Diotic and dichotic frequency discrimination thresholds in musicians and non-musicians: relationships between perception, musical ability and self-evaluated competence

Pitch perception provides important information for musical and vocal communication. Numerous studies have shown that musical training and expertise are associated with better pitch processing, however, it is unclear what types of pitch percepts are plastic with music training. The current study addresses this issue by measuring discrimination thresholds of Musicians (n=20) and Non-musicians (n=18) to diotic (same sound to both ears) and dichotic (different sounds to each ear) stimuli created from four types of acoustic computations:1) pure sinusoidal tones, PT; 2) four-harmonic complex tones, CT; 3) iterated rippled noise, IRN; and 4) interaurally correlated broadband noise, called “Huggins” or “dichotic” pitch sounds, DP. Frequency Difference Limens (DLF) in each condition were obtained via a 3-alternative-forced-choice adaptive task requiring selection of the interval with the highest pitch, yielding the smallest perceptible fundamental frequency (F0) distance (in Hz) between two sounds. Music skill was measured by an online test of musical Pitch, Melody and Timing (International Laboratory for Brain Music and Sound Research, https://www.brams.org/en/onlinetest/). Musicianship, length of music experience and self-evaluation of musical skill were assessed by questionnaire. Results showed musicians had smaller DLFs in all four conditions and that thresholds were related to subjective and objective musical ability. In addition, self-report of musical ability was shown to be a significant variable in group classification, suggesting that the neurobehavioral profile of musicians includes self-evaluation of musical competence.


37
Musical training is associated with better pitch encoding and perception (for review see (1)). In music,

38
pitch is the quality that most strongly defines the perception of the melodic contour. Each note in a grade 9, age 14-15) and 3) a total of at least 5 years in formal music education. 20 subjects fulfilled the 87 criteria for MU group inclusion, with the remainder 18 subjects grouped into Non-musicians (NM).

88
Group characteristics of age, music education, self-ratings and objective measures of musical skill (i.e.

89
online aptitude test, for description see below) are presented in Table 1.

91
STIMULI. Sounds were 300 ms in duration, with two 60-ms raised cosine ramps for onset and offset.
92 Figure 1 shows time waveforms (left panels) and frequency spectra (right panels) for the 440 Hz

93
(standard) stimuli used in the study. 440 Hz was chosen because it is a familiar musical note (A4) that 94 elicits strong phase-locking.

95
In order to test binaural mechanisms, we created a dichotic pitch (DP) stimulus, often called "Huggins' 96 pitch," which consists of dissimilar right and left inputs to make a dichotic estimation of a sound's pitch 97 (18,19). DP stimuli were created with the Binaural Auditory Processing Toolbox for MATLAB8 using a 98 transition width of 16%. DP sounds were made of white noise, diotic at all frequencies except for a 99 narrow band at the F0 (440 Hz), over which the interaural phase transitioned progressively through 360°.

100
Individuals were familiarized with DP perception through five online Demonstrations

101
(https://web.stanford.edu/~bobd/cgi-bin/research/dpDemos/). Self-reported music education, music skill and listening frequency measures obtained via questionnaire and are reported in years. Only 8 Non-musicians had previous music education. Self-reported music skill was rated on a scale from 1-9, with 1 being "novice" and 9 denoting "professional". Music listening frequency was rated on a scale from 1-9 with 1 being "never" and 9 "all the time". Melody, Timing, Pitch and Average/Total musical skill scores obtained via online aptitude test (www.brams.org) and are reported in percent correct. Figure 1 In contrast, a pure tone (PT), shown in panel B of Figure 1 is the product of a sinusoidal function.

108
Sinusoids are thought to be encoded by place code mechanisms because they elicit narrow bands of 109 maximal activation at specific places in the tonotopic map of the cochlea. At lower frequencies (<~2 kHz) 110 elicit additional phase-locked temporal codes at the frequency's period. Pure tones consisted of sinusoids 111 at a fundamental frequency (F0) of 440 Hz, chosen.

