Absolute pitch memory: Its prevalence among musicians and. dependence on the testing context

Transcription

1 Absolute Pitch Memory 1 Absolute pitch memory: Its prevalence among musicians and dependence on the testing context Yetta Kwailing Wong 1* & Alan C.-N. Wong 2* Department of Applied Social Studies, City University of Hong Kong, Kowloon Tong, Hong Kong 1 Department of Psychology, The Chinese University of Hong Kong, Shatin, Hong Kong 2 Citation: Wong, Y. K. & Wong, A. C.- N. (2014). Absolute pitch: Its prevalence among musicians and dependence on the testing context. Psychonomic Bulletin & Review, 21(2), Word count: 3990 Keywords: pitch processing, auditory, music training, context, multimodal, expertise *Corresponding Authors: (1) Yetta Kwailing Wong Y7414, Academic Building I, Department of Applied Social Studies City University of Hong Kong Tat Chee Avenue, Kowloon, Hong Kong yetta.wong@cityu.edu.hk Phone number: (2) Alan C.-N. Wong 344 Sino Building Department of Psychology The Chinese University of Hong Kong, Shatin, Hong Kong alanwong@psy.cuhk.edu.hk Phone number:

2 Absolute Pitch Memory 2 Abstract Absolute pitch (AP) is widely believed to be a rare ability possessed by only a small group of gifted and special individuals ( AP possessors ). While AP has fascinated psychologists, neuroscientists and musicians for more than a century, no theory can satisfactorily explain why this ability is so rare and difficult to learn. Here we show that AP ability appears rare because of the methodological issues of the standard pitchnaming test. Specifically, the standard test unnecessarily poses a high decisional demand on AP judgments and uses a highly inconsistent testing context to one s musical training. These extra cognitive challenges are not central to AP memory per se, and have thus led to consistent underestimation of AP ability in the population. Using the standard test, we replicated the typical findings that the accuracy for general violinists was low (12.38%; chance level = 0%). With identical stimuli, scoring criteria and participants, violinists attained 25% accuracy in a pitch-verification test in which the decisional demand of AP judgment was reduced. When the testing context was increasingly similar to their musical experience, verification accuracy improved further and reached 39%, three times higher than that for the standard test. Results were replicated with a separate group of pianists. Our findings challenge current theories about AP, and suggest that the prevalence of AP among musicians has been highly underestimated in prior work. A multimodal framework is proposed to better explain AP memory.

3 Absolute Pitch Memory 3 Introduction To the majority of people, labeling the pitch of an isolated tone is difficult (Takeuchi & Hulse, 1993), unless they are given an external pitch reference beforehand (Ward, 1999). Years of explicit musical training do not make this task any easier for professional musicians (Athos et al., 2007; Levitin & Rogers, 2005; Zatorre, 2003). A small group of people, however, can label or produce isolated tones accurately and effortlessly. They are typically called absolute pitch (AP) possessors, and conventional estimates suggest that only 1 out of people have this ability (Takeuchi & Hulse, 1993). This remarkably rare ability has been considered a special talent and musical endowment for gifted musicians (Deutsch, 2002; Ward, 1999; but see Miyazaki, 1993), and has fascinated musicians, psychologists and neuroscientists for more than a century (Deutsch, 2002; Levitin & Rogers, 2005; Takeuchi & Hulse, 1993; Ward, 1999). Different theories have been proposed to explain the differences between AP possessors and non-possessors. One theory suggests that, in adulthood, pitch memory is organized along a free-floating helix, so it is impossible for general musicians to name a pitch without external reference (Ward, 1999). Only for AP possessors, the helix is somehow well anchored with a permanent pitch label, which explains their ease in absolute pitch labeling (Ward, 1999). However, this account does not explain why the general population has highly precise pitch memory for familiar songs and recordings of popular television shows (Halpern, 1989; Levitin, 1994; Schellenberg & Trehub, 2003). In contrast, a widely accepted working hypothesis suggests that general listeners have AP memory to a considerable extent (Halpern, 1989; Levitin, 1994; Schellenberg & Trehub, 2003), while only the AP possessors can associate the represented pitches with

