EXPRESSION OF PIANO TIMBRE: GESTURAL CONTROL, PERCEPTION AND VERBALIZATION

Transcription

1 EXPRESSION OF PIANO TIMBRE: GESTURAL CONTROL, PERCEPTION AND VERBALIZATION Michel Bernays, Caroline Traube Laboratoire Informatique, Acoustique, Musique (LIAM), Observatoire International de la Création et des Cultures Musicales (OICCM), Faculté de musique, Université de Montréal, Canada International Laboratory for Brain, Music and Sound Research (BRAMS), Montreal, Canada Centre for Interdisciplinary Research in Music Media and Technology (CIRMMT), Montreal, Canada ABSTRACT Timbre is a key to musical expressivity in virtuosic pianistic performance, discussed amongst professionals through an extensive, yet very abstract and emotional, vocabulary. This study first aims to determine the degree of consensus of this vocabulary among pianists. Furthermore, timbre s quantitative and functional characteristics are yet to be explored. We thus intend to identify timbre correlates in a performer s gesture on the keyboard. A professional pianist was asked to play three short pieces, with eight different timbres as successive instructions to colour the performances. The audio recordings were used as stimuli for a timbre identification task, in which 17 other pianists had to label the timbre with the descriptor they deemed most fitting. The results, more than three times above chance, were significantly improved by taking into account the semantic proximity ratings between descriptors. This is thus indicative of a semantic consistency in the perception and description of piano timbre. Then, from key position and hammer velocity data collected in the performances through the computer-controlled recording grand piano Bösendorfer CEUS, we computed numerous gesture features as cues of dynamics, articulation, synchronism, key touch and attack and pedal use, among which several were significantly correlated with the timbre performed. This is a step towards a comprehensive gestural mapping of piano timbre that could yield significant improvements in piano pedagogy and software modelization. 1. INTRODUCTION As music performance scientific studies have been extending ever farther along with technological progress, many expressive features in piano performance such as intensity, timing, fingerings or articulation were thoroughly and systematically explored by Repp (1996; 1997), Parncutt & co. (Clarke, Parncutt, Raekallio & Sloboda, 1997; Parncutt, Sloboda & Clarke, 1999; Parncutt & Troup, 2002), and Goebl (2003; Goebl, Bresin & Galembo, 2005; Goebl, Flossmann & Widmer, 2009) among others, thus updating the seminal works of Ortmann (1929/1962) or Seashore (1936; 1937). Thanks to them, we now better understand the role of gesture in expressive piano performance. However, among the musical features most essential in piano performance, timbre has mostly been left out, as this complex and multidimensional parameter still lacks a proper clear-cut, quantified and functional definition. Manifold timbre perceptual spaces were built and their dimensions' acoustical correlates identified (Grey, 1977; McAdams, Winsberg, Donnadieu, De Soete & Krimphoff, 1995; Hajda, Kendall, Carterette & Harshberger, 1997). However such studies only accounted for timbre as the cue of instruments' identities. Timbre as tone-quality (Marozeau, 2004), which the instrumentalist can control, has not been examined in these studies. Still, timbre's musical relevance is set alight within the piano treatises of master teachers, who intend to guide the pianist in finding his sound with their empiric knowledge of the most adequate gesture the pianist ought to use to create those sonic nuances from which the emotions can arise. Notorious twentieth century masters such as Matthay (1932), Neuhaus (1958/1971), Kochevitsky (1967) highlighted the importance of gesture in developing one s proper sound. Indeed, virtuosic pianists know how to use many subtleties of gesture and articulation between notes, within a chord, arpeggio or melodic line in order to colour their performances in timbre, in a much finer way than the sole intensity of key stroke to which piano timbre control is often thought to be contrived by the mechanical constraints of a single, isolated key. Articulation is indeed required to create what we call aggregate timbres: a combination of at least two sonic elements into one resulting auditory object where articulation the relation between one note to the next is of prime importance; or, as Heinrich Neuhaus puts it, sound molecules emerging from atom-notes (Neuhaus, 1958/1971, p.67). As a consequence of this advanced control, timbre happens to be a meaningful and well-defined parameter amongst high level performers, who have developed over the years of practice a precise and refined motor control of the instrument, as well as an acute perceptive sensibility to slight sonic variations. This fine perception of timbre results in an extensive vocabulary built to describe all the nuances a performer can detect. However, the piano learning process, at the professional level, is essentially devoted to the transmission, from master to student, of abstract concepts through subjective and metaphorical verbal descriptions, most often in the form of adjectives such as bright, harsh, shimmering, velvety Timbre is thus directly linked to the emotion it ought to instill, thus bypassing its concrete ways of production through instrumental gesture. Yet, despite no such objective sources to rely on, pianists seem able to avoid inter-individual misunderstandings in timbre description. Few studies have been devoted to such timbre verbalization. Von Bismarck (1974), Faure (2000), Stepanek (2006) had sound stimuli matched to timbre verbalizations either over bipolar or PREPRESS PROOF FILE 1 CAUSAL PRODUCTIONS

