subject 2) completed all of the psychophysical studies reported here. The results from two other subjects (one male, one female) that completed part of these experiments are consistent with the other three. All experiments were performed in a darkened sound- and echo-attenuated sound booth (inner dimensions: 6.5′×8.5′, Industrial Acoustics, Bronx, NY; lined with 3″ Sonex foam) with the same acoustic apparatus, stimulus delivery, and data acquisition systems as described in detail elsewhere (34). Briefly, subjects faced the center of a 15-speaker array spanning ±28° in 4° increments along the horizontal meridian. Acoustic stimuli were generated by using a Tucker-Davis Technologies system, and all parameters of the experiment were controlled by a personal computer. Acoustic stimuli consisted of 200-msec (5-msec rise/fall) 750 Hz tones and 3,000 Hz tones. Stimulus intensity (45 dB SPL) was varied over a 4 dB range for every stimulus presentation to reduce detectable differences in the speaker transformation functions. Visual stimuli consisted of dim red or green LEDs located 1° above each of the 15 speakers.

Fig. 1A shows the experimental strategy for documenting the ventriloquism aftereffect. First, the ability of each subject to localize these stimuli in absolute space was measured. Second, a period of training with an eight-degree disparity between the auditory and visual stimuli was presented for approximately 20–30 min. Third, the absolute localization ability was measured again. The only visual stimuli presented during the course of the experiment were from the LEDs, and the only other acoustic stimuli were from brief instructions from the experimenter presented from behind the subject between the first and second parts of the experiment.

To define the absolute localization ability (Fig. 1B), the subject’s head position was measured by using a headband mounted to a variable potentiometer that measured the head orientation. Subjects were in complete darkness except for a dim-fixation LED located at the central position (12° up) that would blink on/off until the subjects oriented their head directly toward zero degrees in azimuth and elevation (±1°). The fixation light was then extinguished, and 500–1500 msec later a tone or noise was presented from 1 of the 15 speakers of the array. The subjects turned their head to face the perceived location of the stimulus. Single trial estimates were taken as the head position measured 1,950–2,000 msec after the onset of the tone stimulus. This time period was well after minor adjustments by the subjects and before the subjects’ returned their head to the center position. Each session consisted of 15 trials at each of 9 locations, which typically lasted 15–20 min.

After collection of these baseline data, subjects were exposed to a 20- to 30-min “training” period (Fig. 1C). Training stimuli consisted of paired 200-msec duration light stimuli and 200-msec acoustic stimuli identical to the stimuli used to collect the baseline data. The light and the sound were always at the same relative spatial locations, either the same (0°) or with the light offset by 8° to the right of the sound (+8). These stimuli were presented at a rate of two per second in sets of five stimuli from the same location, then a different location was randomly selected. To ensure that the subjects attended to these stimuli, either the third, fourth, or fifth stimulus of the set of five for any given location was randomly presented at 35 dB SPL (10 dB quieter) on approximately one of every five locations. The subjects pressed a button to initiate the stimulus presentations and released the button when they detected a change in the intensity of the stimulus. The instructions given to the subjects were to attend to the intensity of the acoustic stimuli and to release the button when the stimulus intensity decreased.

The training period presented 2,500 stimuli, corresponding to approximately 20–30 min. Immediately after this training period, the subjects again performed the same task as for the baseline data, in which they oriented their head toward the

FIG. 1. Psychophysical paradigms used to demonstrate the ventriloquism aftereffect. (A) The experimental strategy. Subjects first had their absolute localization ability measured, then were exposed for a period of 20–30 min to paired visual and auditory stimuli, then had their absolute localization ability remeasured. (B) The absolute localization paradigm used pre- and posttraining for each subject. (C) Training paradigm for the ventriloquism aftereffect. The 15-speaker array with a corresponding LED for each speaker is schematically illustrated. The subject pressed a button to initiate a trial. Throughout the course of the session a light and sound were simultaneously presented in a consistent spatial relationship. In this example the light is located two positions (8°) to the right of the sound. The subject was asked to attend to the intensity of the acoustic stimuli and to release the button on detection of a decrease in intensity.

same acoustic stimulus in complete darkness. If the exposure to a mismatch in the acoustic and visual stimuli was effective in driving representational changes, there should be a shift in the localization estimates in the session immediately posttrain-