The phonemes /ba/ and /pa/ differ mainly in the delay between opening of the vocal tract and vibration of the vocal cords. This is the voice-onset time (VOT), illustrated in . Using a computer, it is possible to generate artificial sounds that vary between a natural /ba/ (VOT = 0 ms) and a natural /pa/ (VOT = 50 ms). How are these intermediate sounds perceived? Do they sound partway between /ba/ and /pa/, sometimes like one and sometimes like the other, or like something else entirely? Test yourself with (~10 minutes).
Your results probably show an abrupt transition between your perception of /ba/ and /pa/. Given what you have learned about sensory systems, this may seem odd. When we perceive simple stimuli such as color or pitch, a small change in the stimulus leads to a small change in perception. With the phonemes, a small change in VOT did not change your perception until that change crossed a boundary, when it caused a large change in your perception. This is called categorical perception, and you have just determined your labeling function for VOT. This function revealed two distinct categories, /ba/ and /pa/.
continuous
categorical
To show that perception is truly categorical, one must test discrimination as well as labeling. Not only must there be a sharp boundary between the two categories, one must be less able to discriminate between pairs of stimuli that come from the same category. Test your discrimination function with (~10 minutes).
Your results probably show a peak in discrimination when the stimuli in a pair fell on opposite sides of the boundary determined by the labeling test. A typical set of results looks like . You have now demonstrated categorical perception of /ba/ and /pa/. This has been shown for many pairs of phonemes that vary in VOT or some other aspect. However, categorical perception is not absolute. If you listen to the stimuli many times, you may begin to hear differences between stimuli that belong in the same category.
What do you expect to happen to your labeling and discrimination functions if you selectively adapt your /ba/ detectors? Eimas and Corbit (1973) predicted that adaptation would shift the categorical boundary and discrimination peak to the left (toward /ba/). You can approximate their experiments with the following, which repeats /ba/ or /pa/ 10 times before each trial.
Computer-generated /ba/ and /pa/ sounds were provided by Bob McMurray, at the time a graduate student at Rochester University.
To demonstrate categorical labeling without doing the full experiment, you can use the following links to play through the stimuli in order from -10 to +50 ms VOT.
You should find that your perception shifts from /ba/ to /pa/ abruptly along this continuum.
,
,
,
,
,
,
,
,
,
,
,
,
.
To quickly test the hypotheses of Eimas and Corbit (1973) described in Further Exploration, listen to the −10 ms VOT (/ba/) repeatedly, then listen to a VOT that you previously found to be just barely on the /ba/ side of the VOT spectrum. (Hold the shift key when clicking an audio link to make the sound loop repeatedly, then click the link again to stop it.)