Adaptation in the processing of interaural time differences revealed by

Adaptation in the processing of interaural time differences revealed by the auditory localization aftereffecta) Makio Kashinob) and Shin’ya Nishida Information Science Research Laboratory, NTT Basic Research Laboratories, 3-1, Morinosato Wakamiya, Atsugi, Kanagawa 243-0198, Japan

~Received 25 February 1997; revised 10 September 1997; accepted 2 March 1998! Two experiments were conducted involving the auditory localization aftereffect, in which the perceptual lateralization of a test sound having an interaural time difference ~ITD! shifts away from that of a prior adapting sound having a different ITD. First, the frequency selectivity of the aftereffect was examined for sinusoids presented through headphones, with various combinations of adapter and test frequencies below 800 Hz, using the method of constant stimuli. The magnitude of the aftereffect was found to be largest when the frequencies of the two tones were similar, and virtually disappeared at a frequency difference of one-half octave. Second, the ITD selectivity of the aftereffect was examined for 400-Hz sinusoids. Subjects’ judgments of lateralization were measured directly in terms of the perceived azimuth of the test tone for various combinations of adapter and test ITDs in the range of 6625 m s. The magnitude of the aftereffect was found to be largest when adapter and test ITDs differed by approximately 250 ms. These results were successfully simulated by an interaural cross-correlation model having gain control. The results are consistent with the idea that the gain of ITD-selective units, located after binaural interaction but before across-frequency integration, is changed by recent input. © 1998 Acoustical Society of America. @S0001-4966~98!03806-5# PACS numbers: 43.66.Pn, 43.66.Qp, 43.66.Mk @RHD#

INTRODUCTION

This paper describes two experiments involving the auditory localization aftereffect, performed to investigate processing channels in the auditory system and their dynamic adaptation during auditory spatial processing. The aftereffect involves shifts in the apparent location of a test sound following prolonged exposure to an adapting sound. The aftereffect was initially reported in 1920’s, but has been the subject of only a few studies ~Flu¨gel, 1920–1921; Bartlett and Mark, 1922–1923; James, 1936; Be´ke´sy, 1960; Thurlow and Jack, 1973!. It has been shown that the displacement of the test sound is away from the adapter. It has also been shown that the localization aftereffect can be produced not only by an adapting sound that has only an interaural level difference ~ILD!, but also by an adapting sound that has only an interaural time difference ~ITD! ~Thurlow and Jack, 1973!. However, the nature of the localization aftereffect remains largely unclear, and no effort has been made to interpret the aftereffect in relation to the mechanisms of binaural processing. We revisited the localization aftereffect for two reasons. First, we expected that the localization aftereffect would provide useful tools to analyze perceptual processing channels and their interaction in auditory spatial processing. The usefulness of aftereffects as a psychophysical tool has been proven in the study of vision, where various aftereffects have been known for years. For instance, apparent shifts in spatial frequency or orientation provide convincing evidence for the existence of processing units ~or channels! in the early stages a!

Portions of this research were presented at the 131st meeting of the Acoustical Society of America. Electronic mail: [email protected]

b!

3597

J. Acoust. Soc. Am. 103 (6), June 1998

of visual processing, each selectively tuned to a certain range of spatial frequencies and orientations ~Blakemore and Sutton, 1969; Coltheart, 1971!. The selectivity of aftereffects, as well as that of visual masking, has been widely used to measure the tuning of such processing channels. In hearing, on the other hand, it is primarily masking that has been used to analyze processing channels, and only a few studies have measured aftereffects ~Kay and Matthews, 1972; Green and Kay, 1974; Regan and Tansley, 1979; Tansley and Suffield, 1983; Grantham, 1989; Shu et al., 1993!. Electrophysiological studies have shown that ITDs and ILDs are extracted separately in channels that are frequency selective at early levels of binaural interaction. Neurons selectively tuned to specific ITDs have been found in the brainstem nuclei of some animals such as the owl ~Carr and Konishi, 1990! and the cat ~Yin and Chan, 1990!. However, direct evidence for the existence of such mechanisms in the human auditory system has been lacking, due in part to the difficulty of physiological measurements. One of our purposes is to determine if the selectivity of the auditory localization aftereffect is consistent with the idea that there are ITD-selective units in the human auditory system. The second reason we revisited the localization aftereffect is that it represents dynamic adaptation in auditory spatial processing. Recently, there is increased interest in the perception of spatially dynamic sounds, as exemplified in the study of binaural sluggishness and the precedence effect ~for review, Grantham, 1995!. However, dynamic changes in the mechanisms responsible for spatial perception have been paid little attention, except for a series of studies by Hafter and his colleagues ~Hafter et al., 1988!. They found that, in the ITD detection task for click trains, information from