2 unipolar scales, or by free description and corpus analysis and tried to fit the resulting semantic dimensions to perceptual spaces, with relative success. Cheminée, Gherghinoiu & Besnainou (2005) studied free verbalization of piano sounds. Semantic analysis revealed the specificity of the pianists lexicon (as a consensual linguistic subset) built upon an affective and axiological vocabulary following two axes: percussion and resonance. Bellemare, in her study of piano timbre verbalization (Bellemare & Traube 2006), interviewed eight professional pianists and asked them to choose ten common piano timbre descriptors, their synonyms and antonyms, and their process of technical production. Close to a hundred timbre-descriptive adjectives were thus gathered. In the next series of interviews, the twenty most salient descriptors were defined and organized by semantic and technical similarity. The piano timbre verbalization side of this study intends to follow in these footsteps. As for the gestural control of piano timbre, despite the tremendous empiric awareness of timbre high-level pianists elicit, they often have no clear and assertive consciousness of its producing gesture. Only Ortmann (1929/1962) delved into the relations between piano timbre verbal descriptors and characteristics of touch, and solely for a single note hardly a real musical context. There is a lack of literature about timbre verbalization and its gestural expression in piano performance. Thus relations between the concept of timbre as expressed through those verbal descriptions and its gestural control in performance have yet to be explored. 2. AIMS The study thus aims to elicit the link between piano timbre and the gestural control of its production, with an approach that is mainly three-fold. The first step focuses on the verbal description of piano timbre, its typology and underlying metaphorical and technical meaning. The semantic relationships between descriptors were explored in order to build a semantic space of piano timbre verbal description. We then aimed at determining whether those verbal descriptions were consensually agreed upon and remained consistent from piano performance to the listening experience. A timbre identification and labelling experiment was set up to test a group of pianists' ability to identify timbre in audio recordings of controlled performances each of which was played following a precise timbre instruction. From those timbre-featured performances, we were able to collect key motion and hammer velocity data, thanks to the computer-controlled acoustic grand piano Bösendorfer CEUS and its high-precision optical sensors. From those raw, mechanical data, we defined higher-level features that could describe instrumental gesture on the keyboard in a musically relevant fashion as cues of dynamics, articulation, synchronism, key touch and attack and pedal use. Those features and their values are then to be correlated with the timbre performed. 3. METHOD As the central task in the experimental process, a professional pianist was asked to perform short musical pieces multiple times with different timbre instructions given by specific verbal descriptors. Prior to this, our exploratory study required a few preliminary steps. First, we refered to (Bellemare & Traube, 2006) and selected the eight most common and salient piano timbre descriptors. A local composer (student at Université de Montréal) then wrote three short original pieces, which could in all be fitted to those eight timbres. The attribution of timbres was as follows: First piece: bright, round, distant, harsh, dark. Second piece: full-bodied, round, harsh, dark. Third piece: shimmering, matte. Three was the minimal number of pieces required to illustrate those eight timbre nuances. Cross-comparisons of the round, harsh and dark timbres are nonetheless enabled between the first and second pieces. The pieces would last roughly twenty, thirty and fifteen seconds respectively. 3.1 Performances recordings These pieces were performed on the Bösendorfer CEUS piano, in the BRAMS facilities, by a professional pianist, doctoral graduate in piano performance from Université de Montréal (designated by MG in the rest of this article). Thirty-four performances were recorded during four sessions. Raw data of keys and pedals positions and maximum hammer velocities were collected through the piano's CEUS system, with the highest accuracy of any recording piano both in time (2ms time steps) and range (250 steps, i.e. a 32µm accuracy for keys position). Audio recordings were gathered as well, with the setup displayed in Figure 1. More details on the sound takes can be found in (Bernays & Traube, 2010). Figure 1: Recording apparatus 2