4 Absolute Pitch Memory 4 verbal names (Brancucci et al., 2009; Deutsch, 2002; Levitin & Rogers, 2005; Schellenberg & Trehub, 2003; Vanzella & Schellenberg, 2010). Nonetheless, the puzzle remains as to why general listeners encounter such a great difficulty associating represented pitches with names. Associating names with established concepts should be a trainable skill, as demonstrated by children learning to name familiar objects in new languages (Gathercole & Baddeley, 1990), and adults learning to name novel objects with non-words within hours (Wong, Palmeri, & Gauthier, 2009; Wong, Folstein, & Gauthier, 2011). Why, then, do general musicians, after having spent years of explicit training in associating pitches with labels, still fail to name the pitches that are represented in auditory memory? Why did most of the intensive AP training in adulthood result in limited success (Brady, 1970; Cuddy, 1970; Takeuchi & Hulse, 1993; Ward, 1999)? Is the low AP performance of general musicians really an issue of naming, or does it stem from other causes? To address this question, let us consider how AP ability is typically measured. To express one s AP ability, one can label a sounded note verbally (e.g., this tone is a C ), produce a specific pitch by singing, reproduce a sounded note on an instrument, etc. (Takeuchi & Hulse, 1993; Zatorre, 2003). Among these, the most common and standard way to assess AP ability is by the pitch-naming test, in which observers name the pitch of tones (mostly sine wave tones) presented in isolation (Athos et al., 2007; Takeuchi & Hulse, 1993; Zatorre, 2003). There are at least two reasons why this standard pitch-naming test may be suboptimal. Firstly, to name a tone, listeners have to choose a name out of twelve possibilities (the twelve semitones in the Western music scale). However, it has been

5 Absolute Pitch Memory 5 shown in other cognitive tasks that, with the stimulus set held constant, handling more alternatives places a higher decisional demand on the listeners, leading to lowered performance (Churchland, Kiani, & Shadlen, 2008; Niwa & Ditterich, 2008). Secondly, the testing context is highly inconsistent with musicians experience. Context here does not mean an external reference (e.g., another tone and its pitch label as in relative pitch tasks); instead it refers to the circumstances involved in the presentation of a tone. For example, with extensive training, musicians represent musical notes as multimodal objects by automatically integrating information from visual, auditory, somatosensory and motor modalities (Wong & Gauthier, 2010; Zatorre & Beckett, 1989; Zatorre, Chen, & Penhune, 2007). However, the standard test eliminates much of the important information that is integrated into the pitch concept, such as the timbre, the playing posture, the fingering associated with the tone, and the visual image of the fingering and the instrument. Such a difference in context between learning and testing impairs performance in memory tasks (Godden & Baddeley, 1975; Smith & Vela, 2001). Even for AP possessors, performance are impaired with contextual inconsistencies during testing, such as the use of different timbres and pitch registers (Levitin & Rogers, 2005; Takeuchi & Hulse, 1993). If we accept AP as simply the ability to identify the pitch of an isolated tone (Athos et al., 2007; Deutsch, 2002; Levitin & Rogers, 2005; Schellenberg & Trehub, 2003; Takeuchi & Hulse, 1993; Vanzella & Schellenberg, 2010; Ward, 1999; Zatorre, 2003), the high decisional demand and the contextual inconsistencies between testing and one s musical experience seem extraneous and may have unnecessarily prevented general musicians from expressing their AP memory.

6 Absolute Pitch Memory 6 To test this hypothesis, we devised a pitch verification test for assessing AP ability. Participants judged if an isolated tone matched with a given pitch label. The pitch label was either the same as the presented tone, or differed by one or three semitones. This verification test used the most stringent scoring standard for AP ability (Takeuchi & Hulse, 1993; Zatorre, 2003). Firstly, it used isolated tones without any external pitch reference. Secondly, we calculated the pitch verification accuracy with only the most difficult trials, those with one semitone difference between the label and the tone. Failures to discriminate between neighboring semitones (e.g. treating a C as a C# ) were regarded as errors, the most stringent scoring standard used in the literature (Takeuchi & Hulse, 1993; Zatorre, 2003; Deutsch et al., 2006; Lee & Lee, 2010). The major advantage of the verification test is that the number of possible answers is reduced to two (match or non-match), thus reducing the decisional demand on the participants. All participants also performed the standard test and were scored with an identical standard for assessing how the difference in decisional demand affects AP performance. To examine whether contextual consistency affects AP judgment in general musicians, we manipulated testing contexts during verification. In Experiment 1, 36 violinists took the test at four levels of increasing contextual similarity to musical experience. The basic level used sine wave tones identical to those used in pitch naming. The second level used tones in violin timbre. At the third level, participants viewed video clips showing another person playing the labeled pitch, which either matched with the testing violin tones or not. At the fourth level, participants held the playing posture and correct fingering themselves without producing any sound according to the labeled pitch, which either matched with the testing violin tones or not. It was predicted that, with

