A model was proposed as an alternative to current models for categorical perception. The model proposed that a single (common) factor causes both a peak in the discrimination function and a categorical dichotomy and thus the correlation between the two. This common factor producing categorical perception is a stable, sharp (internally or externally imposed) limitation along the continuum represented by the independent variable. The model was used to predict the existence of categorical perception for flickering visual stimuli and to create categorical perception along the auditory intensity dimension. We show that the common-factor model more adequately explains existing demonstrations of categorical perception involving the temporal onset of acoustic stimuli (i.e., stop consonants, sawtooth waveforms, and noise-buzz sequences) than do other existing models.Categorical perception refers to the apparent responding to stimuli only in absolute terms. The degree to which these stimuli are differentially identified is correlated with, and assumed to determine, their discriminability (e.g.

The ABX procedure, which has been used extensively in speech perception research, provides a great deal of face validity. Unlike the classic same—different (or AX) procedure, each member of the pair of stimuli to be discriminated are presented on every trial. Furthermore, in theory, the estimates of discriminability are relatively free from response bias. Recently, this procedure has received considerable criticism for being contaminated, on a theoretical basis, by memory effects and thus providing diminished estimates of the discriminability between stimuli. In a series of experiments, we compared (a) the typically employed “random” ABX procedure (random relative relationship between A and B), (b) a “fixed” ABX procedure (fixed relative relationship between A and B), and (c) the same—different (AX) procedure in terms of psychometric functions and isosensitivity contours using an intensity discrimination task. The results are discussed in terms of the relative sensitivity of the procedures and their susceptibility to contamination by response bias.

Until recently, categorical perception was considered to be a unique property of speech perception universal to all languages. A number of studies have demonstrated this phenomenon with complex, nonspeech auditory stimuli, but not with simple auditory stimuli. Categorical perception now is considered to be a property of complex acoustic signal processing which is of central importance to speech perception. This paper describes two studies demonstrating categorical perception: one in the auditory and one in the visual modality. In the vision study, the frequency of a single flickering light was varied. In the auditory study, the intensity of a tonal stimulus was varied in the presence of a fixed pedestal. Both studies find (a) clear ‘labeling functions separated by a sharp boundary; (b) discrimination functions with a peak at the category boundary and chance performance within categories; and (c) a high degree of correspondence between the actual discrimination data and the data predicted on the basis of the labeling data. The interpretation of these studies provides an added dimension to understanding the nature of categorical perception.

Our laboratory's recent work on the nature of masking has revealed that stimulus change can cause both forward and backward maskinglike effects. In these previous demonstrations, a decrease or shift in the noise contained in the signal channel constituted the stimulus change. To demonstrate that these masking effects were attributable to stimulus change and not the interaction of off-frequency transients, we designed the present experiment in which the stimulus change occurred in the nonsignal ear. In this case, stimulus change consisted of a shift in correlation of the noise between the two ears by inverting the contralateral noise at various times (4 < Δt < 200 msec) prior to the onset of the potential signal. Relative to the unchanging (+1 or −1) contralateral noise control conditions, the detectability of the signal was significantly decreased independent of the direction of change in correlation (+1 to −1, or −1 to +1). Effects of differing rise-fall times (1, 2.5, and 5 msec) of noise change were also examined. The results are discussed in terms of current masking theory. [This research was supported by a grant from NINDS.]

Research from our laboratory has shown that instability or changes in masking noise which are temporally contiguous with the potential signal will result in nonsimultaneous masking effects. The present experiment involves a monaural signal (1000 Hz) embedded in a broadband noise. Noise in the other earphone either was continuous, with an interaural correlation of +1 or −1 (+1 and −1 control conditions), or shifted from one correlation to the other T-ms following signal offset (T = 2, 5, 10, 20, 40, 80, 160 ms). Relative to both control conditions, both postsignal correlation-shift conditions produced greater masking, with the amount of masking decreasing as a function of increasing values of T. The results are discussed in terms of the role of noise parameter stability in masking paradigms. [Research supported by grant NS 10995 from NINCDS.]