Enhanced auditory temporal gap detection in listeners with musical training Open Access

Musicians demonstrate enhanced abilities in various perceptual tasks, e.g., pitch discrimination (Micheyl et al., 2006) and speech-in-noise perception (Parbery-Clark et al., 2009). Most studies conventionally used the Western-classical musician model (e.g., Kraus et al., 2014). Here, we asked if South Indian-classical (Carnatic) musicians demonstrate the same kind of superiority in auditory perception. Knowledge of this would be beneficial to gain a more complete understanding of the unique perceptual abilities of musicians. Listeners trained in Carnatic music use different pitch structures and temporal patterns compared to those found in Western music (Krishnaswamy, 2004; Wade, 2001). Of interest here is auditory temporal resolution.

One of the frequently used psychoacoustic approaches to study temporal resolution is gap detection. Gap detection threshold (GDT) represents the sensitivity to detect the presence of a silent temporal gap between two markers. GDTs can be measured in within- and across-channel paradigms. In the within-channel task, the relevant perceptual process involves discontinuity detection, most likely the detection of the onset of the trailing marker, executed on the activity in the channel triggered by the stimulus (Oxenham, 1997). In the across-channel task, the relevant operation is the relative timing of the offset of activity in the channel representing the leading marker and the onset of activity in the channel representing the trailing marker (Phillips et al., 1997). To date, only two studies measured GDTs in musicians; one in Western- and the other in Indian-classical musicians (Mishra and Panda, 2014; Zendel and Alain, 2012). While Mishra and Panda (2014) reported a significant group difference between musicians and nonmusicians, the study of Zendel and Alain (2012) presented a minimal difference between the two groups of listeners aged below 30 years; group differences appeared to emerge around 40 years of age (see their Fig. 2). A potentially problematic issue with these two studies is that they used within-channel GDTs to assay temporal processing. Within-channel gap detection tasks have the inherent problem that the limiting factor may be intensity coding rather than temporal coding per se (Shailer and Moore, 1985). Compared to within-channel task, the across-channel gap detection task is more relevant to categorical perception of speech—voice onset time (VOT) perception (Elangovan and Stuart, 2008). Elangovan and Stuart (2008) showed that across-channel gap detection, but not within-channel thresholds, were related with the phonetic boundaries of voiced-voiceless speech sounds.

To our knowledge, musician enhancement for across-channel gap sensitivity is yet to be investigated. Characterizing across-channel gap detection in South Indian-classical musicians may be particularly interesting as Carnatic music style uses microtonal inflections that are not typical in Western-classical music (Forney and Machlis, 2011; Krishnaswamy, 2004). Microtonal inflections may involve some fine processing of the relative timing of acoustic events in different frequency ranges similar to across-channel gap detection. In addition, measuring across-channel GDTs in Carnatic musicians may provide increased external validity for the prediction that musicians have superior auditory temporal sensitivity. The objective of this study was to compare the temporal resolution, assayed by across-channel gap detection thresholds, between Carnatic musicians and nonmusicians. Specifically, two hypotheses were tested: musicians would exhibit lower GDTs compared to nonmusicians and the inter-subject variability of GDTs would be lower in musicians compared to nonmusicians. The second prediction was based on findings from perceptual learning studies that showed reduced variability following auditory training (e.g., Moore and Amitay, 2007).

Forty-four otologically healthy listeners with hearing thresholds of 15 dB hearing level or lower at octave frequencies between 250 and 8000 Hz participated in this study. All listeners had “high average” intelligence quotient (110–119) on the Wechsler adult intelligence scale (third edition). Sixteen participants (12 females and four males) between ages 18 and 31 years were self-identified musicians affiliated with the Ksheera Sagara School of Music, Chennai. Musicians were academically trained in Carnatic (South Indian-classical) music, began their training during childhood before 10 years of age (mean = 6.4 years), and had consistently practiced for the previous three years. While their musical performance ability was not evaluated nor was considered as a criterion for inclusion in the study, the information on the duration of musical experience was obtained. Data from these musicians were previously reported for another study (Mishra and Panda, 2014). Twenty-eight listeners (23 females and five males) between ages 18 and 32 years were categorized as nonmusicians and had no formal training (zero years) in music of any style. Nonmusician listeners had no reported medical conditions and/or disabilities. All participants provided informed consent according to the SRM University Medical Research Ethics Committee. Experimental procedures were conducted in an acoustically treated double-room setup.