7 Absolute Pitch Memory 7 identical sine wave tones, performance of general musicians ( non-ap possessors using conventional definition) should be better in the verification test compared with the standard naming test because of the relief in decisional demand. Their performance should further improve when the testing context becomes more similar to one s musical experience. In Experiment 2, we replicated the experiment with a separate group of 34 pianists using similar procedures, except that the video clip context was not introduced. Experiment 1 Methods Participants Thirty-six violinists (13 males, 23 females; M age = 21.6, s.d. = 3.0) were recruited in Hong Kong. On average, the violinists started violin lessons at 8 years old and had 12.8 years of playing experience. All had passed the Grade Seven examination or above of the Associated Board of the Royal Schools of Music (ABRSM). All participants reported normal or corrected-to-normal vision and hearing. They gave informed consent according to the ethics guidelines of the Chinese University of Hong Kong, and were remunerated for their participation. Materials and Stimuli The experiment was conducted on Mac Minis using Matlab (Natick, MA) with the PsychToolbox extension (Brainard, 1997; Pelli, 1997). Eighteen violin tones from D4 to G5 played by a volunteer violinist were videotaped in a soundproof room. For each tone, the playing posture was identical, and the fingering and the movement of the bow

8 Absolute Pitch Memory 8 could be seen clearly. A microphone was put at about 8cm from the violin and directly connected to a PC, which recorded the violin tones in Audacity 1.3. The sampling rate of the tones was 44100Hz. No quantization was performed as the precision of the tones was checked during recording by a tuner. Sine wave tones in the same pitch range and identical to those in prior AP tests (Bermudez, Lerch, Evans, & Zatorre, 2009) were used. All clips were edited in Audacity such that they lasted for 1 second with a 0.1-second linear onset and 0.1-second linear offset and were of the same magnitude. Procedure Pitch Verification Test. Participants first took a pitch verification test. In each trial, a pitch name was presented on a computer screen. When the participant was ready, the experimenter pressed a key to start the tone. Participants judged if the tone matched with the pitch name within 6 seconds. The pitch name matched with the tone for half of the trials. For the other half, the pitch names were either -3, -1, +1 or +3 semitones from the presented tones with equal probabilities. Ten practice trials with sine wave tones were provided with feedback before the test. There was no feedback during the test. The violinists took the verification test in four levels of contextual similarity to music training experience. The first level used sine wave tones ( sine ). The second level used tones in violin timbre ( timbre ). For the third level ( video ), a pitch label was presented with a silent video clip showing another violinist playing the labeled pitch on a violin. The fingering of the violinists could be clearly seen and always matched with that required for the pitch label. A violin tone was presented simultaneously with the video and the label. Participants had to verify whether the violin tone they heard was the same

9 Absolute Pitch Memory 9 or different from that indicated by the pitch label and the video. For the fourth level ( posture ), when the pitch name was presented, participants held the correct fingering and playing posture corresponding to the pitch label without the bow touching the string and producing any sound. When the posture was ready, the experimenter played a violin tone and participants judged whether the tone matched with the pitch name or not. The condition order was counterbalanced across participants. Participants brought their own violin to the experimental session. There were 144 trials for each condition and thus 576 trials in total. Standard Pitching-Naming Test We created an abridged version of the standard pitch-naming test (Takeuchi & Hulse, 1993; Zatorre, 2003). Sine wave tones from B3 to A#5 were used since this pitch range was the most familiar for violinists. In each trial, a tone was presented, followed by a picture showing the 12 possible pitch names coupled with 12 letter keys. Participants named the pitches by key presses with no time limit. No feedback was provided. Each pitch was tested twice and there were 48 trials in total, presented in a randomized order. Analyses The most stringent scoring standard was used in both verification and pitchnaming, in which failures of discriminating between neighboring semitones (e.g. treating a C as a C# ) are regarded as errors (Takeuchi & Hulse, 1993; Zatorre, 2003). To match the difficulty level between the two tests, only trials with the pitch name deviating

10 Absolute Pitch Memory 10 from the presented tones at -1, 0, or +1 semitones were used for the calculation of accuracy for pitch verification. Performance was measured using two dependent variables: (i) accuracy corrected for guessing and (ii) sensitivity according to the signal detection theory. First, accuracy corrected for guessing was included for comparing the naming and verification tests since the chance-level performance for the two tasks was different. For accuracy in pitch naming, chance performance was 8.33% (12 alternatives), and correction was conducted using Percent Correct = [(Raw Percent Correct 8.333)/( )] 100. For accuracy in verification, chance level was 50% (match/mismatch), and thus the following formula was used: Percent Correct = [(Raw Percent Correct 50)/(100 50)] 100. After correction for guessing, the measure offered the same metric for comparing performance in naming and verification tests (i.e., 0% and 100% representing chance and perfect performance respectively). A potential shortcoming of the use of percent correct is that it is unknown if the effects were driven by differences in sensitivity or in bias. Therefore sensitivity was also used to compare between different context conditions (note that there was no A for the naming test). We used A as a non-parametric measure of sensitivity according to the signal detection theory without the assumption of normality or that of equal variance (Stanislaw & Todorov, 1999, Wong et al., 2011, 2012). It is calculated as: # A'=.5 + sign(h F) (H F)2 + H F & % ( $ 4max(H,F) 4HF ' where H and F represent hit rate and false alarm rate respectively.