0001-4966/98/103(6)/3597/8/$10.00

© 1998 Acoustical Society of America

3597

later-arriving portions becomes progressively less effective when information is presented at high rates. At a glance, this may seem to be a loss of information, but Hafter et al. ~1988! pointed out that ‘‘binaural adaptation’’ may prevent the perceptual system from being overloaded by redundant directional information. Also in vision, we can find many examples in which adaptation promotes efficient information processing in the real world. For instance, light adaptation improves the effective resolution of brightness using sensors which have a limited dynamic range, by shifting the operation point according to the recent brightness distribution in the environment ~Werblin, 1973!. Barlow ~1990! argues that adaptation reduces redundancy in sensory messages and makes the system specially sensitive to new associations. We thought that dynamic adaptation in sound localization may also play an important role in efficient directional information processing, and the existence and nature of such adaptation could be revealed by the localization aftereffect. If correct, then models of auditory spatial processing should incorporate the mechanisms that underlie adaptation. To clarify the nature of the localization aftereffect, we examined how the magnitude and direction of the aftereffect depend on the frequency relationship and the ITD relationship between the adapter and test sound in experiments 1 and 2, respectively. To best focus on the mechanisms processing ITDs, we used sinusoids presented through headphones with frequencies below 800 Hz, where ITDs of ongoing waveforms act as a major determinant of the perceptual lateralization of tones. We found that the aftereffect shows strong frequency selectivity comparable to that of auditory filters, and that the aftereffect is largest when the two tones are separated by an ITD of around 250 ms. The psychophysical results were then successfully simulated using an interaural cross-correlation model with a newly introduced gain control function. These results are consistent with the idea that there are neural channels that have selectivity both in ITD and in frequency in the human auditory system, and that changes in the sensitivity of such channels produce the localization aftereffect.

the offset of each test tone!, the adapter tone was presented again for 5 s to reinforce the adaptation. This time pattern of adaptation ~60-s initial adaptation and 5-s adaptation between trials! was determined based on informal observations of conditions required to obtain a stable aftereffect throughout the session. The adapter frequency was either 200, 283, 400, 566, or 800 Hz, and the test frequency was either 283, 400, or 566 Hz. The frequency combination was fixed throughout a session. In no-adaptation sessions, the adapter tones were replaced with the same duration of silence. Twenty measurements were conducted for each data point per subject. The proportion of ‘‘right’’ responses was obtained as a function of test ITD for each frequency combination. To ensure that the aftereffect reflects binaural processing of ongoing ITDs, other cues that may affect sound source lateralization were removed. In the previous studies, ITDs of ongoing waveforms and ITDs of onsets and envelopes ~if available! were co-varied ~Bartlett and Mark, 1922–1923; James, 1936; Thurlow and Jack, 1973!. In the present experiment, a raised-cosine ramp ~0.05 s! was applied at every onset and offset of the stimuli. The ramps were synchronous across the left and right channels, eliminating onset and envelope ITD cues. The A-weighted sound pressure level at both ears was adjusted to 55 dB. The stimuli were synthesized digitally on a computer ~Macintosh Quadra 950!. The sampling rate was 48 kHz and quantization was 16 bit. The stimuli were converted to analog signals using a Digidesign Pro Tools Audio Interface, and presented through Sennheiser HDA 200 headphones in a sound-insulated booth. Four young-adult listeners participated, all of whom had quiet thresholds within 15 dB of the ANSI, 1969 standard at all audiometric frequencies. They had experience in psychoacoustic tasks other than sound lateralization judgments, and received several hours of training on sound lateralization judgments of sinusoids used in the experiment.

B. Results and discussions I. EXPERIMENT 1: FREQUENCY SELECTIVITY OF THE LOCALIZATION AFTEREFFECT A. Method

We measured the magnitude of the aftereffect with various combinations of adapter and test frequencies. Psychometric functions were obtained for left–right judgment of test using the method of constant stimuli, with and without adaptation. The shift of the psychometric function between the adaptation and no-adaptation conditions corresponds to the magnitude of the aftereffect. In each adaptation session, an initial adapter tone having an ITD ~either 2375 or 375 ms; Positive values indicate that the right-ear signal leads! was presented for 60 s, followed by a 0.25-s test tone. The ITD of the test tone was chosen randomly on each trial from the five values ~2250, 2125, 0, 125, or 250 ms!. Subjects’ task was to judge whether the test tone was perceived at the left or right in a two-alternative forced-choice paradigm. Following each test tone ~1 s after 3598

J. Acoust. Soc. Am., Vol. 103, No. 6, June 1998

Figure 1 shows the results when both adapter and test tones were 400 Hz. With no adapter tone, a subjective center ~50% response! was brought about by a test tone having an ITD near zero. With the left-leading ~right-leading! adapter, the psychometric function, estimated using the maximum likelihood procedure ~Bock and Jones, 1968!, shifted left ~right!. Although there were individual differences in the absolute magnitude of the aftereffect, the overall tendency was similar for all subjects. The shift of the psychometric functions at the 50% points was 108 ms in average for the leftleading and right-leading adapters across four subjects. Figure 2 shows the magnitude of the aftereffect ~shift of psychometric functions! as a function of the adapter and test frequencies. For all test frequencies and all subjects except one case ~subject: HO, test frequency: 566 Hz!, the magnitude of the localization aftereffect was found to be largest when the frequencies of the adapter and test tones were identical. It decreased as the frequency difference increased, and virtually disappeared at a difference of one-half octave. M. Kashino and S. Nishida: Auditory localization aftereffect

3598

FIG. 3. Mean lateralization judgments as a function of the test ITD for a trained subject ~MK! when no adapter was presented and when the adapter ITD was 0. Error bars indicate the standard error of the mean. The data for the other four subjects were similar. FIG. 1. Proportion of ‘‘right’’ responses as a function of test ITD when both the adapter and test sounds were 400-Hz tones ~the adapter ITD was 2375 or 375 ms!, and when only the 400-Hz test tone was presented as a noadapter control condition. Individual data and the average of the four subjects are shown. Dots indicate the obtained data, error bars indicate the standard error of the mean, and curved lines indicate the psychometric functions estimated using the maximum likelihood procedure.

II. EXPERIMENT 2: ITD SELECTIVITY OF THE LOCALIZATION AFTEREFFECT A. Method

We measured the magnitude of the localization aftereffect for 400-Hz sinusoids using various combinations of adapter and test ITDs. Subjects indicated the azimuth of the intracranial image of the test tone on a half-circle arc for various combinations of adapter and test ITDs in the range of

6625 m s and also for a no-adapter control condition. They were instructed to ignore the distance and elevation of the perceived image. The azimuths of the marked points from the center were measured ~positive value: right of center, negative value: left of center!. The difference between each adapter condition and the no-adapter condition represents the magnitude of the aftereffect. In each adaptation session, an initial adapter tone having an ITD ~either 2375, 2125, 0, 125, or 375 ms! was presented for 60 s, followed by a 0.252s test tone. The ITD of the test tone was chosen randomly on each trial from the the eleven values ~2625, 2500, 2375, 2250, 2125, 0, 125, 250, 375, 500, or 625 ms!. Following each test tone ~4 s after the offset of each test tone!, the adapter tone having the same ITD as the initial adapter tone was presented again for 5 s to reinforce the adaptation. Subjects made their decision during the silent interval between the test tone and the next adapter tone. Twenty measurements were conducted for each data point per subject. In no-adaptation sessions, the adapter tones were replaced with the same duration of silence. Three subjects ~HA, NK, and HO! who had participated in experiment 1 and two new subjects ~MK and HM! participated in experiment 2. MK was the first author. HM was similar in experimental experience and hearing sensitivity to the three subjects who had participated in experiment 1. Three of the five subjects ~HA, HM, and HO! participated in the no-adaptation condition and only one adaptation condition ~adapter ITD 50!. The other two subjects ~MK and NK! completed all the conditions. The apparatus of the experiment was the same as in experiment 1. B. Results and discussions

