Classifying and quantifying individual differences in phonetic categorization patterns

Abstract

A key component to speech perception is the accurate categorization of phonemes, which is classically measured by labeling functions in forced-choice tasks. The common assumption is that poor categorization drifts toward chance performance at 50%. However, the current study shows that individuals cluster in distinct patterns that diverge from that assumption. In an experiment where listeners categorized phonemes whose frequency spectra were acutely or gradually shifted upwards (reminiscent of cochlear implants), some listeners were better able to maintain distinct perceptual phonetic categories despite increasing amounts of frequency mismatch, while others could recalibrate their perception commensurate with the shift, up to a certain point before categorization ultimately broke down in a variety of ways. This study highlights the importance of classifying categorization patterns at the individual level, rather than averaging across groups, and invites reconsideration of how to quantify and interpret overlooked behavioral response patterns that signify when and how phonetic categorization breaks down. This study emphasizes provides novel ways to mathematically analyze qualitatively unique signatures of categorization breakdown (asymmetry, bias, all-or-none judgments, etc.) that are potentially ripe for studying individual differences in strategy and capacity for overcoming specific challenges relating to hearing or adaptation to acoustic degradation.