11 Absolute Pitch Memory 11 Results Using the standard naming test, we replicated the typical observation in the literature. Seven violinists were identified as typical AP possessors with 70% accuracy or above (for prior work adopting similar accuracy level for identifying AP possessors without correction for guessing, see Athos et al., 2007; Bermudez & Zatorre, 2009; Levitin & Rogers, 2005; Takeuchi & Hulse, 1993). As shown in Figure 1A, the remaining 29 violinists had a low accuracy in naming sine wave tones, with a median of 6.82% and a mean of 12.38% (Fig. 1A), consistent with past observations that only a small group of individuals perform well in the standard pitch naming test (Athos et al., 2007; Takeuchi & Hulse, 1993). We excluded data from the seven AP possessors due to the obvious ceiling issue and analyzed the data from the remaining 29 violinists. Accuracy was reported in terms of both percent correct and A, except when comparisons with the standard naming test were involved (as there was no A for the naming task) Insert Figure 1 here The verification test showed much better performance than the standard naming test. Accuracy improved drastically from 12.38% in pitch-naming to 24.65% in verification of sine-wave tones, [t(28)=3.48, p=.0017, d=.64], indicating that simply reducing the decisional demand allowed violinists to better express their AP ability. Further analyses revealed the importance of a musical testing context in measuring AP ability. Accuracy for posture was better than that for sine [% correct:

12 Absolute Pitch Memory 12 t(28)=4.28, p=.0002, d=1.39; A : t(28)=3.90, p=.0005., d=.72], timbre [% correct: t(28)=3.38, p=.0021, d=.62; A : t(28)=3.21, p=.0032., d=.59] and video [% correct: t(28)=2.33, p=.0270, d=.43; A : t(28)=3.37, p=.0022, d=.62]. Although performances for the latter three conditions were not significantly different (% correct: ps >.12; A : ps>.62), a trend analysis showed that verification performance improved linearly when the testing context was increasingly similar to their own musical experience, from sine, timbre, video to posture [% correct: 24.65%, 26.69%, 31.35%, 39.08%, t(28)=4.23, p=.0002; A :.680,.683,.694,.760, t(28)=3.37, p=.0021]. As a group, the violinists performed better than chance level in all conditions (ps<.0001). Notably, for the best-performing third of the violinists in posture, accuracy averaged 66.11%, approaching the conventional definition of the AP possessors (Athos et al., 2007; Bermudez et al., 2009; Levitin & Rogers, 2005; Takeuchi & Hulse, 1993). The standard test was not sensitive enough to detect the AP ability possessed by some of the violinists. In the standard pitch-naming test, the accuracy of the lowerperforming third of the violinists was not different from chance (mean = 0.25%, p=.681). Yet in the verification test, the accuracy of the same participants became significantly better than chance with identical sine wave tones (mean = 9.05%, p=.001), and rose to 22.43% in the posture context. In other words, their AP ability, which is reliably above chance, can be observed with appropriate testing contexts, but not with the most commonly used standard AP test. In sum, simply reducing the decisional demand allows general musicians to better express their AP ability, and performance further improves with musical testing contexts. This is in stark contrast to the typical observation that only a small group of individuals

13 Absolute Pitch Memory 13 possesses AP (Athos et al., 2007; Takeuchi & Hulse, 1993). Next, we examined if these results could be replicated with a separate group of pianists. Experiment 2 Method Thirty-four pianists (4 males, 30 females; M age = 20.0, s.d. = 1.84) were recruited in Hong Kong. On average, they started learning piano at 6.4 years old and had 12.5 years of playing experience. All had passed the Grade Eight examination or above of the ABRSM. The testing methods and procedures were identical to that of Experiment 1 except for the following. First, 18 piano tones of the same pitches as the violin tones recorded with an electric keyboard (Yamaha S31) were used. Second, the pianists performed in the sine, timbre and posture conditions (with the video condition dropped) with an electric keyboard (Yamaha S31). There were 432 trials in total in the verification test. Results Results in Experiment 1 were replicated with pianists. First, using the standard pitch-naming test, we identified three pianists as AP possessors with 70% accuracy or above (Athos et al., 2007; Bermudez et al., 2009; Levitin & Rogers, 2005; Takeuchi & Hulse, 1993). As shown in Figure 1B, the remaining 31 pianists showed a typical low accuracy, with a median of 4.55% and a mean of 11.07%. We excluded data from the three AP possessors due to the obvious ceiling issue and analyzed the data from the remaining 31 pianists. Accuracy was again reported in terms of both percent correct and

