The sound of a category
In the 1950s, researchers at Haskins Laboratories, now located in New Haven, asked a fundamental question: how does the brain convert a continuous stream of sound into the discrete units of language we call phonemes? They focused on the distinction between voiced and voiceless consonants, like "ba" and "pa". The main physical difference between these sounds is a brief delay called Voice Onset Time (VOT). VOT is the time between the puff of air released when the lips part and the beginning of vocal cord vibration. For "ba", the vibration starts almost immediately (a VOT near 0 milliseconds), while for "pa" there is a longer delay (around 50 milliseconds).
Using a device called the Pattern Playback, invented at the lab by Franklin S. Cooper, researchers could precisely control this timing. The machine converted hand-painted spectrograms, visual representations of sound, back into audible speech. The team, including Alvin Liberman, created a series of sounds with VOTs that increased in small, equal steps, from a clear "ba" to a clear "pa". When they played these sounds to listeners, the perception did not change gradually. Instead, people reported hearing "ba" up to a certain point, and then their perception abruptly flipped to "pa". They had identified a perceptual boundary. This phenomenon, where a continuous physical variable is perceived in discrete categories, was named categorical perception.
The brain's filing system
The experiments at Haskins demonstrated that the human brain does not process all acoustic differences equally. For English speakers, the boundary between "ba" and "pa" occurs at a VOT of about +25 to +30 milliseconds. Two sounds with VOTs of 5 and 15 milliseconds are both clearly identified as "ba". A sound with a VOT of 40 milliseconds is clearly "pa". Listeners find it very difficult to distinguish two sounds that fall on the same side of this boundary, even if the physical difference in VOT is significant. They can, however, easily distinguish two sounds with a much smaller physical difference if they fall on opposite sides of the boundary.
This effect is not universal; the location of the boundary changes depending on language. Spanish speakers, for instance, have a different VOT boundary that distinguishes their voiced and voiceless stops. Further research has shown that categorical perception is not limited to speech. It appears in how we perceive colors and musical tones, suggesting it is a general principle of how the brain handles sensory information. Studies have even shown that one-month-old infants and chinchillas perceive the "ba"/"pa" continuum categorically, pointing to deep biological roots for this method of processing sound. The work started here showed a mechanism of perception: the brain actively sorts messy, continuous sensory input into neat, useful categories.