112
We also tested an iterated noise (IRN) stimulus which evokes a pitch perception that is primarily reliant 113 on temporal information (25)(26)(27). IRN stimuli, shown in Figure 1C, were created from Gaussian

117
Finally, we used a complex tone with three harmonic overtones (CT), which most closely resembles the 118 sound a musical instrument makes and relies on a combination of place and temporal codes. Complex 119 tones ( Fig. 1D)

134
Musician self-identification (e.g. "Are you a musicians?"), 2) Self-Report of Music Listening Frequency 135 on a scale of 1-9 3) Self-Report of Musical Skill on a scale of 1-9 4) Age of Music Start and 5) Years of 136 Consistent Practice. Group means and standard deviations are shown in Table 1.  Table 1). Examination of the detrended SR Musical Skill scores showed that one NM rated themselves 141 >1 Standard Deviation from normal and one MU rated themselves >-1 Standard Deviation from normal.

142
Examination of the detrended BRAMS pitch scores showed that one individual from each group scored >-143 1 Standard Deviation from normal. Given that a skew in distribution was observed for two measures, we 144 provide observed power for each test and only conducted tests that were robust to the assumption of 145 normality (29, 30).

146
The group difference hypothesis was tested using a set of mixed repeated-measures ANOVAs

168
Group means show that musicians had lower thresholds for each sound type category (Fig. 2, Supp. Table   169 1). Bar graphs in Figure 2 illustrates the group mean values for each sound type category, showing lower 170 means for the musician group in all categories, relative to Non-musicians. Taken together, the data show 171 that musicians can hear smaller pitch differences that Non-musicians in all four pitch-evoking sound type

185
Examination of group means showed that threshold variance is smaller in MU than NM in the DP and PT 186 condition (Supp. Table 2).

Relationships between pitch discrimination thresholds, self-reports and musical aptitude measures
188 Pearson's correlations show that better discrimination thresholds are associated with a higher self-report 200 musical skill (scaled between 1-9, with 1 being novice, 9 professional). Higher self-report is associated 201 with smaller (better) thresholds. Right column shows relationships between pitch discrimination 202 thresholds behavioral scores obtained from the BRAMS musical skills test (objective). Higher score is 203 associated with smaller (better) thresholds.  220 Discussion

221
We have answered two main questions in this study: 1) Are musicians better at perceiving specific pitch-

229
To answer the first question, DLF data were subjected to a RMANOVA with four within-subject factors 230 of sound type and two between-subject factors of group. Results showed group differences across all 231 sound types, with the greatest differences for dichotic and pure tone stimuli. These data refute our initial 232 hypothesis that pitch-related temporal encoding mechanisms would be most impacted by musicianship; 233 instead suggesting that music-related plasticity is not restricted to types of pitches. The greatest difference 234 between Musician and Non-musician discrimination thresholds in the dichotic condition suggests that higher-order mechanisms, such as those requiring a combination of sound across the ears, are greatly 236 impacted by musical training.