14 Absolute Pitch Memory 14 A, except when comparisons with the standard naming test were involved (as there was no A for the naming task). Performance in the verification test for sine wave tones was 21.27%, significantly higher than that in the standard naming test [t(30)=2.93, p=.0063, d=.52], indicating the importance of reducing the decisional demand of the AP test. Furthermore, accuracy for posture was higher than that for sine [% correct: t(30)=3.04, p=.0048, d=.54; A : t(30)=3.35, p=.0022, d=.60], and marginally higher than that for timbre [% correct: t(30)=1.84, p=.0755, d=.33; A : t(30)=2.94, p=.0967, d=.30]. Trend analysis indicated that verification accuracy increased linearly from sine, timbre to posture [% correct: 21.27%, 27.60%, 31.96%, t(30)=3.27, p=.0026; A :.657,.696,.728, t(30)=3.19, p=.0033]. As a group, the pianist performed with above-chance accuracy in all conditions (ps<.005). For the best-performing third of the pianists in posture, accuracy averaged 57.41%, again getting close to the conventional definition of AP possessors (Athos et al., 2007; Bermudez et al., 2009; Levitin & Rogers, 2005; Takeuchi & Hulse, 1993). The standard test was again insensitive to reliable AP abilities of some of the pianists. In the standard pitch-naming test, the lower-performing two-thirds of pianists did not perform above chance (mean = 0.65%, p=.540). However, their accuracy rose above chance in the verification test with identical sine wave tones (mean = 16.84%, p<.0001), and improved to 26.81% in posture. Overall, we replicated findings in Experiment 1 with a separate group of pianists, including the substantial improvement on AP performance by reducing the decisional demand and by providing a familiar musical context during testing.

15 Absolute Pitch Memory 15 Combined Analyses of Violinists and Pianists To know if the context effect was robust across musicians with different levels of AP memory, we examined the sensitivity (A ) data for participants in both experiments, including the AP possessors. We first collapsed the violinists and pianists, and divided them into four groups ( AP possessors, upper third, middle third, and lower third) using their standard naming test performance. Then for each group we compared A in the different context conditions ( sine, timbre, and play ). Figure 2 shows the performance of these groups Insert Figure 2 here All groups in general showed an increase in A as the context became richer, except for the AP possessors whose performance was at ceiling. An analysis of variance (ANOVA) showed only significant main effects of context [F(2,132)=8.58, p=.0003, η p =.104] and group [F(3,66)=32.48, p<.0001, η p =.59] but not their interaction [F(6,132)=1.69, p=.126]. However, separate ANOVAs for the groups showed a context effect in the upper third [F(2,42)=4.68, p=.014, η p =.18], the middle third [F(2,40)=5.86, p=.005, η p =.22], and the lower third [F(2,32)=4.37, p=.020, η p =.21], but not in the AP possessors [F(2,18)=1.81, p=.190]. Similarly, trend analyses also showed that A increased linearly with a richer context in the upper third [t(21)=2.67, p=.014], the middle third [t(20)=3.23, p=.003], and the lower third [t(16)=2.41, p=.027], but not in the AP

16 Absolute Pitch Memory 16 possessors [t(9)=1.30, p=.224]. While a context effect was not found among the AP possessors, it remains to be seen if the AP possessor would also show a context effect given a larger sample size (10 in our current sample) and performance away from ceiling. For others, the context effect seems to hold irrespective of their level of performance. General Discussion This study asked why AP has been found to be a special and rare ability possessed only by a few individuals but not by the general musicians. Using a pitch verification test with reduced cognitive demands, we demonstrated that general musicians have a much better AP memory than previously estimated by the standard pitch-naming test. With identical stimuli, our sample of musicians attained 22.9% accuracy in verification, two times as high as that in pitch-naming (11.7%). Performance was further enhanced when the testing context became increasingly similar to musical experience. Accuracy increased linearly to 35.40% in the posture condition, a three-fold increase of that in pitch-naming. All these differences were obtained in the same participants with the same scoring strengency. These findings challenge current theories about AP. First, AP memory is much more prevalent than previously estimated, leading to questions about how rare and special AP is (Athos et al., 2007; Levitin & Rogers, 2005; Takeuchi & Hulse, 1993; Ward, 1999; Zatorre, 2003). Second, the findings disagree with the theory proposing that general musicians cannot associate absolute pitches with verbal names (Brancucci et al., 2009; Deutsch, 2002; Levitin & Rogers, 2005; Schellenberg & Trehub, 2003; Vanzella & Schellenberg, 2010). During verification, participants were required to match the tones

