Michael I. Miller

Three Changing Verbs: bite, ride, and shrink

One of Raven McDavid’s favorite aphorisms was that theory can wait, data cannot. Yet McDavid believed firmly that success in linguistics depends less on coming up with the right answers than in asking the right questions, a view that implies a clearly formulated theoretical foundation from which questions can be asked. The primary purpose of this paper is to present some data, but I would also like to consider the implications of the data for a descriptive and historical account of American English. My controlling assumption is the frequent observation that morphological variation in both British and American spoken grammar has leveled rapidly, especially since World War II. By focusing on a small, easily controlled data-set, I would like to explore several broader questions in American and British dialectology relating to this fact. I have four questions. First, leveling is such a pervasive feature of American English, and apparently has been since the eighteenth century, that my initial question is less about the extent dialect leveling has proceeded in urban American speech than how the forms observable in a narrow urban field relate to the known forms in other areas of the English-speaking world. My second question, therefore, is what constant factors correlate with the differences between the urban data and data from elsewhere? Third, what maintains the relic forms that persist in urban speech? And fourth, what are the relationships among the countervailing forces that work either to maintain relics or introduce innovations? A full answer to these questions would require

Encoding, Data Definition, and Command Syntax for Linguistic Survey Data

Dialectologists generally work with three kinds of quantified data displays: frequency tables, cross-classification (or &dquo;contingency&dquo;) tables, and summary tables, such as lists of averages. We often combine these numerical tables with graphic devices, like bar charts or scatterplots. And we usually present these displays along with statistical tests and measures, such as Student’s T. In previous papers (Miller 1988a, 1988c, 1988d) I have shown how off-the-shelf statistical packages can manipulate linguistic survey data to produce the data displays we need, along with a rich array of statistical tests and measures. This paper describes how to organize the input in order to get that output. When dialectologists collect what they call &dquo;data&dquo;, the work sheet or protocol for any given informant might look something like Figure 1. These transcriptions represent a considerable abstraction from raw phenomena, since they depend not only on theories about the relationship between writing and speech but also on technical decisions about the application of a specific phonetic alphabet. These data are already &dquo;encoded&dquo; in a form readable by other dialectologists, but dialectologists seldom present data in this raw form to other linguists. The dialectologist’s next step has usually been to produce a &dquo;list manuscript&dquo; which gathers together all the responses to a single query (Figure 2).