Across-channel GDTs were measured using a customized version of Adaptive Tests of Temporal Resolution written in visual basic^® version 6.0 that plays stored sound files (Lister et al., 2011). The software was run using a Dell laptop with a Conextant CXT5047 sound card (Conexant, Irvine, California, USA) and stimuli were delivered using a calibrated MAICO MA52 audiometer (MAICO Diagnostic, Denmark). An external display was connected to the laptop to record the participant's responses.

The stimulus comprised a silent gap flanked by two narrowband noise markers with quarter octave bandwidths, geometrically centered on 2000 Hz before the gap (leading marker) and 1000 Hz after the gap (trailing marker). The onset of the leading and offset of the trailing markers were cosine-square ramped to create 10-ms rise and fall times respectively. Similarly, the offset of the leading marker and the onset of the trailing marker were shaped with a cosine-squared window to give 1-ms rise-fall times. The leading marker was 300 ms in duration and the trailing marker varied randomly in duration between 250 and 350 ms to control extraneous duration cues. Gap step sizes varied depending on the gap durations. For gap durations between 1 and 100 ms, the stimuli had 1-ms steps. Similarly, 2-ms steps were used for gap durations between 102 and 200 ms and 5-ms steps were used for gap durations between 205 and 400 ms. If the listener gave two correct responses in a row, the gap duration in the target interval was decreased by a factor of 1.2. If the listener selected an incorrect response once, the gap duration in the target interval was increased by a factor of 1.2. The appropriate sound file was selected within these limitations by the adaptive paradigm. For instance, if the adaptive paradigm calculated a gap size ≥ 1.5 and ≤ 2.4, a 2-ms gap sound file was played, and, if the paradigm calculated a gap size between 103.5 and 105.4, then a sound file with a 104-ms gap was played (Lister et al., 2011).

The GDTs were measured using an adaptive two-interval, two-alternative, forced-choice psychophysical approach targeting 70.7% correct responses (Levitt, 1970). In the standard interval, the leading and trailing markers separated by a fixed 1-ms gap were presented. This 1-ms gap ensured similar gating transients in both intervals. In the target interval, two markers separated by a gap of adaptively varying duration were presented. The standard and target intervals were presented in a random order. The listeners were instructed to select the target interval. Two correct responses in a row decreased the gap and an incorrect response increased the gap. The initial gap duration was set at 80 ms. The stimuli were delivered to the listeners’ right ears using TDH 39P headphones at 70 dB sound pressure level. GDTs were defined as the geometric mean of the final six reversals of the total eight reversals. All listeners were provided with a familiarization run prior to actual measurements.

The data followed a normal distribution assessed by the Shapiro–Wilk test of normality (nonmusicians statistic = 0.97, p = 0.53; musicians statistic = 0.96, p = 0.64). Levene's test for equality of variances showed no significant difference in variance of GDTs between nonmusicians and musicians (F = 2.61, p = 0.11). An independent-samples one-tailed t-test revealed a statistically significant difference in GDTs between nonmusicians (mean = 41.75 ms, SD = 13.79) and musicians (mean = 21.65 ms, SD = 9.36; t(42) = 5.17, p < 0.001, d = 1.73). Figure 1 shows GDTs for musicians and nonmusicians. These results indicate that musical training has a significant impact on GDTs. Specifically, musicians have lower GDTs compared to nonmusicians. Across-channel GDTs as a function of years of musical experience is shown in a bivariate scatterplot in Fig. 2. Pearson's product moment correlation analysis showed a significant negative correlation between years of musical experience and GDTs (r = −0.49; one-tailed p = 0.027; n = 16). However, when the two potential outliers (GDTs = 7.45 and 9.03 ms; two filled circles in Fig. 2) were excluded the correlation was not significant (r = −0.14; one-tailed p = 0.32; n = 14). Because small sample size could create potential issues related to power, we do not discuss correlation results. Nonmusicians’ GDTs were not used for correlational analysis because their inclusion could yield inadvertent results since the years of musical experience for every listener was zero.

This study presents simple yet novel findings that South Indian-classical musicians, compared to nonmusicians, have enhanced sensitivity to detect the presence of a silent temporal gap between two spectrally dissimilar markers. Although a few studies demonstrated that musicians possess enhanced temporal perception skills, including within-channel gap detection (Mishra and Panda, 2014; Rammsayer and Altenmüller, 2006; Zendel and Alain, 2012), the current findings are unique for at least two reasons. First, the across-channel GDTs—a direct psychoacoustic measure of temporal resolution that overcomes potential concerns of intensity coding associated with within-channel task—have not specifically been measured in musicians regardless of music culture. Second, the use of musicians trained in non-Western (South Indian-classical) music is particularly important; musicians’ perceptual advantages are not limited to Western-classical music, on which much of the previous research on musical training has narrowly focused.

The across-channel gap detection task involves the detection of a silent interval between the offset of excitation in one perceptual channel and an onset in a distant perceptual channel. Extraneous cues, such as spectral splatter, could affect across-channel gap detection. Some listeners, particularly musicians, may be better at detecting the spectral splatter in the signal interval. This extraneous spectral cue was controlled by using a fixed 1-ms gap between the leading and trailing markers in the standard interval. However, possibility of some qualitative differences, though less likely, between the spectral cues available in the standard and target intervals cannot be completely eliminated.

On average, across-channel GDTs in musicians were half (0.52 times) the size of that in nonmusicians. In contrast, within-channel GDTs in musicians (mean = 3.53 ms) were 0.24 times smaller than that in nonmusicians (mean = 4.67 ms) (Mishra and Panda, 2014). The across-channel GDTs in nonmusician listeners are consistent with previous studies that used similar methods (Lister et al., 2011). Lister et al. (2011) reported the mean across-channel GDT in young normal hearing individuals to be 42.6 ms. The performance on pitch discrimination tasks for nonmusicians has been shown to improve after a few hours of training (Micheyl et al., 2006). Although we provided one familiarization run to all listeners, we did not train the nonmusicians. Thus it remains to be seen whether after brief gap-detection training nonmusicians can achieve musician-like performance on gap detection tasks.

Contrary to our expectation, the inter-subject variability in GDTs in musicians was no different than nonmusicians, suggesting that natural variation in temporal resolution also exist among musicians, and may also imply a continuum of GDTs from nonmusicians to musicians. Present data, like other perceptual studies (Micheyl et al., 2006; Parbery-Clark et al., 2009; Rammsayer and Altenmüller, 2006; Zendel and Alain, 2012), cannot confirm if the musician listeners underwent musical training in part because of pre-existing or predisposition to superior auditory skills.

Many perceptually relevant tasks mimic the across-channel gap detection paradigm, e.g., detecting the VOT to perceptually discriminate voiced stop /b/ and voiceless stop /p/ consonants (Elangovan and Stuart, 2008). Lower gap detection thresholds observed in musicians are consistent with previous reports that showed improved neural encoding of temporal cues of speech (Chobert et al., 2012; Parbery-Clark et al., 2009).

In summary, the inclusion of Carnatic musicians in the current study provided an alternative model to further validate the findings of superior auditory temporal resolution in musicians. This study lays the groundwork for investigating various auditory and cognitive mechanisms in Indian-classical musicians, and possibly comparing with Western-classical musicians. Such work may be beneficial to gain a complete understanding of the mystique that surrounds the perceptual skills of musicians. From an applied perspective, it highlights the importance of considering musical experience when assessing a person's gap detection threshold.

The authors thank the musicians for their time and effort in participating in this study. They greatly appreciate Jennifer Lister for generously providing the gap detection test software (attr^®). The authors appreciate discussions with Moumita Choudhury on the differences between Indian- and Western-classical music.