17 Absolute Pitch Memory 17 with a pitch name. A failure in retrieving the names of the tones should have resulted in chance performances for all conditions. Note that our goal was not so much to determine the exact prevalence of AP in the population as to point out how AP has been consistently underestimated with the use of the standard test, which may have led to a misinterpretation of the prevalence of AP, at least among musicians. The prevalence of AP in our posture condition cannot be explained by the fact that our Asian participants speak tonal languages (Cantonese), which has been associated with exceptionally high AP occurrence rate (Deutsch et al., 2006; Lee & Lee, 2010). First, the effect of different tests and musical contexts were revealed with a within-subjects design, therefore these results cannot be explained by the specific demographic backgrounds of the participants. Moreover, the rate of AP occurrence was low compared with prior reports with tonal language speakers. Specifically, with similar onset age of musical training and AP criterion (at least 70% accuracy in the pitch-naming test), only 14% of our 70 musicians were considered AP possessors, which was more comparable with that in the American population (about <12%), but far lower than the 40%-70% in tonal language speakers reported in prior studies (Deutsch et al., 2006; Lee & Lee, 2010). These confirmed that our results cannot be explained by the especially high AP occurrence rate among tonal speakers. Is the context-dependent AP true AP? While there is no consensus regarding the importance of context on AP judgment in the literature (e.g., Ward, 1999; Zatorre & Beckett, 1989), most researchers adopt a simple definition for AP, i.e., the ability to identify or produce the pitch of a tone without

18 Absolute Pitch Memory 18 external reference (Athos et al., 2007; Deutsch, 2002; Levitin & Rogers, 2005; Schellenberg & Trehub, 2003; Takeuchi & Hulse, 1993; Vanzella & Schellenberg, 2010; Ward, 1999; Zatorre, 2003). Under this definition, AP ability does not preclude the use of contextual cues, e.g., context-dependent AP has been defined as one of the forms of AP (e.g. AP only for the timbre of one s instrument; Levitin & Rogers, 2005). Regardless of what definition of true AP one adopts, the importance of contextual cues on AP memory should be recognized. While the idea that context influences AP performance is not new, all previous studies have focused on the effects on AP possessors (Levitin & Rogers, 2005; Takeuchi & Hulse, 1993). Our findings demonstrate that acknowledging the contextual influence on AP memory among general musicians reveals a very different picture about the prevalence of AP in the population. The multimodal framework for understanding AP We believe that AP memory representation can be better understood if one takes into account the multimodal nature of musical experience (e.g., Zatorre & Beckett, 1989). For musicians, the concept of pitch is not only tied to the frequency of the sound, but also highly associated with multimodal information, including other dimensions of the auditory input (e.g. timbre), somatosensory and motor input (e.g. fingering, arm positions or lip vibration), visual input (e.g. musical notation or the instrument), and other specific contexts (e.g. the melody and arrangement of a familiar song; Halpern, 1989; Levitin, 1994; Schellenberg & Trehub, 2003; Bermudez et al., 2009; Wong & Gauthier, 2010; Zatorre & Beckett, 1989). In this multidimensional representation of pitch, all other associations may serve as cues to specify the pitch information, as well as specifying

19 Absolute Pitch Memory 19 information in other modalities. For example, motor prediction in music playing may incorporate online sensory information (e.g., the auditory tone, visual musical notation and somatosensory vibration) to adjust and refine the motor plan and execution in the feedforward-feedback loop of information flow (Desmurget et al., 2000). Such associations are largely shaped by music training and experience. Under this framework, AP memory representation refers to the multimodal representation of pitch when external pitch references are absent. Therefore, performance in the standard pitch-naming test depends on how well one can extract pitch information from this multimodal pitch representation, which is constrained by at least two factors. One is the quality of this multimodal pitch representation. The other is how well one can dissociate pitch from its multimodal associations. This framework predicts that AP memory is best expressed with an ideal musical context, e.g., when musicians are allowed to play the instrument as normal, such that musicians can naturally compare the testing tone with the multimodal pitch representation. In this case, the variability of AP performance is mainly determined by the quality of multimodal pitch representation. However, typical AP tests have eliminated at least part of the musical context. For example, the use of sine wave tones almost completely deprive one of the musical context, while our posture condition excluded the arm movement of the bow and the string vibration during sound production. In these cases, for a given quality of multimodal pitch representation, performance is modulated by the ability of dissociating pitch from the multimodal context. Those who find it easy can excel at the AP tests