3 3.2 Timbre identification test The performer himself was submitted right after the last recording session to the timbre identification task: labelling the timbre in the audio recordings of his performances. The same task was assigned to seventeen pianists in a collective listening test of thirteen selected audio excerpts. Among the participants, sixteen were currently related to Université de Montréal, and eleven were Quebec natives, which we acknowledge covers only a limited portion of the worldwide pianistic community. Three successive listening rounds consisted in, respectively, preliminary familiarization with the stimuli, a first random-order presentation with the assignment of writing down a free verbal description of the timbre, and a second random-order presentation of the audio stimuli with forced choice between the eight timbre descriptors possible. Additionally, in order to perform a semantic analysis of the identification results, we asked afterwards the 17 participants to fill in an online form with their evaluation of the pairwise semantic proximity between descriptors. Ratings were given out on six-step, zero to five Likert scales, with each step associated to an evaluative, unambiguous term (from very different to very close). Those ratings could then be used as a weighting system to the timbre identification results. 3.3 Gesture data processing Gestural information contained in the raw data collected as mentioned in paragraph 3.1 is restricted to gesture expression at the keyboard level, as strictly transferred to the keys. This discards the ancillary components of gesture pianists elicit through complex patterns of hand, arm and body movements aimed at pressing the keys in the way musically intended. By focusing on gesture as seen through the keyboard actions, we wish to identify the features paramount to obtaining a precise timbre nuance, regardless of a performer s gesture idiosyncrasies. The dataset was processed through a series of Matlab functions, aimed at regaining the most musical information possible out of it. The first relevant items extracted were the consecutive, non-zero key depression occurrences, which were grouped as notes. For each of these notes, several physical, mid-level gestural features were computed such as their mean and maximum key depression, sknewness of depression profile, corresponding maximum hammer velocity (MHV) and pedals use, delays between note onset and MHV or maximum key depression instants as a reduction of the complete information available. Notes were then regrouped according to their onsets synchronism falling into a 50ms window. These new items were labelled elements. Existing note features were recomputed as extrema, means, deviations and distributions per element, and within-element features of overlap and synchronism between notes were added. For each performance, overlaps and inter-onset intervals between elements were calculated. Additionally, as elements cover the whole register, they were cut by separating notes farther than one octave. This simple heuristic is aimed at representing the repartition of notes between hands (no ninth chords were included in the pieces). And to sum it all up, each performance was described by statistics of means, rates and standard deviations between elements. In all, 164 statistics were obtained (times two with the addition of octave-segregated elements statistics). Their computation was either inspired by literature on piano gesture and articulation, or emerged from the novel information provided by the CEUS system which we tried and exploit at the utmost. The 27 timbre-coloured-performances dataset thus compiled once the outliers discarded was pre-processed to enable between-pieces comparisons, then submitted to exhaustive statistical analyses of variance, means, rank, and linear or monotonous, non-parametric correlations. Timbre was used as factor or grouping variable. This process was first performed over the whole dataset, then over each of the 28 pairs of timbres (this is intended to compensate for the lack of robustness of the ANOVA post-hoc comparisons in this particular design). 4. RESULTS 4.1 Timbre identification First, the performer himself managed to identify the timbre in the audio recordings almost flawlessly (but for his confusing the terms round and full-bodied). This implies his conception of timbre is consistent between performance and listening. In the main timbre identification test, the overall success rate of the participants in the forced-choice task was 38.3%. While this doesn t elicit a complete consensus on timbre verbal identification among pianists, this is still more than three times above chance level (12,5%), with a high significance (p<<0.01) and thus indicates a similar verbal identification pattern among the participants. There is a large discrepancy between participants (between 2 and 11 right answers upon 13 timbres, mean 4.82, SD 2.13 i.e. 16.4%). The success rate also varies considerably with timbre (from 0 to 11 good answers over 17 participants, mean 6.3, SD 3.84 i.e. 22.5%). Indeed, the round timbre could not be identified in either of the excerpts in which it appeared, whereas harsh (53% over 3 presentations) and bright (47% over 2 presentations) timbres were easier to label. As for the free description task, the chosen terms semantic fields tend to echo the expected answer s (although some excerpts yield more ambiguous or antonymous terms). Those timbre identification, forced-choice test s results were then processed with regard to the semantic proximity between descriptors (evaluated by the participants as explained in paragraph 3.2). With timbre identification answers rescaled between +1 (expected descriptor) and -1 (antonym), chance level was fixed at The answers yielded an overall semantic accuracy of 0.39, which is much more significant (p<<0.001) and way above chance. This suggests, if not a complete consensus over piano timbre verbal description, at least a certain consistency between participants regarding the semantic meaning attached to piano timbre upon listening. Timbre identification results are further elicited in (Bernays & Traube, 2010). 3

