The sensory encoding of the duration, interval, and order of different stimulus features provides vital information to the nervous system. The present study focuses on the influence of practice on auditory temporal-interval discrimination. The goals of the experiment were to determine (1) whether practice improved the ability to discriminate a standard interval of 100 msec bounded by brief 1 kHz tones from longer intervals, and, if so, (2) whether this improvement generalized to different tonal frequencies or temporal intervals. Learning was examined in 14 human subjects using an adaptive, two-alternative, forcedchoice procedure. One hour of training per day for 10 d led to marked improvements in the ability to discriminate between the standard and longer intervals. The generalization of learning was evaluated by independently varying the spectral (tonal frequency) and temporal (interval) components of the stimuli in four conditions tested both before and after the training phase. Remarkably, there was complete generalization to the trained interval of 100 msec bounded by tones at the untrained frequency of 4 kHz, but no generalization to the untrained intervals of 50, 200, or 500 msec bounded by tones at the trained frequency of 1 kHz. Thus, these data show that (1) temporalinterval discrimination using a 100-msec standard undergoes perceptual learning, and (2) the neural mechanisms underlying this learning are temporally, but not spectrally, specific. These results are compared with those from previous investigations of learning in visual spatial tasks, and are discussed in relation to biologically plausible models of temporal processing.
Key words: psychophysics; hearing; perceptual learning; auditory processing; temporal processing; humanThe sensory encoding of temporal information such as the duration, interval, and order of different stimulus features provides vital information to the nervous system. This is clearly illustrated by the ever-increasing evidence of the importance of temporal cues in the perception of speech. The identification of individual consonant-vowel syllables correlates with the interval between air release and vocal cord vibration (e.g., "ba" versus "pa"; Lisker and Abramson, 1964), the duration of frequency transitions (e.g., "ba" versus "wa"; Liberman et al., 1956), and the silent time between consonants and vowels (e.g., "sa" versus "sta"; Dorman et al., 1979). Furthermore, prosodic cues such as pauses, the duration of speech segments, and speaking rate influence semantic content (Lehiste et al., 1976). Indeed, speech can still be understood even when the available cues are primarily temporal (Shannon et al., 1995), but not when the temporal cues are removed either by manipulations of the speech signal (Drullman et al., 1994a,b) or by impairments of temporal processing in the perceiver (Tallal and Piercy, 1973).Most of the temporal cues involved in speech perception fall on a time scale of tens to hundreds of milliseconds. Relatively little is known about the neural mechanisms und...