20 Absolute Pitch Memory 20 regardless of impoverished or rich contexts, while others are impaired to different degrees when musical cues are eliminated. In this framework, the AP possessors are those who have great quality of multimodal pitch representation and a high capability of extracting pitch from its musical context. Such extraction is seldom perfect, therefore contextual influences on their AP judgment can still be observed (Levitin & Rogers, 2005; Takeuchi & Hulse, 1993). The non-ap possessors are those who have limited ability in dissociating pitch from the context in general. Thus they perform poorly at the standard pitch-naming test. However, with a rich musical context, their AP judgment becomes above-chance or even highly accurate according to each individual s quality of the multimodal pitch representation. The ability to dissociate pitch from its context may also explain the higher occurrence of AP among individuals with Autism or with Williams Syndrome (Sacks, 1995), since both disorders have been associated with the preference for processing local features while ignoring contextualized meanings and relationships (Bellugi et al., 2000; Happe, 1999). This framework generates new questions regarding the mechanisms of AP. For example, AP ability is associated with an early onset of musical training (Takeuchi & Hulse, 1993). One should clarify whether the benefit of early musical training is about establishing AP memory of better quality, developing pitch memory independent of the musical context, or both. One should also clarify the nature of the shift of pitch processing from an absolute to relative basis during development (Stalinski & Schellenberg, 2010; Takeuchi & Hulse, 1993) since musicians maintain both types of abilities in adulthood, as suggested by the prevalence of AP. Besides, while AP ability is associated with hemispheric asymmetry of the platnum temporale in terms of size

21 Absolute Pitch Memory 21 (Keenan, Thangaraj, Halpern & Schlaug, 2001; Schlaug, Jancke, Huang, & Steinmetz, 1995), it is unclear whether such platnum temporale asymmetry supports better AP memory, context-independent retrieval of pitch, or both. And while widespread multimodal brain regions are engaged by visual presentation of musical notes (Wong & Gauthier, 2010), it remains to be seen if stronger connectivity can be observed among these regions for those with better AP memory. Last but not least, our finding that all musicians carry AP memory to a certain degree seems to be at odd with the prior findings that associated AP ability with very few or even a single genetic locus (Drayna, 2007). Further studies should clarify the contribution of carrying the AP-related genes to AP memory, e.g., leading to better AP memory, or context-independent retrieval of pitch information, or both. These are important questions for understanding the genesis of AP. Finally, this framework raises concerns about using the standard test for measuring AP memory. Our results demonstrate how the standard test consistently underestimates AP ability compared with the verification test using identical sine wave stimuli, because the standard test includes extraneous decisional demand that is noncentral to AP memory. Unless researchers would like to specifically study one s ability in the standard test (i.e., one s AP ability under high decisional demand), one should consider minimizing the extraneous decisional demand of the standard test when assessing AP memory in future.

22 Absolute Pitch Memory 22 Acknowledgement This research was supported by grants from the Chinese University of Hong Kong (Direct Grant ) to Alan Wong. The authors declare no conflict of interest. We thank Helene Wong Hoi Shan for her help in violin tone production, Crystal Yuen for her help in data collection, and Patrick Bermudez for providing the sine wave tones used in the standard AP test.

23 Absolute Pitch Memory 23 References Athos, E. A., Levinson, B., Kistler, A., Zemansky, J., Bostrom, A., Freimer, N., et al. (2007). Dichotomy and perceptual distortions in absolute pitch ability. Proc Natl Acad Sci U S A, 104(37), Baharloo, S., Johnston, P. A., Service, S. K., Gitschier, J., & Freimer, N. B. (1998). Absolute pitch: an approach for identification of genetic and nongenetic components. Am J Hum Genet, 62(2), Bellugi, U., Lichtenberger, L., Jones, W., Lai, Z., & St George, M. (2000). I. The neurocognitive profile of Williams Syndrome: a complex pattern of strengths and weaknesses. J Cogn Neurosci, 12 Suppl 1, Bermudez, P., Lerch, J. P., Evans, A. C., & Zatorre, R. J. (2009). Neuroanatomical correlates of musicianship as revealed by cortical thickness and voxel-based morphometry. Cereb Cortex, 19(7), Bermudez, P., & Zatorre, R. J. (2009). The absolute pitch mind continues to reveal itself. J Biol, 8(8), 75. Brady, P. T. (1970). Fixed-scale mechanism of absolute pitch. The Journal of the Acoustical Society of America, 48(4(2)), Brainard, D. H. (1997). The psychophysics toolbox. Spatial Vision, 10, Brancucci, A., Dipinto, R., Mosesso, I., & Tommasi, L. (2009). Vowel identity between note labels confuses pitch identification in non-absolute pitch possessors. PLoS ONE, 4(7), e6327. Churchland, A. K., Kiani, R., & Shadlen, M. N. (2008). Decision-making with multiple alternatives. Nature Neuroscience, 11(6),

24 Absolute Pitch Memory 24 Cuddy, L. L. (1970). Training the absolute identification of pitch. Perception & Psychophysics, 8(5A), Deutsch, D. (2002). The puzzle of absolute pitch. Current Directions in Psychological Science, 11(6), Deutsch, D., Henthorn, T., Marvin, E., & Xu, H. (2006). Absolute pitch among American and Chinese conservatory students: Prevalence differences, and evidence for a speech-related critical period. Journal of Acoustical Society of America, 119(2), Drayna, D. T. (2007). Absolute pitch: a special group of ears. Proc Natl Acad Sci USA, 104(37), Gathercole, S. E., & Baddeley, A.,. D. (1990). Phonological memory deficits in language-disordered children: Is there a causal connection? Journal of Memory and Language, 29, Godden, D. R., & Baddeley, A. D. (1975). Context-dependent memory in two natural environments: On land and under water. British Journal of Psychology, 66, Halpern, A. R. (1989). Memory for the absolute pitch of familiar songs. Memory & Cognition, 17, Happe, F. (1999). Autism: cognitive deficit or cognitive style? Trends Cogn Sci, 3(6), Keenan, J. P, Thangaraj, V., Halpern, A. R., & Schlaug, G. (2001). Absolute pitch and planum temporale. NeuroImage, 14,

25 Absolute Pitch Memory 25 Lee, C.-Y., & Lee, Y.-F. (2010). Perception of musical pitch and lexical tones by Mandarin-speaking musicians. Journal of Acoustical Society of America, 127(1), Levitin, D. J. (1994). Absolute memory for musical pitch: evidence from the production of learned melodies. Percept Psychophys, 56(4), Levitin, D. J., & Rogers, S. E. (2005). Absolute pitch: perception, coding, and controbersies. Trends in Cognitive Sciences, 9(1), Miyazaki K. (1993) Absolute pitch as an inability: Identification of musical intervals in a tonal context. Music Perception, 11(1), Niwa, M., & Ditterich, J. (2008). Perceptual decisions between multiple directions of visual motion. The Journal of Neuroscience, 28(17), Pelli, D. G. (1997). The videotoolbox software for visual psychophysics: transforming numbers into movies. Spatial Vision, 10, Sacks, O. (1995). Musical ability. Science, 268(5211), Saffran, J. R., & Griepentrog, G. J. (2001). Absolute pitch in infant auditory learning: evidence for developmental reorganization. Dev Psychol, 37(1), Schellenberg, E. G., & Trehub, S. E. (2003). Good pitch memory is widespread. Psychol Sci, 14(3), Schlaug, G., Jancke, L., Huang, Y., & Steinmetz, H. (1995). In vivo evidence of structural brain asymmetry in musicians. Science, 267(5198), Smith, S. M., & Vela, E. (2001). Environmental context-dependent memory: a review and meta-analysis. Psychon Bull Rev, 8(2), Takeuchi, A. H., & Hulse, S. H. (1993). Absolute pitch. Psychol Bull, 113(2),

26 Absolute Pitch Memory 26 Vanzella, P., & Schellenberg, E. G. (2010). Absolute pitch: effects of timbre on notenaming ability. PLoS ONE, 5(11), e Ward, W. D. (1999). Absolute Pitch. In D. Deutsch (Ed.), The Psychology of Music (pp ). San Diego, California: Academic Press. Wong, A. C.-N., Palmeri, T. P., & Gauthier, I. (2009). Conditions for facelike expertise with objects: Becoming a ziggerin expert - but which type? Psychological Science, 20(9), Wong, Y. K., Folstein, J. R., & Gauthier, I. (2011). Task-irrelevant perceptual expertise. Journal of Vision, 11(14:3), Wong, Y. K., & Gauthier, I. (2010). A multimodal neural network recruited by expertise with musical notation. J Cogn Neurosci, 22(4), Zatorre, R., & Beckett, C. (1989). Multiple coding strategies in the retention of musical tones by possessors of absolute pitch. Memory & Cognition, 17, Zatorre, R. J. (2003). Absolute pitch: a model for understanding the influence of genes and development on neural and cognitive function. Nat Neurosci, 6(7), Zatorre, R. J., Chen, J. L., & Penhune, V. B. (2007). When the brain plays music: auditory-motor interactions in music perception and production. Nat Rev Neurosci, 8(7),

27 Absolute Pitch Memory 27 Figure 1. Box plots showing accuracy in the standard AP test and the verification test for violinists (A) and pianists (B), after excluding participants defined as AP possessors by their performance in the standard naming test. Y-axes show accuracy after correction for guessing (see Methods), with 0% and 100% representing chance and perfect performance respectively. Limits of the boxes represent 25th and 75th percentiles, the cross and solid line inside the box represent the mean and median respectively, ends of whiskers represent one standard deviation above and below the mean.

28 Absolute Pitch Memory 28 Figure 2. Verification performance for three context conditions. The y-axis shows sensitivity (see Methods), with chance performance at 0 and perfect performance at 1. The violinists and pianists were collapsed and then divided into four groups based on their performance in the standard naming test.