Isoglosses and Predictive Modeling

Abstract

THE PRACTICE OF DRAWING ISOGLOSSES to chart areal variation has come in for good amount of criticism that, I would suggest, has sometimes been ill founded because of basic misconception of what isoglosses indicate. Isoglosses, and the maps on which they are charted, are conceptual models, not reproductions of reality. Just because some criticisms have been flawed, however, does not excuse those who have drawn isoglosses without paying due attention to the different possibilities for what the lines could mean within their models. There is greater need today than ever before for map makers and their audiences to be clear about what lines on maps can mean: there is greater emphasis now on quantitative analysis of dialect features and thus need to understand how both quantitative and qualitative analyses can be displayed with lines on maps. As step toward that end, I propose here to discuss the nature of lines on maps as models and to suggest appropriate roles for qualitative isoglosses and for two sorts of lines on maps with quantitative meaning. An isogloss is, or should be, nothing more than LIMIT OF OCCURRENCE, line drawn on map which separates, on one side of the line, some set of speakers who use some particular linguistic form from, on the other side, another set of speakers who do not use the form. Schneider (1988) points out that there has been heated and unresolved debate about isoglosses and their relationship to other iso lines since the turn of the century (17679). As Chambers and Trudgill (1980) suggest, the isogloss is not materially different from HETEROGLOSSES, where two lines are drawn to indicate the same sort of boundary (104-05). Chambers and Trudgill, however, criticize the notion of the isogloss because it is imprecise: speakers who were not surveyed may not be accurately labeled by the isogloss, because the line cuts arbitrarily through what is in fact unknown territory (105). This criticism is amplified by Francis (1983) in his introduction to the notion (3-7). He cites the isoglossic maps and the detailed chart of responses from Kurath (1949) for pailand bucketin the Atlantic states and concludes that a closer look at the actual data on which the line is based reveals that it is by no means as sharp and clear as it is sometimes made to appear (4). Francis notes the existence of localities where both words are used, the existence of scattered instances of one word in areas dominated by the other, and potential difference in meaning between the forms; he then asserts,