Titrating effects of acoustic variability on context effects and psychometric function slopes in speech categorization

In temporal contrast effects (TCEs, also termed speaking rate normalization), a fast context sentence makes the onset of the subsequent target word sound longer, and vice versa. King, Sharpe, Shorey, and Stilp (2022 ASA) reported that variability in the talkers or lexical contents of context sentences each decreased TCE magnitudes. However, in that experiment, acoustic variability was not tightly controlled. Using the same stimuli as King et al., listeners completed three blocks, hearing one talker speaking the same sentence (1 Talker/1 Sentence) or 200 different sentences (1/200), or 200 different talkers speaking 200 different sentences (200/200). On each trial, listeners heard one (fast/slow) context sentence, then identified the subsequent target word as “deer” or “tier.” In one experiment, speaking rates were matched across 1/200 and 200/200 blocks; TCE magnitudes were comparable and psychometric slopes did not differ. In another experiment, all slow sentences were set to 2.67 syllables/sec and all fast sentences to 8 syllables/sec. Differences due to sentence variability (1/1 versus 1/200) were extinguished; TCE magnitudes were smaller and psychometric slopes were shallower in the 200/200 block. Thus, multiple levels of variability must be considered in how they shape speaking rate normalization. [Work supported by NIH R01DC020303.]