American Social Dialectology: A Statistical Appraisal

BOUT A GENERATION has passed since Glenna Pickford (1956) criticized the Linguistic Atlas for errors in sampling, and since then nearly every sociolinguistic study in the United States has attempted to be more systematic in selecting respondents.' Even reviews of sociolinguistic research such as that of Baubkus and Viereck (1973) have tended to stress the importance of adequate and reliable sampling techniques. Regardless of the correctness of Pickford's criticism, it does seem clear that one lesson statistics has taught the sociolinguist is that there are right and wrong ways to gather data. What seems to be absent, however, is sufficient critical attention to the problem of analyzing the data once the sample is properly chosen.2 In fact, only in the past decade has American social dialectology benefited from statistical analysis. Lawrence Biondi (1975), Ralph Fasold (1972), Crawford Feagin (1979), and Walt Wolfram (1974) in one way or another all use chi square (X2) tests to evaluate the reliability of results. Fasold (1972, p. 33), as a matter of fact, notes with some surprise that neither Labov (1966, 1969) nor Wolfram (1969) used statistical tests on their results: