Permissions for Reuse

The effect of attention on adaptation produced by amplitude modulation (AM) was examined. In different experimental conditions, listeners' AM detection thresholds for a 2 kHz test tone were measured after exposing them to an adapting sound that was presented simultaneously with speech distractors. Magnitude of an aftereffect, calculated as the elevation of the thresholds caused by adaptation, was smaller when the listeners shift attention away from the adaptor to the distractor voice than when they attended to the adaptor. The results suggest that the AM of unattended sounds may not be fully analyzed compared to that of attended sounds. ©2009 Acoustical Society of America

Contents

• Introduction
• Method
  • A.Participant
  • B.Stimuli and equipment
  • C.Procedure
• Results and discussion
• Acknowledgments
• REFERENCES
• FIGURES

Almost all sounds in natural environments, including our voices, have an amplitude envelope that varies over time. This time-dependent variation [amplitude modulation (AM)] plays important roles in auditory perception, as is seen in speech perception (e.g., Shannon et al., 1995) and in perceptual grouping of frequency components (Darwin and Carlyon, 1995). To understand the processing of amplitude envelopes, several researchers have examined characteristics of perceptual adaptation produced by AM. For example, Kay and Matthews (1972) reported that the AM detection threshold increased following listeners' exposure to a long, modulated tone (i.e., adaptor). Wojtczak and Viemeister (2003) used a matching procedure to show that suprathreshold perception of modulated tones was also changed by adaptation. Moreover, to some extent, such aftereffects of AM exposure appear to be selectively linked to both the rate of modulation and the carrier frequency of the adaptor sound (Kashino, 1998; Richards et al., 1997; Wojtczak and Viemeister, 2003). To some, such findings suggest that aftereffects of AM adaptation arise from specific auditory processing of AM (e.g., Wojtczak and Viemeister, 2003).

An important problem related to adaptation by AM concerns the involvement of attention. Does auditory processing show constant adaptation irrespective of the direction of the listener's attention? In vision, several studies have reported that adaptation is indeed affected by attention (e.g., Yeh et al., 1996). For example, Chaudhuri (1990) examined the effect of attention on the motion aftereffect elicited by a moving array of random dots. The duration of the aftereffect was shorter when the observers shifted attention from the dots to alphanumeric symbols superimposed on the dots, compared to when they attended to the moving dots.

In the present report, the author examined possible effects of attention on adaptation by AM. Following research in visual perception, the author predicted that AM adaptation should be less pronounced when the listener's attention is directed to sounds other than the adaptor sounds during an adapting phase than when attention is focused directly on the adaptors.

Nine observers, two males and seven females, with no reported history of hearing problem, participated in the experiment. They were university students and paid for the participation.

Sound stimuli were generated on a personal computer (Apple, Macintosh G4) equipped with a sound card and digital to analog converter. They were presented to individual participants over headphones (Sennheiser, HDA200). All stimuli were digitized in 16 bits and were played at 44.1-kHz sample frequency. Measurements were conducted in a sound-proof room.

A sinusoidally amplitude-modulated (SAM) tone with a carrier frequency of 2 kHz was served in both the test stimuli and the adaptor. Two types of AM test stimuli were created based, respectively, on two different modulation frequencies: 16 and 51 Hz. Durations of the test (SAM) tones were always 0.5 s. The modulation frequency of the adaptor was 16 Hz; its depth of modulation was −3 dB. To prepare for the manipulation of the listener's attention, the level of the adaptor was occasionally changed during presentation. The fluctuation of sound level lasted for 1 s (during this interval, the sound was damped linearly for 0.4 s, then kept at minimum sound level of −8 dB for 0.2 s, and was ramped linearly for 0.4 s to the original level). The time interval between fluctuations was randomly chosen from three candidates: 1.4, 2.4, and 3.4 s. The adaptor and the test stimuli were always presented through the left channel.

Three types of distractor sounds were used: female voices, male voices, and 5313-Hz tone. Distractors were presented simultaneously with an adaptor sound. The voice distractors were always presented at the opposite ear from the adaptor. The female voice distractor consisted of utterances of numbers (1–10 in Japanese, and the average duration was 0.32 s); ordering of numbers was randomized with replacement. The utterances were separated by silent intervals whose duration varied randomly among three values: 0.50, 0.60, and 0.65 s. The male voice distractor was composed of sentences in a speech corpus (Speech Database Committee of Acoustical Society of Japan, 1990); this continuous distractor was included to render a focus on numbers (in the female voice) more difficult. Both voices were filtered with a band-stop filter to reduce their energy around 2 kHz. Unlike voice distractors, the tone distractor was presented through the same channel as the adaptor. The rationale for adding a tone distractor was to lower adaptor salience. It has been reported that listeners have difficulty in attending to one of the two sounds simultaneously presented at one ear, when other sound was presented at the opposite ear (Kidd et al., 2003). All stimuli were presented at a sound pressure level of 61 dB, although each voice distractor was presented at 60 dB. All stimuli were gated on and off with 0.02 s cosinusoidal amplitude ramps.

Thresholds for AM detection were measured using a two-interval forced choice procedure. Equally often a SAM tone occurred in one of two 0.5 s intervals; a silent 0.3 s interval separated two observation intervals. Participants reported which interval contained a SAM tone. Feedback was not presented (e.g., Richards et al., 1997). In conditions where an adaptor was presented, a long adaptor (80 s duration) was presented at the start of a block of trials. In addition, shorter adaptors (40 s durations) were presented between trials when 25 s have passed since the last presentation of the adaptors.

In addition to a no-adaptor control condition, three adapting conditions were examined. In the adaptor condition, adapting stimuli were presented accompanied only by the tone distractor. In two different adaptor-plus-voices conditions, the adaptor and all three types of the distractors were simultaneously presented; these two conditions were acoustically identical; however, they differed in attentional set and the listeners were imposed on attentional tasks during adaptor presentation (in addition to AM detection). One of the two conditions, the attend-to-adaptor condition, required participants to report the level fluctuations of the adaptor by pressing a key during the adaptor presentation. The other condition was the attend-to-distractor condition. In this condition, participants were required to judge (by pressing one of two keys) whether an utterance of the female voice expressed a number that was even or odd.

All participants received the control condition and the adaptor condition. However, they received only one of the two adaptor-plus-voices conditions; five were assigned to the attend-to-adaptor condition and four were assigned to the attend-to-distractor condition. Since both these groups received the same adapting stimuli but different attentional instructions, the design of the experiment was a 2×3×2 mixed factorial design in which two levels of test modulation rates and three levels of the adapting settings (no-adaptor, adaptor with a tone distractor, and adaptor with three distractors) were crossed with two levels of the instructions.

A trial block contained 96 trials (2 modulation rates×6  depths  of  modulation×8 iterations); a block typically lasted 10 min in conditions that presented the adaptor. Each participant completed eight blocks for each condition (in the attend-to-distractor condition, one participant with time constraints received only seven blocks). A single session consisted of two blocks of the control plus (following) four blocks of either the adaptor condition or the adaptor-plus-voices condition. Different adapting conditions (i.e., the adaptor and the adaptor-plus-voices conditions) were alternately tested in separate days, their start conditions being counterbalanced across participants, following a first session that contained practice blocks.

Psychometric functions for each of the two attention groups were plotted using the average data points [Fig. 1(a)]. In both groups, the functions in the 16 Hz modulation condition (upper row) shifted to the right when the adaptor was presented (circles) compared to the no-adaptor control (triangles). The shift indexes the aftereffect due to AM adaptation.

Figure 1.

When the distractor voices were presented, reduction in the performance by the adaptation was softened at several tested modulation depths in the attend-to-distractor group (squares at upper right panel). To compare the effect of the adaptor across conditions, the magnitude of the aftereffect was calculated for each participant as a threshold difference between an adapted condition and that in the control condition (the threshold used corresponds to the modulation depth associated with 75% correct responding, specified by the cumulative normal functions fitted using least squares). The average magnitude in the 16 Hz test condition, plotted in Fig. 1(b), shows effects of attention condition and adaptor. When distractor voices were not presented, the magnitude of an AM aftereffect was almost constant in both the attend-to-distractor group and the attend-to-adaptor group (two bars at the right side). However, when distractor voices were presented, the magnitude was less pronounced in the attend-to-distractor group than in the attend-to-adaptor group (two bars at the left side). A 3×2 analysis of variance (ANOVA) using the threshold values (for the 16 Hz condition, the results of two factor analysis, not three, were reported for description brevity) revealed a significant interaction between the adaptor conditions and attention instructions, F(1,7)=6.47, p<0.01. Multiple comparisons, using Tukey's honestly significant differences tests, indicated that the thresholds in the adaptor condition increased significantly relative to those in the no-adaptor control condition (p<0.01). This increment was not modulated by the presences of distractor voices in the attend-to-adaptor group (p>0.1); however, the threshold decreased in the attend-to-distractor group (p<0.01). These results indicate that adaptation becomes less pronounced when listeners shift attention away from the adaptor.

When the modulation frequency of the adaptor (16 Hz) differed from that of the test stimuli (51 Hz), the proportion of correct responses did not largely differ across conditions [Fig. 1(a), bottom rows]. As before, a 3×2 ANOVA using the thresholds was conducted. No statistically significant effects of the adaptation variable, the attention variable, and their interaction were found (p>0.1 in all cases). The results are in line with the above-mentioned conjecture that the change in the magnitude of the aftereffect in 16 Hz condition is due to the shift of attention, not to the general influence of the distractor task.

Because the number of presentations of adapting stimulus varied over blocks depending on the responding pace of each participant, it is possible that results of Fig. 1 were influenced by this variable. However, we found no systematic differences in the number of the adaptor presentations between the attend-to-adaptor group and the attend-to-distractor group (8.4 and 8.7 on average, in the adaptor-plus-voices condition, respectively). This indicates that the number of the adaptor presentation cannot explain the difference in the magnitude of the aftereffect.