4 4.2 Gesture data analysis The full-dataset analysis revealed repeatedly significant (at the 5% threshold) correlations between timbre and 31 features. From those physical, piano mechanism-related characteristics, we extracted the musical meanings, with regards to the explanations found in the piano action literature. This process was facilitated by the fact the features were originally chosen for computation with their possible musical interpretation in mind. In the end, this resulted in 28 timbre-discriminative musical attributes, split into seven categories, and listed in Table 1. Musical Musical attributes domain Key dynamics consistency between elements' notes Dynamics Variations in the register/hand repartition of dynamics Legato/ Degree of legato or staccato play staccato Variations in the degree of legato or staccato play Speed of attack homogeneity between elements' notes Variations in the register/hand repartition of the speed of key attacks Central voice emphasis within hand of the sharpness of key attacks Key attack Sharpness of attack homogeneity between elements' notes Register/hand repartition within elements of the sharpness of key attacks Variations in the register/hand repartition of the sharpness of key attacks Variations in notes durations repartition Relative duration of notes within elements Repartition of key depressions in hand/register Key elements depress Variations in the repartition of types of touch profiles Variations in the lengths of decays within elements Length of decay Core synchronism of notes within elements Offsets synchronism within hand elements Note Offsets timing consistency within elements offsets Offsets synchronism consistency within hand timing Variations in offsets timing consistency Offsets timing within elements Onsets timing consistency Note Onset synchronism of notes within elements onsets timing Variations of the onset synchronism of notes within elements Use of the forte/resonance pedal Pedals Variation in the use of the soft pedal Variation in the use of the forte/resonance pedal Table 1: Timbre-discriminative musical attributes as set alight by statistical analysis of 27 performances illustrating 8 timbres. Through timbre pairwise analysis, several features were highlighted as differentiating one timbre from another, and their value assessed as higher or lower. From the confusion matrix thus established, the recurring features that could separate one timbre from every other were selected. Each of the eight timbres was thus described with some distinctive features set at a peculiarly high or low value. Musical attributes are displayed in Table 2. Timbre Musical features Key dynamics disparity in elements and registers Consistent speed and sharpness of key attacks in Bright elements and between registers Synchronous offsets within elements Heavy use of the forte/resonance pedal Consistent key dynamics in elements and registers Consistent key depressions profiles within elements Dark Long decays Consistent offsets synchronism within elements Consistently synchronous onsets within elements Key dynamics consistency in elements and registers Legato play Slow, disparate speeds and sharpnesses of key attacks in elements and between registers Emphasized sharpness of attack in central voice Distant Disparate note durations in elements and registers Long, consistent decays Disparate repartitions of the types of touch Asynchronous and disparate offsets Synchronous onsets within elements Heavy and monotonous use of the soft pedal Disparate sharpnesses of attacks in elements and Fullbetween registers bodied Somewhat legato play Key dynamics disparity between registers Low, disparate core synchronism of notes within elements Harsh Disparate key depressions profiles within elements Disparate, asynchronous offsets within elements Consistent onset timing within elements Light use of the forte/resonance pedal Sharper key attacks in the low register Matte Consistent note duration repartition in registers Heavy use of the soft pedal Asynchronous offsets within elements Round Heavy use of the forte/resonance pedal Shimm Key dynamics disparity in elements and registers -ering Heavy & monotonous use of the forte/resonance pedal Table 2: Timbres gesture profile based on timbre pair-wise analysis and selection of recurring features. 4

