MELODIES PERFORMED WITH WESTERN

Transcription

1 Evaluation of Operatic Performances 489 LAY LISTENERS CAN EVALUATE THE PITCH ACCURACY OF OPERATIC VOICES PAULINE LARROUY-MAESTRI Max-Planck-Institute for Empirical Aesthetics, Frankfurt, Germany DOMINIQUE MORSOMME University of Liège, Liège, Belgium DAVID MAGIS Fonds de la Recherche Scientifique-FNRS, Liège, Belgium DAVID POEPPEL Max-Planck-Institute for Empirical Aesthetics, Frankfurt, Germany & New York University LAY LISTENERS ARE RELIABLE JUDGES WHEN evaluating pitch accuracy of occasional singers, suggesting that enculturation and laypersons perceptual abilities are sufficient to judge simple music material adequately. However, the definition of pitch accuracy in operatic performances is much more complex than in melodies performed by occasional singers. Furthermore, because listening to operatic performances is not a common activity, laypersons experience with this complicated acoustic signal is more limited. To address the question of music expertise in evaluating operatic singing voices, listeners without music training were compared with the music experts examined in a recent study (Larrouy-Maestri, Magis, & Morsomme, 2014a) and their ratings were modeled with regard to underlying acoustic variables of pitch accuracy. As expected, some participants lacked test-retest reliability in their judgments. However, listeners who used a consistent strategy relied on a definition of pitch accuracy that appears to overlap with the quantitative criteria used by music experts. Besides clarifying the role of music expertise in the evaluation of melodies, our findings show robust perceptual abilities in laypersons when listening to complex signals such as operatic performances. Received: November 24, 2015, accepted October 17, Key words: music expertise, pitch accuracy, melodic perception, singing voice, opera MELODIES PERFORMED WITH WESTERN operatic technique yield a complex acoustic signal (see Larrouy-Maestri, Magis, & Morsomme, 2014b; Sundberg, 2013, for reviews). Therefore it is unsurprising that the perception of pitch accuracy is not solely related to the information outlined in a musical score (i.e., the pitch relationship between tones), which is the main musical criterion used to estimate singing proficiency in occasional singers (see Dalla Bella, 2015, or Pfordresher & Larrouy-Maestri, 2015, for discussions on this topic). Operatic performances also include several acoustic parameters, such as vibrato or singer s formant, developed through operatic vocal training. In order to examine the construct of pitch accuracy in operatic voices, Larrouy-Maestri et al. (2014a) asked professional musicians to rate the pitch accuracy of Happy Birthday performances sung with an operatic singing technique. This study showed that pitch accuracy evaluation was not based exclusively on the precision of performed music intervals but on a complex combination of performance variables (i.e., tempo, fundamental frequency of the starting tone) and quality parameters (i.e., energy distribution, vibrato rate, and vibrato extent). For instance, large pitch interval deviations influenced the judges rating only when associated with the pitch height of the starting tone or with the tempo. The results confirmed the complexity of singing accuracy perception but also highlighted the high intra and interjudge reliability and the objectivity of the listeners (i.e., judgments relying mainly on the quantifiable variables mentioned above). Interestingly, these qualities (i.e., objectivity and consistency) are not exclusive to professionals with intense formal training. Recently, Larrouy-Maestri, Magis, Grabenhorst, and Morsomme (2015) observed that lay listeners behaved similarly when evaluating the pitch accuracy of melodies performed by occasional singers. This finding demonstrates perceptual abilities in lay listeners comparable to musically trained listeners and appropriate use of implicitly learned musical rules (e.g., Bigand & Poulincharronnat, 2006; Hannon & Trainor, 2007; Marmel, Tillmann, & Dowling, 2008). However, examining the evaluation of occasional singers might not Music Perception, VOLUME 34, ISSUE 4, PP , ISSN , ELECTRONIC ISSN BY THE REGENTS OF THE UNIVERSITY OF CALIFORNIA ALL RIGHTS RESERVED. PLEASE DIRECT ALL REQUESTS FOR PERMISSION TO PHOTOCOPY OR REPRODUCE ARTICLE CONTENT THROUGH THE UNIVERSITY OF CALIFORNIA PRESS S REPRINTS AND PERMISSIONS WEB PAGE, DOI:

2 490 Pauline Larrouy-Maestri, Dominique Morsomme, David Magis, & David Poeppel suffice to conclude that lay listeners are objective (i.e., relying on quantifiable definitions of pitch accuracy) and reliable (i.e., consistent). Indeed, more challenging conditions could lead to different results (Besson & Faïta, 1995) and therefore reflect the limits of either perceptual abilities or use of internalized musical rules when listening to complex auditory signals. Evaluating operatic performances is challenging for two main reasons. First, opera is not a popular music style. According to the National Endowment for the Arts (2008), only 2.1% of the adult population of the United States attend an opera performance per year. Although there are operatic voice contests (e.g., Queen Elisabeth International Music Competition of Belgium), such events are rare compared to the numerous casting shows and talent contests in the media. Therefore, one can infer that laypersons are not particularly familiar with this specific style and its acoustic features. For instance, the singer s formant consists of increasing the energy between 2 and 4 khz (Barnes, Davis, Oates, Chapman, 2004; Omori, Kacker, Carroll, Riley, & Blaugrund, 1996; Thorpe, Cala, Chapman, & Davis, 2001) and is quite typical for this vocal technique (Larrouy- Maestri et al., 2014b; Stone, Cleveland, Sundberg, & Prokop, 2002). Developed during training (Brown, Rothman, & Sapienza, 2000; Lundy, Roy, Casiano, Xue, & Evans, 2000; Omori et al., 1996), this feature is considered a vocal quality by singing teachers (Ekholm, Papagiannis, & Chagnon, 1998; Garnier, Henrich, Castellengo, Sotiropoulos, & Dubois, 2007) and allows singers to be heard over an orchestra without electronic amplification. Since most popular music styles make use of amplification, this acoustic parameter is expected to be quite unfamiliar to lay listeners. Note that other parameters such as vibrato also considered a vocal quality developed through training (Mürbe, Zahnert, Kuhlisch, & Sundberg, 2007) and appreciated in highly trained voices (Garnier et al., 2007) are not exclusive to opera but occur in other singing styles too. Lay listeners might thus be acquainted with this feature even if they are not opera listeners. Second, the definition of pitch accuracy in operatic singing voices is based on the interaction of several parameters (Larrouy-Maestri et al., 2014a), in contrast with the definition of pitch accuracy of melodies performed by occasional singers (Larrouy-Maestri, Lévêque, Schön, Giovanni, & Morsomme, 2013; Larrouy-Maestri et al., 2015). In addition, parameters such as the spectral composition of the sound (e.g., Hutchins, Roquet, & Peretz, 2012; Russo & Thompson, 2005; Vurma, Raju, Kuuda, 2010; Warrier & Zatorre, 2002) or vibrato (van Besouw, Brereton, & Howard, 2008) influence pitch perception. The combination of these parameters in a single signal may enhance the complexity of the task for laypersons and challenge their perceptual abilities, which are typically deemed weaker than music experts (e.g., Kraus & Chandrasekaran, 2010; Micheyl, Delhommeau, Perrot, & Oxenham, 2006; Schellenberg & Weiss, 2013; Tervaniemi, Just, Koelsch, Widmann, & Schroger, 2005; Varnet, Wang, Peter, Meunier, & Hoen, 2015; but see Larrouy- Maestri et al., 2015; Vanden Bosch der Nederlanden, Hannon, & Snyder, 2015, for contradictory findings). To investigate whether music expertise is a prerequisite for objective and consistent evaluation of pitch accuracy in challenging conditions, we asked lay listeners to evaluate, in a test-retest paradigm, the pitch accuracy of operatic singers previously analyzed in Larrouy-Maestri et al. (2014a); we then modeled their performance with regard to the acoustic predictors of pitch accuracy definition and to the music experts data of the reference study (Larrouy-Maestri et al., 2014a). Method PARTICIPANTS Twenty-two participants were paired by gender (8 women) and age (M ¼ years, SD ¼ 11.64) with the expert listeners of the reference study (Larrouy- Maestri et al., 2014a). They were recruited in Belgium and France on the basis of an extended questionnaire. None mentioned possessing absolute pitch, they reported no history of choral singing, no history of formal music training (or a maximum of 2 years of training and no practice during the past 5 years) and no particular affinity for music (attending less than one concert a week and actively listening to music less than two hours a day). We applied additional inclusion criteria before enrolling participants in the present study: (a) bilateral hearing threshold of 20 db SPL at 500, 1000, 2000, and 4000 Hz, screened with pure tone audiometry (Madsen Xeta, GN Otometrics, Denmark), (b) no congenital amusia (tested with the Montreal Battery of Evaluation of Amusia of Peretz, Champod, & Hyde, 2003), and (c) the ability to perform the song Happy Birthday with the appropriate melodic contour. MATERIAL The material consisted of 14 performances from the corpus (last sentence of a cappella Happy Birthday performed with an operatic singing technique by 14 female professional singers, from 21 to 66 years old, M ¼ 34.86, SD ¼ 12.72). The performers reported on average years of formal vocal training (SD ¼ 3.58), regular singing practice, and

3 Evaluation of Operatic Performances 491 TABLE 1. Mean (SD), Minimum, and Maximum of the Acoustic Parameters for the 14 Singing Performances Evaluated Acoustic parameters Mean (SD) Minimum Maximum Pitch interval deviation (cents) (35.72) Tempo (bpm) (6.96) F0 (Hertz) (79.51) Energy distribution (ratio) 1.69 (0.28) Vibrato rate (Hertz) 5.77 (0.55) Vibrato extent (cents) (44.19) Note: All the acoustical data analyses were performed on a Macintosh (Mac OS X, Version ) with AudioSculpt 2.9.4v3 and OpenMusic 6.3 softwares (IRCAM, Paris, France) using a Short Time Fourier Transform (STFT) analysis. The pitch interval deviation measurement was computed with regard to equal temperament and reported as a continuous variable. Absolute values of the deviations between the non-repeated tones of the short melodies were preferred to signed values in order to avoid the cancellation of enlarged/compressed intervals and therefore the underrepresentation of pitch deviations in trained singers. Note that energy distribution, vibrato rate and extent were quantified on the basis of the longest tone (i.e., last tone) of the last sentence of Happy Birthday (length from 1.13 to 1.98 seconds, M ¼ 1.45, SD ¼ 0.09) in order to get enough acoustic information for adequate measurements. public solo performances in classical style at the time of the recording. As summarized in the table, each performance was analyzed with respect to pitch interval deviation (mean absolute value of the deviation between interval performed and expected, in cents), tempo (in beats per minute), F0 of the starting tone (in Hertz), energy distribution (ratio between the khz band and the khz band, associated with the presence of singers formant), vibrato rate (number of quasiperiodic modulations of the F0 per second, in hertz) and extent (amplitude of the F0 variations within the same tone, in cents). Note that the pitch interval deviation measurement provides an adequate representation of deviations (i.e., great variability in Table 1) over short melodies (i.e., not long enough to measure the precision over time or the respect of the tonal center). Individual values for each performance can be found in the reference paper (Larrouy-Maestri et al., 2014a). PROCEDURE As in the reference study, the 14 selected performances (average duration of a sequence: M ¼ 6.29 s, SD ¼ 0.76) were rated using a pairwise comparison paradigm. Ninety-one pairs of performances (i.e., N*(N-1)/2) were presented via headphones (K271 MKII, AKG, Vienna, Austria) at a comfortable loudness level. For each pair, participants pushed buttons on a graphical user interface to listen to each sequence, as often as they wished, and at their own pace, and indicated which performance in a pair was more in tune, considering only pitch accuracy as a parameter and not by judging the quality of the voice. One point was attributed to the melodic sequence judged to be more in tune, zero points to the other sequence, and 0.5 points to both sequences when the pair was evaluated as being equally in tune. This way, each judge classified the 14 singers excerpts ranging from least in tune to most in tune. A test-retest procedure was applied after 8 to 15 days (M ¼ 9.32 days). ANALYSIS To examine intrajudge reliability, the 14 performances were ranked from worst to best, separately by each nonexpert judge, at two timepoints (test and retest). Then, rank differences (test minus retest) were computed for each performance, and the sample variance of differences was computed for each judge. This yields 22 sample variances, each variance being computed on 14 rank differences. Large variances indicate large dispersion in rank differences, and thus large variability in the rankings on test and retest and consequently low intrajudge reliability. Cut-off values for sampling variances in reliability analyzes were similar to those derived from expert judges of the reference study (Larrouy-Maestri et al., 2014a): Variances were compared to a chi-square distribution by selecting the degrees of freedom that maximized the p value of a Kolmogorov-Smirnov test. Then, a threshold was set as the quantile of the chisquared distribution with a lower tail probability of.95. A similar process was used to measure interjudge reliability, but using only data of the first test, disregarding the retest. Sample variances were computed for each singer on the set of rankings derived from each nonexpert judge. Low variance indicates low variability in the judges rankings of the singers, and thus good interjudge reliability. In order to compare lay listeners and professional musicians definition of pitch accuracy, expert and nonexpert judges ratings were modeled by combining the currently analyzed data from nonexpert judges with the data collected in the reference paper, Larrouy-Maestri et al. (2014a). In this multiple linear model with judges ratings as independent variables, all acoustic factors (pitch interval deviation, tempo, F0 of the starting tone,

4 492 Pauline Larrouy-Maestri, Dominique Morsomme, David Magis, & David Poeppel A B Variance of rank differences Variance of ranks of singer performance Judge Singer FIGURE 1. Intra and interjudge reliability. The left panel (A) illustrates the variance of rank differences (test minus retest) of the nonexpert judges (N ¼ 22). The right panel (B) highlights the variances of ranks of singers performances, using only nonexpert judges not flagged with intrareliability issues (N ¼ 16). Dashed horizontal lines depict the threshold for music experts in the reference study (Larrouy-Maestri et al., 2014a). energy distribution, vibrato rate and vibrato extent) as well as a group effect (with categories expert and nonexpert ) were included as covariates. All possible main effects and pairwise interactions were set up in the complete model, while the effect of a single judge was not modeled. The complete model was simplified by removing all nonsignificant terms, leading to the simplest model retained for analysis. Significant effects and their direction were assessed by Wald tests with appropriate contrast matrices across model parameters and limiting chi-square distributions (Rao, 1973). Note that in this analysis the Wald statistics are referred to as Q statistics, due to the use of contrast matrices. Results and Discussion INTRA AND INTERJUDGE RELIABILITY IN LAY LISTENERS Six of the 22 nonexpert participants exhibited high variability in their rank differences and exceeded the expert-based threshold (Figure 1A). This indicates weaker intrajudge reliability in the group of nonexpert judges. Note that the music experts in Larrouy-Maestri et al. (2014a) did also not show perfect intrajudge reliability (e.g., two of them seemed to change their rating strategy between test and retest). However, judges who use a similar strategy on both test and retest evidently share a similar strategy with the other judges. Indeed, Figure 1B shows that all sample variances are below the limit threshold derived from the reference study, indicating good overall interjudge reliability in the nonexpert group. Thus, most of laypersons are capable to develop a common construct of pitch accuracy despite the complexity of the signal and despite their limited experience as listeners to this specific style of singing. LAY LISTENERS VERSUS PROFESSIONAL MUSICIANS DEFINITION OF PITCH ACCURACY The final model calculated to explain the judges ratings has an R 2 coefficient of.53, very close to the coefficient of the full model (.532), meaning that the acoustic variables under study explained more than half of the variance of the ratings. The final model (illustrated in Figure 2) can be interpreted as follows. First, all covariates are included in the model, but none of them as simple main effect. Judges ratings cannot be fully described with individual covariates but with a complex interaction structure instead. Second, several covariates partly explain the ratings as pairwise interactions, though such interactions are independent of the particular group of judges (experts or nonexperts). All those pairwise interactions involve the tempo on one hand, and on the other hand either pitch interval deviation, Q(1) ¼ 7.87, p ¼.005, F0 of the starting tone, Q(1) ¼ , p ¼ <.001, vibrato rate, Q(1) ¼ , p <.001, or vibrato extent, Q(1) ¼ 41.49, p <.001. Note that the limited contribution of the pitch interval deviation parameter in the model could not be attributed to low variability among singers (see Table 1).

5 Evaluation of Operatic Performances 493 Pitch interval deviation Vibrato extent experts only laypersons only Tempo Vibrato rate stronger for experts Energy distribution F0 of the starting tone FIGURE 2. Illustration of the pairwise interactions predicting the increase of judges rating of pitch accuracy. Solid lines represent the pairwise interactions of covariates, which have similar effects for both groups of judges. Dashed lines represent the pairwise interactions of covariates that have different effects according to the respective group of judges (lay versus expert listeners). Equal signs stand for similar value of the parameter whereas positive symbols correspond to an increase in the value of the parameter and negative symbols a decrease in the value of the parameter. The direction of the distinct expertise effects is specified in grey text. Directional interpretations of these pairwise interactions are as follows (increase/decrease is highlighted in Figure 2 with positive and negative symbols): At a given level of pitch interval deviation, ratings will increase with a decrease of tempo; while at a given level of either F0 of the starting tone, vibrato rate or vibrato extent, ratings will increase with an increase of tempo. The overall results highlight the complexity of the definition for experts and nonexperts as well as the large overlap between the two groups, as illustrated by the solid lines in Figure 2. In addition to the relative reliability pointed out in the previous analysis, lay listeners seem to use criteria that are also important to the music experts, despite the complexity of the signal to be evaluated and the unfamiliar character of operatic performances for these participants. Finally, several pairwise interactions of covariates have different effects according to the respective group of judges, as shown by the dashed lines in Figure 2. These pairwise interactions involve: Tempo and energy distribution, Q(1) ¼ 25.47, p <.001, pitch interval deviation and F0 of the starting tone, Q(1) ¼ 9.93, p ¼.003, and pitch interval deviation and energy, Q(1) ¼ 31.94, p <.001. Concretely, at given levels of energy, judges ratings increase with tempo, both for nonexpert judges, Q(1) ¼ 77.48, p <.001, and for expert judges, Q(1) ¼ , p <.001, but the increase is more pronounced for experts than for nonexperts. Also, at given levels of pitch interval deviation, ratings increase with F0 of the starting tone in expert judges, Q(1) ¼ 9.42, p ¼.002, but remain constant in nonexpert judges, Q(1) ¼ 0.54, p ¼.46. In addition, at a given level of pitch interval deviation, ratings increase with a decrease in the energy distribution measure for nonexpert judges, Q(1) ¼ 27.69, p ¼.002, but remain constant for expert judges, Q(1) ¼ 2.95, p ¼.09. It is tempting to attribute the differences occurring between the two groups of judges to weaker perceptual abilities in nonmusicians. For instance, the fact that lay listeners give better ratings to lower energy distribution for an equal pitch interval deviation value could be explained by a greater influence of timbre (partly explained by the energy distribution in a signal) on pitch perception (e.g., Hutchins et al., 2012; Russo & Thompson, 2005; Vurma et al., 2010; Warrier & Zatorre, 2002). However, lower pitch perception abilities would also be noticeable due to an expertise effect on other interactions including vibrato rate or extent or tempo, which are also known to affect pitch perception (Prince, 2011; van Besouw et al., 2008). The model here suggests that interactions including these parameters do not differ between the two groups, which weakens the explanation of expertise effects by lower perceptual abilities in laymen, at least as far as listening to operatic voices is concerned. Further studies with materials allowing multiple parameters and their interactions to be controlled are necessary to fully eliminate this explanation. Nevertheless, the present findings already support that the differences observed in our model could be interpreted as subtle differences in the definition of pitch accuracy, as a consequence of specific knowledge of operatic singing performances. Indeed, specific knowledge regarding the technical difficulties of operatic performances could explain the relevance of the interactions pitch interval deviation - F0 of the starting tones (better ratings when the F0 is high, for similar pitch interval deviation) and energy tempo (better ratings when the tempo is fast, for similar energy distribution) for experts. Singer s formant (estimated through the energy measurement) occurs on vocalic sounds and, as fast tempo shortens vowels, music experts would acknowledge the difficulty

6 494 Pauline Larrouy-Maestri, Dominique Morsomme, David Magis, & David Poeppel of singing faster while maintaining energy distribution in the signal. As a consequence, they would grant the performers higher ratings. A similar explanation seems plausible regarding the interaction pitch interval deviation - F0 of the starting tones. Indeed, the development of the vocal range is a key aspect in vocal training, and starting the melody on a high pitch could be seen as a special vocal effort that requires both breath support and control of laryngeal muscles. Here again, specific knowledge about vocal technique itself could lead music experts to incorporate the vocal effort required to sing in a high range for similar precision of the pitch intervals. This would explain why they rated these performances better than lay listeners did. Moreover, less knowledge of laypersons could explain the interaction pitch interval deviation - energy distribution. Since laypersons have limited exposure to operatic voices and knowledge about the performance settings (i.e., no amplification), they would show preference for what they simply know and interpretation of low energy distribution as more adequate for similar interval precision in the signal. Interestingly, our model did not show differences between music experts and nonexperts regarding the contribution of the vibrato rate and vibrato extent parameters. It might be the case that the effect of formal music expertise on music perception is exclusively limited to acoustic parameters that lay listeners simply do not know. The differences between lay and professional listeners seem more attributable to the accrual of domain knowledge about vocal technique than to differences in perceptual abilities or different definitions of pitch accuracy. Conclusion We examined the ability of listeners to evaluate pitch accuracy of complex signals. We applied the approach described in Larrouy-Maestri et al. (2014a) to matched nonexperts. The results demonstrate surprising consistency and objectivity of the listeners despite the complexity of the signal evaluated. Whereas lay listeners make trustworthy judges when evaluating melodies performed by their peers (Larrouy-Maestri et al., 2015), here we observed two different profiles among the lay listeners: listeners who cannot maintain a consistent strategy in evaluating complex voice versus listeners who behave similar to musicians (cf. Larrouy-Maestri et al., 2014a), although they lacked the technical expectations due to the absence of specific knowledge about this singing technique. Therefore, the role of formal music training seems limited when evaluating technical features (e.g., pitch accuracy in the present study). We suggest that using similar methods with a focus on aesthetic features might allow a better understanding of the role of formal music training in music appreciation. Author Note We thank the Centre Henri Pousseur in Liège, Laboratories of Images, Signals and Telecommunication Devices (LIST) in Brussels, and Felix Bernoully for technical support, Marion Nowak and Virginie Roig- Sanchis, for their help with the data collection, and Renan Marcello Vairo Nunes, Matthias Grabenhorst, Tina Roeske, and Wolf Schlotz for helpful comments on a previous version of the manuscript. Correspondence concerning this article should be addressed to Pauline Larrouy-Maestri, Neuroscience Department, Max-Planck-Institute for Empirical Aesthetics, 14 Grüneburgweg, Frankfurt am Main, Germany. pauline.larrouy-maestri@ aesthetics.mpg.de References BARNES, J. J., DAVIS, P., OATES, J.,& CHAPMAN, J. (2004). The relationship between professional operatic soprano voice and high range spectral energy. Journal of the Acoustical Society of America, 116, BESSON, M.,& FAÏTA, F. (1995). An event-related potential (ERP) study of musical expectancy: Comparison of musicians with nonmusicians. Journal of Experimental Psychology: Human Perception and Performance, 21, BIGAND, E.,&POULINCHARRONNAT, B. (2006). Are we experienced listeners? A review of the musical capacities that do not depend on formal musical training. Cognition, 100, BROWN, W.S.,ROTHMAN, H.B.,&SAPIENZA, C. M. (2000). Perceptual and acoustic study of professionally trained versus untrained voices. Journal of Voice, 14, DALLA BELLA, S. (2015). Defining poor-pitch singing: A problem of measurement and sensitivity. Music Perception, 32, EKHOLM, E.,PAPAGIANNIS, G.C.,&CHAGNON, F. P. (1998). Relating objective measurements to expert evaluation of voice quality in Western classical singing: Critical perceptual parameters. Journal of Voice, 12,