FIG. 2. Magnitude of the localization aftereffect as a function of the adapter and test frequencies. Individual data and the average of the four subjects are shown. Error bars indicate the standard error of the mean. 3599


Figure 3 shows the lateralization judgments of a subject ~MK! when no adapter tone was presented and when the adapter ITD was zero. The differences between the two conditions are shown in Fig. 4, along with the other four subjects’ data. There are relatively large individual differences in the absolute size of the aftereffect, but we can still see a M. Kashino and S. Nishida: Auditory localization aftereffect

3599

FIG. 4. Mean lateralization judgments ~adapter ITD50! relative to the noadapter condition as a function of the test ITD for five subjects. When the test ITD is larger than zero, a positive value of the relative lateralization judgment indicates that the test tone shifted away from the adapter tone ~the first quadrant!. When the test ITD is smaller than zero, a negative value indicates the same repulsion effect ~the third quadrant!.

common tendency across subjects. First, the magnitude of the aftereffect was rather small when the ITDs of the adapter and test tones were similar, and was largest when ITDs differed by around 250 ms. Second, most of the data points fell in the first or the third quadrant, indicating that the displacement of the test tone was away from the perceived location of the adapter tone. In other words, the aftereffect enhances the spatial separation between the adapter and test tones. The same tendency was also observed in conditions in which the adapter ITDs were not zero ~Fig. 5!. The magnitude of the aftereffect was largest when ITDs of the adapter and test tones differed by approximately 250 ms. These results support the earlier findings concerning the direction of the localization aftereffect ~Bartlett and Mark, 1922–1923; James, 1936; Thurlow and Jack, 1973!. III. DISCUSSION A. Summary of the results

~1! The localization aftereffect occurs for low-frequency tones that are lateralized based only on ITDs of ongoing waveforms. ~2! Following adaptation, the apparent lateralization of the test tone shifts away from that of the adapter tone, no matter where the adapter tone is lateralized. ~3! The magnitude of the localization aftereffect is, under optimal conditions, about 20 degrees in average subjects. ~4! The localization aftereffect is frequency selective, that is, the magnitude of the aftereffect is largest when the frequencies of the adapter and test tones are identical, and disappears at a frequency difference of one-half octave. ~5! The localization aftereffect is ITD-selective, that is, the magnitude of the aftereffect is smallest when the ITDs of 3600


FIG. 5. Mean lateralization judgments ~adapter ITD52375, 2125, 125, or 375 ms! relative to the no-adapter condition as a function of the test ITD for two subjects ~MK and NK!. When the test ITD is larger than the adapter ITD, a positive value of the relative lateralization judgment indicates that the test tone shifted away from the adapter tone. When the test ITD is smaller than the adapter ITD, a negative value indicates the same repulsive effect.

the adapter and test tones are similar, and is largest when they differ by around 250 ms. B. Underlying mechanism of the localization aftereffect

The current results indicate that the localization aftereffect based only on ITDs is selective both in ITD and in frequency. This suggests that processing units having selectivity both in ITD and in frequency are responsible for the aftereffect. To confirm this idea, we created a model of ITD processing which incorporates adaptation, and compared the results of our computer simulation with the psychophysical data. Our model is based on the widely accepted idea that binaural interaction contains a process to compute the interaural cross correlation separately in each frequency region after processing by the peripheral auditory system ~for review, see Stern and Trahiotis, 1995; Colburn, 1995!. In such models, it is assumed that the interaural cross correlation is computed by a delay-coincidence network, originally proposed by Jeffress ~1948!. Each coincidence detector records coincidences of neural impulses from the two ears after a series of internal time delays. Thus the coincidence detectors can be thought of as ITD channels, each responding selectively to a certain range of ITDs. The short-term average of the set of coincidence outputs, as a function of their internal delay, approximates a short-term cross-correlation function of the neural signals arriving at the coincidence detectors. The cross-correlation functions from different frequency bands are then integrated into a summary cross correlation. M. Kashino and S. Nishida: Auditory localization aftereffect

3600

Subjective lateral position is assumed to then correspond either to the location of peaks of the summary crosscorrelation function or to the location of a centroid along the internal delay axis ~Stern and Colburn, 1978; Lindemann, 1986; Shackleton et al., 1992!. In vision, adaptation is often thought of as a temporary change in the sensitivity of a particular population of single neurons after prolonged stimulation. The sensitivity change is assumed to occur either by fatigue of the activated neurons ~Sutherland, 1961; Coltheart, 1971! or by continued lateral inhibition from adjacent neurons ~Blakemore et al., 1970; Barlow, 1990!. Here we assumed that such a sensitivity change occurs in the coincidence detectors according to the magnitude of previous activation. We implemented the sensitivity change by using independent multiplicative gain controls affecting each frequency-selective coincidence detector. In our model, the sounds reaching the two ears were first bandpass filtered by the ERB filter bank containing 78 frequency channels equally spaced on an ERB-rate scale between 50 and 3000 Hz ~Moore and Glasberg, 1986!. The outputs of the filter bank were then half-wave rectified to mimic the unidirectional response properties of the hair cells. The rectified signals from the left and right ears were then fed into the delay-coincidence network. In the simulation, the adapter and test signals used as input to the model were sampled from the 50-ms portion at the midpoint of the stationary sinusoidal signals used in experiments 1 and 2. Therefore, the shape of the temporal weighting function used to compute the running interaural cross correlation is not crucial here. The coincidence count with internal delay t at frequency f during the time period between t 1 and t 2 is t2

f ~ f , t ! 5 ( X l ~ f ,t ! X r ~ f ,t2 t ! , t5t 1

~1!

where X l ( f ,t) and X r ( f ,t) are the rectified outputs from the left and right ears, respectively, and t 2 2t 1 was set to 50 ms. The variables t and t took discrete values, with a step size of 1/96 000 s. The range of t was between 21 ms and 1 ms, which covered the possible range of ITDs for human listeners. The coincidence count was normalized into the range between 0 and 1, by dividing f ( f , t ) by the maximum coincidence count for the sound in the entire f 2 t plane. The output from the coincidence detector is ¯ ~ f ,t !, c ~ f , t ! 5g ~ f , t ! • f

~2!

¯ ( f , t ) is a normalized where g( f , t ) is a gain function and f coincidence count. The gain, which represents the effect of adaptation, should decrease according to the magnitude of the previous input, and keep changing during adaptation until it reaches an asymptote. However, we do not know the exact temporal course of adaptation yet. Therefore, we assumed that after long adaptation, the gain function becomes g~ f ,t !5

1 , ¯ 11k• f adapt~ f , t !

~3!

where k is a positive constant to control the amount of gain ¯ control, and f adapt( f , t ) is the normalized coincidence count for the adapter at frequency f and interaural delay t. There is 3601


FIG. 6. Behavior of the gain control model in a frequency channel. Before adaptation, the gain function g( f , t ) is unity. As a consequence, the outputs of the coincidence detectors, c ( f , t ), is equal to the cross correlation of the ¯ ( f , t ). The peak of c ( f , t ) for a test tone having signal from the two ears, f an ITD of 375 ms is located at the internal delay of 375 ms. After long exposure to an adapter tone having zero ITD, the gain g( f , t ) changes ~dotted line!, resulting in the shift of the peak location of the coincidence outputs, c ( f , t ). The shift of the peak corresponds to the localization aftereffect.

only one free parameter (k) in this model. Figure 6 illustrates how this model works in each frequency band. Before adaptation ~upper panel!, the outputs from the coincidence detec¯ ( f , t ), since the gain tors c ( f , t ) are equal to the inputs f g( f , t )51. The peak of the coincidence outputs c ( f , t ) for the 400-Hz test tone having an ITD of 375 ms is located at the internal delay of 375 ms. After long exposure to an adapter tone having zero ITD, the gain g( f , t ) changes as is represented by a dotted line, making the peak of the coincidence output c ( f , t ) shift away from the adapter ITD. The perceived lateral position is then determined by the location of the peak in the summary cross-correlation function 78

s~ t !5

( c@ f ~ i !,t #,

i51

~4!

where f (i)’s are equally spaced on an ERB-rate scale between 50 and 3000 Hz ~Shackleton et al., 1992!. The shift of this peak corresponds to the localization aftereffect. Figure 7 represents the multi-channel outputs of the coincidence detectors and the summary cross correlation. The brightness of each cell corresponds to the magnitude of the value at the cell. The top-left panel shows the coincidence outputs for the 400-Hz adapter tone with an ITD of 2375 m s, and the topright panel shows the gain function for the adapter tone. With this adaptation, the activity pattern of the coincidence detectors for the 400-Hz test tone shifts rightward ~middleleft: before adaptation, bottom-left: after adaptation!. On the other hand, the coincidence activities for the 566-Hz test M. Kashino and S. Nishida: Auditory localization aftereffect

3601

FIG. 7. Top-left: The multi-channel activities of the coincidence detectors c ( f , t ) and the summary cross correlation s( t ) for a 400-Hz adapter tone with an ITD of 2375 m s. Top-right: The gain function g( f , t ) for the adapter tone. Middle-left: c ( f , t ) and s( t ) for a 400-Hz test tone with an ITD of 0 ms before adaptation, bottom-left: c ( f , t ) and s( t ) for a 400-Hz test tone with an ITD of 0 ms after adaptation. Middle-right: c ( f , t ) and s( t ) for a 566-Hz test tone with an ITD of 0 ms before adaptation, bottom-right: c ( f , t ) and s( t ) for a 566-Hz test tone with an ITD of 0 ms after adaptation. The brightness of each cell corresponds to the magnitude of the value at the cell.

tone stay the same ~middle-right: before adaptation, bottomright: after adaptation!. We then tested how well the model predicts the aftereffect observed in experiment 2. We adjusted the parameter k to fit subject MK’s data when the adapter ITD was zero, minimizing the squared deviations between the model predictions and the lateralization judgments. To convert ITD to lateralization azimuth, we used an equation Lateralization~deg!5ITD~ms!* 0.096,

~5!

obtained from the linear regression for MK’s lateralization judgment data in the no-adapter condition ~Fig. 3, closed circles!. The top panel of Fig. 8 shows the predicted aftereffect ~solid line! when k51.05 and the experimental data ~closed circles!. Next, we used the same k value to predict the other four adapter conditions ~Fig. 8, middle and lower panels!. The predicted aftereffect captures three important features of the experimental data for all the adapter ITD conditions. First, the aftereffect is close to zero at the adapter ITD. Second, the displacement of the test tone is away from the adapter tone. Third, the aftereffect is largest when they differ by, in the case of this subject, approximately 300 ms. The parameter k does not change the overall shape of the predicted aftereffect; it changes the magnitude of the aftereffect and the location of the peak. For subjects who show a larger aftereffect ~such as NK!, a larger k produces a better fit. Next, we tested if the model predicts the frequency selectivity of the aftereffect observed in experiment 1. We used the same k as the previous simulation, and ITDs of the adapter and test tones were set to 2375 m s and 0 ms, respectively. The magnitude of the aftereffect was expressed in microseconds rather than degrees. The simulation represents 3602


essential features of the data from experiment 1: The predicted aftereffect disappeared at a frequency difference of one-half octave, and was smaller when the test frequency was 566 Hz ~Fig. 9!. This is apparently due to the independent operation of gain control in each frequency band, as is depicted in Fig. 7, rather than operating after acrossfrequency integration. In the current simulation, we assumed that the delaycoincidence network contains many coincidence detectors

FIG. 8. The predicted localization aftereffect ~solid lines! and subject MK’s data ~closed circles! as a function of test ITD in five adapter ITD conditions. The vertical dotted lines represent the adapter ITDs. For details, see text. M. Kashino and S. Nishida: Auditory localization aftereffect

3602

as a group of cross correlaters. Moreover, the frequency selectivity of the localization aftereffect observed in experiment 1 indicates that such ITD-selective units operate independently in different frequency bands. The frequency selectivity of the ITD-selective units appears comparable to that of the auditory filters. It is premature to determine the specific neural site where the adaptation of ITD-selective units takes place in the human auditory system. For now, we can say that the adaptation occurs after binaural interaction, but before acrossfrequency integration. FIG. 9. The predicted localization aftereffect ~solid lines! and subject MK’s data ~closed circles! as a function of the adapter and test frequencies. For details, see text.

with very small (1/96 000 s) steps, and that the apparent sound location was determined by the peak location of the coincidence outputs. However, these assumptions are not necessary to predict the localization aftereffect. For example, the selectivity of the aftereffect would be predicted based on the activity ratio ~or centroid! of the coincidence detectors. In this case, a relatively small number of coincidence detectors are sufficient to predict the psychophysical data, as is the case for color vision ~Boynton, 1979!. For now, it is not possible to determine which strategy is used in sound lateralization. Although this model can predict the localization aftereffect fairly successfully with only one free parameter, there are several limitations. First, in the simulation of ITD selectivity, the discrepancy between the prediction and the experimental data tends to get larger at the edges of the ITD axis. Second, the model cannot capture the left–right asymmetry some subjects showed. Third, and more importantly, the current model does not incorporate the temporal course of gain change. More empirical data are required to modify the model concerning these points. In sum, both the psychophysical and computational results support the idea that the gain of ITD-selective units, which also have frequency selectivity, changes according to recent input, producing the localization aftereffect. C. Physiological basis of the localization aftereffect

There are anatomical and physiological data supporting the idea that ITDs are extracted by a delay-coincidence network separately in each frequency region, and information from different frequency regions is integrated into a single spatial representation at later stages. It has been shown that the principal cells of the medial superior olive ~MSO! perform coincidence detection between phase-locked inputs from the two ears in the cat and in the barn owl ~Yin and Chan, 1990; Carr and Konishi, 1990!. In the barn owl, the across-frequency integration of ITD information is shown to take place at the level of the external nucleus of the inferior colliculus ~ICX!. As for human listeners, there has been no direct physiological data concerning the mechanisms of ITD detection. Our current data provides evidence that the human auditory system contains ITD-selective units, which can be modeled 3603


D. Functional roles of the adaptation in ITD processing

An important task of hearing is to detect changes or differences in sound localization, which may indicate the emergence of a new acoustic event. The adaptive processing revealed by the localization aftereffect may promote the detection of changes or differences around the adapter ITD. This possibility is supported by Kashino ~1998!, who shows that ITD discrimination is selectively improved following adaptation if the adapter and test sounds are similar in the ITD domain and in the frequency domain. The adaptation can be thought of as a recalibration of the mechanism extracting ITDs in order to represent a small change in the recently experienced sound with maximum resolution. ACKNOWLEDGMENTS

We thank Richard M. Warren and Willard R. Thurlow for directing our attention to relevant early studies and sending copies of those papers. We also thank Eric W. Healy, Tatsuya Hirahara, and Hisashi Uematsu for helpful discussion, and Ken’ichiro Ishii for supporting this research. Barlow, H. B. ~1990!. ‘‘A theory about the functional role and synaptic mechanisms of visual aftereffects,’’ in Vision: Coding and Efficiency, edited by C. Blakemore ~Cambridge U. P., Cambridge, England!, pp. 363– 375. Bartlett, F. C., and Mark, H. ~1922–1923!. ‘‘A note on local fatigue in the auditory system,’’ Br. J. Psychol. 13, 215–218. von Be´ke´sy, G. ~1960!. Experiments in Hearing ~McGraw-Hill, New York!. Blakemore, C., Carpenter, R. H. S., and Georgeson, M. A. ~1970!. ‘‘Lateral inhibition between orientation detectors in the human visual system,’’ Nature ~London! 228, 37–39. Blakemore, C., and Sutton, P. ~1969!. ‘‘Size adaptation: a new after-effect,’’ Science 166, 245–247. Bock, R. D., and Jones, L. W. ~1968!. The Measurement of Prediction of Judgment and Choice ~Holden Day, San Francisco!. Boynton, R. M. ~1979!. Human Color Vision ~Holt, Rinehart, and Winston, New York!. Carr, C. E., and Konishi, M. ~1990!. ‘‘A circuit for detection of interaural time differences in the brainstem of the barn owl,’’ J. Neurosci. 10, 3227– 3246. Colburn, H. S. ~1995!. ‘‘Computational models of binaural processing,’’ in Auditory Computation, edited by H. Hawkins and T. McMullin ~SpringerVerlag, New York!, pp. 332–400. Coltheart, M. ~1971!. ‘‘Visual feature analyzers and aftereffects of tilt and curvature,’’ Psychol. Rev. 78, 114–121. Flugel, J. C. ~1920–1921!. ‘‘On local fatigue in the auditory system,’’ Br. J. Psychol. 11, 105–134. Grantham, D. W. ~1995!. ‘‘Spatial hearing and related phenomena,’’ in Hearing, edited by B. C. J. Moore ~Academic, New York!, pp. 297–345. Grantham, D. W. ~1989!. ‘‘Motion aftereffects with horizontally moving sound sources in the free field,’’ Percept. Psychophys. 45, 129–136. M. Kashino and S. Nishida: Auditory localization aftereffect

