An experimental investigation of acoustic task-based carryover effects

Abstract

Carryover effects, where conditions in one experimental task affect behavior in subsequent tasks, have not received much attention in linguistics. This is worrying, given that linguistic elicitation tasks do affect talkers' production (Labov et al. 1972, Warner et al. 2012). This study used before and after within-subjects design. 16 talkers completed three tasks: a memory task, one of a set of intermediate tasks (passage reading, word reading or an interview), and a second memory task. Pitch, intensity and duration of stressed vowels were measured for productions from each task. Nested mixed linear effects models were constructed for each measure. In all models, acoustic measures were used as the response variable and subject and token as random intercepts. For data from the intermediate tasks, the inclusion of task identity significantly improved model fit for pitch (p < 0.05), intensity (p < 0.001) and duration (p < 0.001). In other words, there were very strong task effects. However, when comparing da...