237
Several hypotheses could reasonably explain our findings. One hypothesis is that mechanisms of music-238 related brain plasticity are not restricted to place or temporal code encoding mechanisms in peripheral or 239 brainstem nuclei (11), but may also occur cortically (12), or at least beyond the superior olive where 240 dichotic sounds first combine. Unfortunately, our current data do not permit further elucidation on the 241 veracity of this postulate because we do not have encoding data to test brainstem and cortical plasticity 242 specifically. An alternative hypothesis is that playing music sharpens one's ability to extract pitch 243 percepts in conditions where the pitch strength is less salient, such as the dichotic and iterated rippled 244 noise conditions. If this hypothesis were true, we might expect that the largest differences between the 245 two groups would be in the least salient conditions. Whereas the largest threshold difference is in the 246 dichotic condition (less salient pitch), the second largest threshold difference is observed in the pure tone 247 condition, which has the most salient pitch strength. Although our data do not directly address the issue of 248 pitch strength, the fact that the largest differences are observed with both strong and weak pitch percepts 249 diminishes this hypothesis' likelihood. A third, big picture, hypothesis is that Musicians possess a greater 250 aptitude to learn the task than Non-musicians. If this were true, we would expect Musicians to learn the 251 task faster than Non-musicians. A post-hoc examination of the within-session change in threshold showed 252 that Non-musicians did have more variability, measured by standard deviation (Supp . Table 4). However, 253 mean magnitudes of the within-session change in threshold over the four runs, computed by subtracting 254 the threshold obtained in the first run from the threshold obtained in the last run, did not appear to differ 255 between groups (Supp. Table 4). To verify our observations, we performed two RMANOVAs for sound 256 type and group on the standard deviation and within-session change data. Results showed that Musicians 257 had lower standard deviation in thresholds to dichotic and pure tone stimuli, compared to Non-musicians, 258 but only in the pure and complex tone conditions. No significant differences were observed for within-or 259 between-subject comparisons of the within-session threshold change magnitude. Taken together, these data suggest that acclimatization or learning trajectories from task beginning to end is similar in

261
Musicians and Non-musicians and that musicianship positively influences dichotic and pure tone pitch 262 discrimination, in part by stabilizing threshold reliability.

263
In answering the second question, we showed evidence for a relationship between psychoacoustic pitch 264 discrimination and measures of subjective and objective music ability. The correlation data show that 265 discrimination thresholds across all four pitch types were negatively correlated with a higher subjective 266 rating of musicianship, such that individuals who rated themselves with musical ability closer to 267 "professional" on a subjective scale, could hear smaller pitch differences in all four sound conditions.

268
Conversely, individuals who rated themselves with a lower musical ability (i.e. closer to "novice" on the 284 musical aptitude and self-assessment of musical skill, but may be independent of pitch perception. Taken together, the correlation data show that the ability to discriminate small pitch differences can be reflected 286 in global musical abilities and an individual's evaluation of their own musical aptitude. This implies that 287 sensory thresholds for pitch discrimination underlie, at least in part, one's musical ability and self-288 appraisal of that ability. Furthermore, relationships between sensory threshold for pitch and more broad 289 measures of musicianship are not restricted to a specific mechanism of pitch processing.

290
The Discriminant Analysis allowed us to detect the degree to which our variables discriminate between

291
Musicians and Non-musicians. The variables that contributed most to the predictions of group 292 membership were 1) Self-report of musical ability on a scale of 1-9, 2) Pure Tone DLFs and 3) BRAMS

293
Avg./Total score. While the relationship between pure tone perception, musical aptitude and musicianship 294 is well established, the contribution of a self-report variable is novel as far as the authors' knowledge.

295
Here, we show that self-evaluation of musical competence can be meaningfully applied to classify groups 296 and is related to objective measures of music and perceptual ability. Self-evaluation of competence, or 297 self-competence is defined as the sense of one's capacity. (31) Previous data on this topic show that 298 general self-competence is as associated with measures of cognitive ability such as IQ and academic 299 achievement measured by GPA. (32) Our data support the argument that self-evaluation of competence is 300 a meaningful measure of ability and outcomes (33) and extend into musicianship.

301
In addition to the finding of self-report as a meaningful measure, the discriminant analysis showed 302 common characteristics of musicians include psychoacoustic, musical and self-evaluated abilities. This

303
gives rise to the notion that all three areas may interact to define a person who is talented or skilled in 304 music. It is interesting to note that the self-reported music listening scale did not distinguish between 305 groups. This supports several lines of research showing that active music-making, rather than listening 306 alone, is a catalyst for brain plasticity and internalized perceptual change (23,34,35).

307
In conclusion, this study sheds light on several aspects of musicianship. First, we show that the influence 308 of musicianship is not limited to pitch judgements involving monotic/diotic mechanisms but also includes 309 those that rely on dichotic integration. Second, our data show that basic perceptual thresholds are related