Categorical perception means that a change in some variable along a continuum is perceived, not as gradual but as instances of discrete categories. The test presented here is a classical demonstration of categorical perception for a certain type of speech-like stimuli. The stimuli in the experiment are synthetic syllables in which the second formant is varied in equal steps. Depending on the precise parameter values, the resulting sounds are perceived as `ba', `da' or 'ga'. Although the change is gradual, this is not the way it is perceived. Subjects regularly perceive the different stimuli as being instances of either of the three syllable types, `ba', `da' or `ga'.
This is not enough, however, for perception to be regarded as categorical. It is also necessary that discrimination between stimuli is much more accurate between categories than within them. Ideally, all stimuli in a given category should be perceived as indistinguishable, whereas stimuli from different categories, no matter how close on the continuum, should be perceived as different. A test for categorical perception thus consists of two parts, a labelling task and a discrimination task.
The experiment demonstrated here was first performed in the fifties and was then taken as evidence that speech is perceived differently from other types of auditory stimuli. Later research has shown, however, that categorical perception is not restricted to speech or speech like stimuli but may occur with stimuli that bear no resemblance to speech sounds. It has furthermore been demonstrated that categorical perception may also occur in animals. Categorical perception experiments on Voice Onset Time with chinchillas, for example, have produced results that closely match those obtained in experiments with human subjects.
The experiment you are about to try still suffers from some technical shortcomings. Under certain circumstances there may be short interruptions in the sounds. As far as we are aware this is due to the way sounds are played in Java, and cannot be remedied at the moment. In our experience, though, the results of the tests are not seriously affected by this problem.