Categorical Perception of Speech

Categorical perception of speech is the finding that listeners perceive certain continuously varying acoustic properties of speech sounds, such as the timing between a consonant's release and the onset of vocal cord vibration, not as a smooth gradient but as belonging to sharply separated, discrete phonetic categories, such as hearing a sound as either a clear ba or a clear da with very little perceived middle ground even when the underlying acoustic signal changes smoothly between the two. The psychologist Alvin Liberman and colleagues demonstrated the effect in an influential 1957 study using computer-synthesized speech sounds that varied along a single acoustic continuum, finding that listeners's identification of the sounds shifted abruptly at a boundary and, critically, that listeners could much more easily tell apart two sounds straddling that category boundary than two equally different sounds falling within the same category. This pattern, sharp identification boundaries paired with a discrimination advantage across those same boundaries, is the specific signature researchers use to identify categorical perception, and it distinguishes speech sound perception from many other auditory judgments, such as judging pure tone pitch, which vary much more continuously with the underlying physical signal. The phenomenon has been central to longstanding debates in psycholinguistics over whether speech perception relies on specialized mechanisms unique to language or draws on more general auditory and categorization processes also seen in other perceptual domains.