case of Southern English, the most commonly mentioned North American dialect area (cf. Metcalf 1997), to create a map of language variation is not simply a matter of finding the correct, or even defensible, boundaries for a region. Mapping Southern English must always be a theory-driven enterprise: one's map must necessarily embody those notions about language and dialect, one's own particular set of decisions, that work together to constitute the model of which the map is just an outward expression (see Kretzschmar 1992). Two opposing factors, linguistic perception versus linguistic production, condition the different models, and thus different maps, of Southern English produced by scholars. Unfortunately, length limitations force the omission from this essay of perceptual maps such as those recently developed by Preston (1989, 1997, 1999) and Tamasi (2000, 2001). A second important opposition, however—which is explored here—is whether maps display (describe) or attempt to interpret the variation they encompass (Chambers and Trudgill 1998, 25-29). No map of Southern English is the "best" one, because each has arisen from different assumptions and principles. However, maps can be considered as products of the decisions of their makers, and in those terms their relative success can be assessed. To that end, in this essay I offer a tour of a number of maps of Southern English production, commenting on the special features of each one as a way, in the end, to permit a clearer understanding about how we might see Southern English.

The classic traditional maps of Southern English are based on linguistic production, not mainly perception (although we shall see that perception is not irrelevant to them). The earliest production map of Southern English of which I am aware is Hempl's (1896b) map for to grease/greasy (fig. 1), which is based on 1,600 responses to a postal questionnaire that he had distributed previously (1896a).

The map includes four regions, North, Midland, South, and West, about which Hempl says, "In this report the attempt is made to apply this division of the country (as determined by the replies to several other questions) to the answers so far received as to the pronunciation of 'to grease' and 'greasy.' It will be seen from the accompanying table that the general justice of the division is abundantly verified" (1896b, 439). Hempl's map is descriptive, in that it reports the percentage of [s] pronunciations reported from each location, but it is also interpretive, in that it divides the country with "general justice . . . abundantly verified" into four regions. Hempl considered the evidence of the distribution of this single feature to confirm the division that he had made earlier, because he noticed that the frequencies of its use in the North and South at the extremes, and of the Midland in the middle, are "fairly uniform," while the West is more mixed: "the . . . general division of the country into four sections can be but little out of the way" (1896b, 438; emphasis in original). Thus Hempl's interpretive model creates his divisions on the basis of quantitative tendencies, not self-consistent systems; his dividing lines are not so much dialect boundaries as they are a sort of ad hoc means of separation (Hempl, for instance, often talks about the potential mutual influence of the different areas). However, just where he drew his dividing lines ultimately comes down to his perception of where they ought to go: Should part of Missouri really belong to the South and another part to the West? Should the western parts of his Midland really go with the South or be separately marked as a region because the frequencies there are lower than they are in the eastern part of the Midland? Hempl's model cannot answer questions like these, and his assessment of a "general" division whose "general justice is abundantly verified" can come only from the fact that these dividing...