7 Evaluation of Operatic Performances 495 GARNIER, M.,HENRICH, N.,CASTELLENGO, M.,SOTIROPOULOS, D., & DUBOIS, D. (2007). Characterisation of voice quality in Western lyrical singing: From teachers judgements to acoustic descriptions. Journal of Interdisciplinary Music Studies, 1, HANNON, E.,& TRAINOR, L. (2007). Music acquisition: Effects of enculturation and formal training on development. Trends in Cognitive Sciences, 11, HUTCHINS, S.,ROQUET, C.,&PERETZ, I. (2012). The vocal generosity effect: How bad can your singing be? Music Perception, 30, KRAUS, N.,& CHANDRASEKARAN, B. (2010). Music training for the development of auditory skills. Nature Neuroscience, 11, LARROUY-MAESTRI, P.,LÉVÊQUE, Y.,SCHÖN, D.,GIOVANNI, A., & MORSOMME, D. (2013). The evaluation of singing voice accuracy: a comparison between subjective and objective methods. Journal of Voice, 27, e LARROUY-MAESTRI, P.,MAGIS, D.,GRABENHORST, M.,& MORSOMME, D. (2015). Layman versus professional musician: Who makes the better judge? PLoS ONE, 10, e LARROUY-MAESTRI, P.,MAGIS, D.,&MORSOMME, D. (2014a). The evaluation of vocal pitch accuracy: The case of operatic singing voices. Music Perception, 32, LARROUY-MAESTRI, P.,MAGIS, D.,&MORSOMME, D. (2014b). Effects of melody and technique on acoustical and musical features of western operatic singing voices. Journal of Voice, 28, LUNDY, D.S.,ROY, S.,CASIANO, R.R.,XUE, J.W.,&EVANS, J. (2000). Acoustic analysis of the singing and speaking voice in singing students. Journal of Voice, 14, MARMEL, F.,TILLMANN, B.,&DOWLING, W. J. (2008). Tonal expectations influence pitch perception. Perception and Psychophysics, 70, MICHEYL, C.,DELHOMMEAU, K.,PERROT, X.,&OXENHAM, A.J. (2006). Influence of musical and psychoacoustical training on pitch discrimination. Hearing Research, 219, MÜRBE, D.,ZAHNERT, T.,KUHLISCH, E.,&SUNDBERG, J. (2007). Effects of professional singing education on vocal vibrato: A longitudinal study. Journal of Voice, 21, OMORI, K.,KACKER, A.,CARROLL, L.M.,RILEY, W.D.,& BLAUGRUND, S. M. (1996). Singing power ratio: Quantitative evaluation of singing voice quality. Journal of Voice, 10, PERETZ, I., CHAMPOD, A.-S.,& HYDE, K. (2003). Varieties of musical disorders: The Montreal Battery of Evaluation of Amusia. Annals of the New York Academy of Siences, 999, PFORDRESHER, P.Q.,&LARROUY-MAESTRI, P. (2015). On drawing a line through the spectrogram: How do we understand deficits of vocal pitch imitation? Frontiers in Human Neuroscience, 9, PRINCE, J. B. (2011). The integration of stimulus dimensions in the perception of music. Quarterly Journal of Experimental Psychology, 64, RAO, C. R. (1973). Linear statistical inference and its applications (2nd ed.). New York: Wiley. RUSSO, F.A.,&THOMPSON, W. F. (2005). An interval size illusion: The influence of timbre on the perceived size of melodic intervals. Perception and Psychophysics, 67, SCHELLENBERG, E.G.,&W.WEISS, M. (2013). Music and cognitive abilities. In D. Deutsch (Ed.), The psychology of music (3rd ed., pp ). San Diego, CA: Elsevier. STONE, R.E.,CLEVELAND, T.F.,SUNDBERG, J.,&PROKOP, J. (2002). Aerodynamic and acoustical measures of speech, operatic, and Broadway vocal styles in a professional female singer. TMH-QPSR, 43, SUNDBERG, J. (2013). Perception of singing. In D. Deutsch (Ed.), The psychology of music (3rd ed., pp ). San Diego, CA: Elsevier. TERVANIEMI, M.,JUST, V.,KOELSCH, S.,WIDMANN, A.,& SCHRÖGER, E. (2005). Pitch discrimination accuracy in musicians vs nonmusicians: An event-related potential and behavioral study. Experimental Brain Research, 161, THORPE, C. W., CALA, S. J., CHAPMAN, J.,& DAVIS, P. J. (2001). Patterns of breath support in projection of the singing voice. Journal of Voice, 15, VAN BESOUW, R.M.V.,BRERETON, J.S.,&HOWARD, D.M. (2008). Range of tuning for tones with and without vibrato. Music Perception, 26, VANDEN BOSCH DER NEDERLANDEN, C.M.,HANNON, E.E.,& SNYDER, J. S. (2015). Everyday musical experience is sufficient to perceive the speech-to-song illusion. Journal of experimental psychology: General, 144, e VARNET, L.,WANG, T.,PETER, C.,MEUNIER, F.,&HOEN, M. (2015). How musical expertise shapes speech perception: Evidence from auditory classification images. Scientific Reports, 5, VURMA, A., RAJU, M.,& KUUDA, A. (2010). Does timbre affect pitch?: Estimations by musicians and non-musicians. Psychology of Music, 39, WARRIER, C.M.,&ZATORRE, R. J. (2002). Influence of tonal context and timbral variation on perception of pitch. Perception and Psychophysics, 64,