3603

Green, G. G. R., and Kay, R. H. ~1974!. ‘‘Channels in the human auditory system concerned with the wave form of the modulation present in amplitude- and frequency-modulated tones,’’ J. Physiol. ~London! 241, 29P–30P. Hafter, E. R., Buell, T. N., and Richards, V. M. ~1988!. ‘‘Onset-coding in lateralization: Its form, site, and function,’’ in Auditory Function: Neurological Bases of Hearing, edited by G. M. Edelman, W. E. Gall, and W. M. Cowan ~Wiley, New York!, pp. 647–676. James, H. E. O. ~1936!. The Localisation of Sound ~Medical Research Council, Special Report Series No. 207, H. M. Stationery Office, London!. Jeffress, L. A. ~1948!. ‘‘A place theory of sound localization,’’ J. Comp. Physiol. Psychol. 41, 35–39. Kashino, M. ~1998!. ‘‘Adaptation in sound localisation revealed by auditory after-effects,’’ in Psychological and Physiological Advances in Hearing, edited by A. R. Palmer, A. Rees, Q. Summerfield, and R. Meddis ~Whurr, London!, pp. 322–328. Kay, R. H., and Matthews, D. R. ~1972!. ‘‘On the existence in human auditory pathways of channels selectively tuned to the modulation present in frequency modulated tones,’’ J. Physiol. ~London! 225, 657–677. Lindemann, W. ~1986!. ‘‘Extension of a binaural cross-correlation model by contralateral inhibition. I. Simulation of lateralization for stationary signals,’’ J. Acoust. Soc. Am. 80, 1608–1622. Moore, B. C. J., and Glasberg, B. R. ~1986!. ‘‘The role of frequency selectivity in the perception of loudness, pitch and time,’’ in Frequency Selectivity in Hearing, edited by B. C. J. Moore ~Academic, London!, pp. 251–308.

3604


Regan, D., and Tansley, B. W. ~1979!. ‘‘Selective adaptation to FM tones: Evidence for an information-processing channel selectively sensitive to frequency changes,’’ J. Acoust. Soc. Am. 65, 1249–1257. Shackleton, T. M., Meddis, R., and Hewitt, M. J. ~1992!. ‘‘Across frequency integration in a model of lateralization,’’ J. Acoust. Soc. Am. 91, 2276– 2279~L!. Shu, Z. J., Swindale, N. V., and Cynader, M. S. ~1993!. ‘‘Spectral motion produces an auditory after-effect,’’ Nature ~London! 364, 721–723. Stern, R. M., and Colburn, H. S. ~1978!. ‘‘Theory of binaural interaction based on auditory-nerve data. IV. A model for subjective lateral position,’’ J. Acoust. Soc. Am. 64, 127–140. Stern, R. M., and Trahiotis, C. ~1995!. ‘‘Models of binaural interaction,’’ in Hearing, edited by B. C. J. Moore ~Academic, New York!, pp. 347–386. Sutherland, N. S. ~1961!. ‘‘Figural after-effects and apparent size,’’ Q. J. Exp. Psychol. 13, 222–228. Tansley, B. W., and Suffield, J. B. ~1983!. ‘‘Time course of adaptation and recovery of channels selectively sensitive to frequency and amplitude modulation,’’ J. Acoust. Soc. Am. 74, 765–775. Thurlow, W. R., and Jack, C. E. ~1973!. ‘‘Some determinants of localization-adaptation effects for successive auditory stimuli,’’ J. Acoust. Soc. Am. 53, 1573–1577. Werblin, F. S. ~1973!. ‘‘The control of sensitivity in the retina,’’ Sci. Am. 228, 70–79. Yin, T. C. T., and Chan, J. C. K. ~1990!. ‘‘Interaural time sensitivity in medial superior olive of cat,’’ J. Neurophysiol. 65, 465–488.

M. Kashino and S. Nishida: Auditory localization aftereffect

3604