SLHS 1301 - Speech perception 2 - categorical perception

Summary

The video introduces categorical perception: the idea that listeners often perceive sounds according to the category they belong to, rather than as a smooth series of physical differences.

What categorical perception means

• The presenter compares speech categories to state boundaries. Two people can live close together but have different state license plates if they are on opposite sides of a border. The labels reflect category membership, not distance.
• Similarly, a series of gradually changing stimuli may be heard as belonging to one category until a boundary is crossed, at which point the perceived label changes.
• The color example illustrates the contrast: continuous perception registers each shade of blue as slightly different, while categorical perception treats the blue shades as alike and green as a different category.

Demonstrating categorical perception with speech

The video uses voice onset time (VOT)—the timing between the release of a stop consonant and the start of vocal-fold vibration—to illustrate a continuum between sounds labeled “ba” and “pa.” The sound changes gradually, but listeners tend to report one category or the other rather than an intermediate sound.

Researchers use two experiments to test for categorical perception:

Identification test

• Present one sound at a time and ask the listener to choose between two labels, such as “ba” or “pa.”
• Vary the sound’s VOT in small steps, such as five-millisecond increments.
• At short VOTs, listeners consistently choose one label; at longer VOTs, they consistently choose the other.
• Responses change sharply around the category boundary—roughly 22–23 milliseconds in the example—rather than gradually across the continuum.

Discrimination test

• Present two sounds and ask whether they are the same or different.
• When two sounds are difficult to tell apart, responses should be near 50% correct, or chance performance.
• Discrimination rises sharply for a pair that crosses the category boundary, then drops again for pairs farther along the continuum.
• This produces a discrimination peak at approximately the same point where identification results switch categories.

Together, the classic signs of categorical perception are a steep shift in identification and a discrimination peak at the category boundary.

Other speech continua

The presenter notes that categorical shifts can also occur along continua between other speech sounds, including examples resembling “sha” and “saw,” as well as sequences involving “da,” “ba,” and “cha.” In each case, perception may appear to jump from one sound label to another as the acoustic signal changes gradually.

Limits of the categorical-perception account

Categorical perception is an important concept, but the presenter cautions that it is not a complete description of speech perception:

• Listeners can detect differences between sounds within the same category. This sensitivity can help distinguish talkers, anticipate upcoming words, and track systematic changes in speech.
• Experimental methods can encourage categorical-looking results. For example, a forced-choice test that offers only “ba” and “pa” cannot reveal whether listeners experience a sound as intermediate or as partly belonging to each category.
• Categorical perception is best treated as a useful framework or starting point, not a full account of how speech is perceived.

Multiple acoustic cues

Speech perception depends on a constellation of acoustic cues, not just one feature such as VOT. In comparing the example words “loss” and “laws,” the presenter points to several differences that can contribute to perception:

• Low-frequency energy associated with voicing.
• The frequency pattern and movement of the first formant.
• Vowel duration.
• Fricative duration and its duration relative to the vowel.
• Duration of the “l” and the onset frequency of the second formant.
• Differences in the signal’s overall amplitude-envelope structure.

Which cues matter most can vary; any one cue may take priority over others in a particular context.

Speakers and sources featured

• Presenter/narrator: Explains the concepts and experiments in a lecture-style voiceover.
• Speech-sound examples: The video uses spoken or presented sound stimuli to demonstrate speech categories and acoustic contrasts. No individual stimulus speaker is identified.
• Illustrative characters: Jen, Tara, Jim, and Ruth are mentioned in the state-boundary analogy, not identified as speakers.