Accuracy of performance in two distractor tasks was relatively high throughout the experiment. When attending to the distractor voice, the proportion of correct number identifications was between 0.87 and 0.91 across the listeners. In the attend-to-adaptor condition, the hit rate for detecting a level fluctuation in the adaptor ranged between 0.76 and 0.96, with false alarms being infrequent (8.3 times per block, on average). These results suggest that the listener's attention was manipulated as expected by the instructions.

The observed effects of attention on AM adaptation suggest that AM of an unattended sound is not fully analyzed relative to the level of AM analysis conferred on an attended sound. In natural environments, where the auditory system must solve the difficult problem of sound segregation, it may be beneficial for the system to relax analysis of AM of unattended sounds at some stage of processing. By this, system resources can be concentrated on segregating the target sound. Indeed, the effect of attention observed in the present study might relate to the strategy that the auditory system has adopted for efficient processing.

Bruckert et al. (2006) reported that training reduced the AM adaptation. From the finding, they discussed a possibility that the aftereffect might reflect listener's misuse of the adaptor as a reference for detection (see also Wakefield and Viemeister, 1984). In the present study, although an effort was taken to reduce the similar misuse, the distractors might prevent the listener's use of an incorrect reference and this might explain some of the reduction in the aftereffect.

Although attention affects AM adaptation, the exact nature of attention involved in the present experiment remains unclear. There are several possible interpretations of its nature. The first of which is that the number judgment task required large cognitive load, which in turn would lead to disruption of AM processing. Another possibility is that attending to the distractor voice may hinder the processing of the stimuli presented to the contralateral side, leading to less pronounced adaptation. It would be helpful to examine the nature of attention related to the observed effect for further understanding of the auditory processing of AM.

This work was supported by KAKENHI (Grant No. 19730456) and 21st Century COE programs “Center for Evolutionary Cognitive Sciences,” MEXT, Japan.

References and links

1. Bruckert, L., Herrmann, M., and Lorenzi, C. (2006). “No adaptation in the amplitude modulation domain in trained listeners,” J. Acoust. Soc. Am. 119, 3542–3545. [MEDLINE] first citation in article
   
2. Chaudhuri, A. (1990). “Modulation of the motion aftereffect by selective attention,” Nature (London) 344, 60–62. [Inspec] [MEDLINE] first citation in article
   
3. Darwin, C., and Carlyon, R. P. (1995). “Auditory grouping,” in Hearing, edited by B. C. J. Moore (Academic, London). first citation in article
   
4. Kashino, M. (1998). “A modulation filter bank as a basis for auditory spectro-temporal analysis,” Fifth International Conference on Neural Information Processing, Vol. 3, pp. 1333–1336. first citation in article
   
5. Kay, R. H., and Matthews, D. R. (1972). “On the existence in human auditory pathways of channels selectively tuned to modulation present in frequency-modulated tones,” J. Physiol. (London) 225, 657–667. [ISI] [MEDLINE] first citation in article
   
6. Kidd, G., Jr., Mason, C. R., Arbogast, T. L., Brungart, D. S., and Simpson, B. D. (2003). “Informational masking caused by contralateral stimulation,” J. Acoust. Soc. Am. 113, 1594–1603. [MEDLINE] first citation in article
   
7. Richards, V. M., Buss, E., and Tian, L. (1997). “Effects of modulator phase for comodulation masking release and modulation detection interference,” J. Acoust. Soc. Am. 102, 468–476. [MEDLINE] first citation in article
   
8. Shannon, R. V., Zeng, F. G., Kamath, V., Wygonski, J., and Ekelid, M. (1995). “Speech recognition with primarily temporal cues,” Science 270, 303–304. [Inspec] [ISI] [MEDLINE] first citation in article
   
9. Speech Database Committee of Acoustical Society of Japan (1990). “ASJ-JIPDEC speech corpus.” first citation in article
   
10. Wakefield, G. H., and Viemeister, N. F. (1984). “Selective adaptation to linear frequency-modulated sweeps: Evidence for direction-specific FM channels?,” J. Acoust. Soc. Am. 75, 1588–1592. [MEDLINE] first citation in article
    
11. Wojtczak, M., and Viemeister, N. F. (2003). “Suprathreshold effects of adaptation produced by amplitude modulation,” J. Acoust. Soc. Am. 114, 991–997. [MEDLINE] first citation in article
    
12. Yeh, S. L., Chen, I. P., De Valois, K., and De Valois, R. L. (1996). “Figural aftereffects and spatial attention,” J. Exp. Psychol. Hum. Percept. Perform. 22, 446–460. [MEDLINE] first citation in article
    

Full figure

Fig. 1. (a) Psychometric functions of the attend-to-adaptor group (left side panels) and the attend-to-distractor group (right side panels) for two different test modulation frequencies (16 and 51 Hz). (b) Magnitude of aftereffects. In both of the panels, in the adaptor condition, the two participant groups were exposed to the same adapting setup (adaptor plus a tone distractor) and were not imposed on attentional tasks during the adaptor presentation. All error bars reflect standard errors. See text for details. First citation in article

Up: Issue Table of Contents
Go to: Previous Article | Next Article
Other formats: HTML (smaller files) | PDF ( kB)