Cue weighting for speech categorization changes based on regularities in short‐term speech input.

Abstract

The ability to flexibly adapt long‐term speech category representations to informative regularities in short‐term input is critical for on‐line speech perception. The present experiment investigates how short‐term changes in the variability of two distinct acoustic cues affect the relative weighting of the cues for speech categorization. Native English adults distinguish the vowel categories /ae/ and /E/ using both spectral and duration cues. A spectral continuum from /ae/ to /E/ was crossed with a duration continuum between the same vowels to synthesize a 2‐D grid of words ranging between “set” and “sat.” Baseline categorization data were collected from trials sampling the full grid of words; for each trial, listeners selected whether they heard set or sat. Then, they received short‐term exposure to trials drawn only from subsets of stimuli for which one cue was held constant while the other cue exhibited the full range of variability. Results reveal that listeners shift their relative cue weights, compared to baseline, to rely more on the highly variable cue for categorization. This suggests that listeners track the variability present across multiple acoustic cues and dynamically adjust speech categorization to reflect this short‐term statistical regularity. [Work supported by NIH and NSF.]