Uh, What Should We Count?

Abstract

One of the most dramatic discourse–pragmatic changes in twentieth–century English has progressed under the radar of laypeople and (until recently) linguists: the rise of um as the predominant variant of the “filled pause” variable (UHM) at the expense of uh. We investigate UHM at an early stage of change to determine what triggered its rise. We employ the variationist method to examine UHM in the Farm Work and Farm Life Since 1890 corpus of oral histories (recorded in 1984 with elderly farmers in Ontario, Canada). Nearly 5,000 tokens were extracted and coded for speaker birth year, gender, region, and utterance position. The overall frequency of um among the farmers is 11 percent. We find no significant effect of gender (12 percent for women, 10 percent for men). In one region, there is an effect of birth year. Lastly, we find no effect of utterance position. Looking at the frequency of each variant per 1,000 words, however, we see that, while the rate of um remains relatively stable, the rate of uh increases rapidly with year of birth, particularly with non–initial tokens produced by female speakers. Our results indicate that this data covers the first stage of this change.