Examining the consistency of spectral context effects across frequency regions in speech perception

Speech categorization is influenced by spectral contrast effects, or the perceptual magnification of spectral differences between successive sounds. Spectral contrast effects result in the categorization of a target sound being biased away from spectral properties in the preceding acoustic context. Because of the remarkable consistency in the magnitudes of these contrast effects within the same frequency region, we hypothesized that they would also show a stable relationship across different frequency regions. In this study, normal-hearing listeners' phoneme categorization and contrast effects were assessed where phonetic contrasts were driven by changes in low-frequency F1 (“big”-“beg”), mid-frequency F3 (“dot”-“got”), or high-frequency regions (frication spectrum in “sheet”-“seat”). On each trial, listeners heard a precursor sentence that was filtered to emphasize energy in the lower or higher range for one of these frequency regions, followed by a target word that hinged on that filtered frequency region. Spectral contrast effects influenced categorization in each frequency region, as expected. However, effect magnitudes were not reliably correlated with each other across frequency regions. This further clarifies perception-in-context on a broader scale, as using spectral context during speech categorization may be consistent within a single frequency region but not across different frequency regions.