5 Some additional, interesting remarks can arise from an in-depth scrutiny. First, he distant timbre can be set apart by many, and much more salient, features; this was the slowest and softest timbre performed; however, tempo and intensity revealed no significant correlation with timbre overall, and the selected features were pre-processed so as not to depend directly on tempo and intensity; the possibility of any indirect link has yet to be explored. On the other hand, full-bodied and round timbres especially were assigned few and mid-valued discriminative features, which concurs with the idea exposed in (Bellemare & Traube, 2006) that those timbres are somewhat medium, with no salient acoustic or perceptive feature. Even more telling is the fact that no gestural feature could set full-bodied and round timbres apart. They were also the most difficult timbres to identify in the perception test, and the performer himself could hardly hear a difference. As for the matte and shimmering timbres, there are few, but quite salient discriminative features. Finally, timbre pairwise analysis highlighted the same relevant features as the full-dataset analysis, with a few timbre-specific additions, most notably the sheer (slow) speed and sharpness of key attacks for a distant timbre, and the (heavy) use of the soft pedal for the matte and distant timbres. Within the limits of this exploratory study, we could in the end single out several gesture descriptors spreading over seven general aspects of piano playing, and through which every timbre performed (amongst the eight analyzed) can be identified. We could thus obtain in this context a gestural mapping of piano timbre. 5. DISCUSSION We obtained significant indications of the ways of expression pianists use in their gesture on the keyboard to control piano timbre. However, as an exploratory study, the results shall not be generalized since, first, the sample size of timbre-coloured performances is yet too small. While the direct influence of tempo, intensity and musical discrepancy between pieces were eliminated by data pre-processing, side effects could have lingered. Furthermore, the assumed musical correlates of the gesture features will be tested experimentally. We will indeed verify whether performance instructions of such musical attributes yield the expected effects on the computed features. We will also use those timbre-discriminative features to modelize timbre-coloured performances, and test their perceptive timbre validity. Besides, we will use new shorter pieces from now on, so that timbre consistency is ensured throughout, and compare several pianists performances, in order to dismiss individual expressive idiosyncrasies. New timbres to examine will be chosen: in a large-scale evaluation test of timbre descriptors semantic proximity, the fourteen most common descriptors were submitted to many pianists, who provided pairwise semantic similarity ratings. Those ratings will be used in building a semantic space through multidimensional scaling. The minimal number of descriptors that can define all this space s dimensions will then be selected, which shall give out the most diverse and encompassing set of piano timbre descriptors. In all, we wish to get a better knowledge of pianists expressivity at the timbre level, and especially of the relations between words and gesture for piano timbre expression. Such findings may yield the development of new didactic methods, wherein indications of the precise gesture required to produce the adequate timbre nuance would complement its abstract, metaphoric expression traditionally employed. Furthermore, gesture as well as verbalization results could be applied in virtual piano software, as a modeling parameter for the playing style/timbre color, and as instructions in user interfaces, respectively. 6. ACKNOWLEDGMENT We wish to thank all the participants to this pilot study, and also all our collaborators: Madeleine Bellemare for her interviews of pianists about piano timbre and her involvement in the project, Sylvie-Anne Ménard for composing on command the pieces we used, Douglas Eck for granting us use to the Bösendorfer CEUS piano and for his invaluable help, Mathieu Gaudet for his piano expertise, and Dominic Thibault, sound technician, for his taking care of the recording process and of all things regarding the piano. 7. REFERENCES Askenfelt, A. & Jansson, E.V. (1991). From touch to string vibrations. II: The motion of the key and hammer. Journal of the Acoustical Society of America, 90, Bellemare, M. & Traube, C. (2006). Investigating piano timbre: Relating verbal description and vocal imitation to gesture, register, dynamics and articulation. In Proceedings of ICMPC9: The 9 th International Conference on Music Perception and Cognition. Bologna, Italy. Bernays, M. & Traube, C. (2010). Expression of piano timbre: gestural control, perception and verbalization. In Proceedings of CIM09: The 5 th Conference on Interdisciplinary Musicology. Paris: Delatour France. (in press). Von Bismarck, G. (1974). Timbre of steady sounds: a factorial investigation of its verbal attributes. Acustica, 30(3), Cadoz, C. Timbre et causalité (1991). In J.B. Barrière (Ed.), Le timbre: métaphore pour la composition (pp ). Paris: IRCAM & Christian Bourgeois. Caron, S., Traube, C. & Lacasse, S. (Eds.). (2007). Le timbre musical : Composition, interprétation, perception et réception. Les cahiers de la Société québécoise de recherche en musique, 9(1 2). Cheminée, P., Gherghinoiu, C. & Besnainou, C. (2005). Analyses des verbalisations libres sur le son du piano versus analyses acoustiques. In Proceedings of CIM05: The 2 nd Conference on Interdisciplinary Musicology. Montreal, Canada. 5

6 Clarke, E.F., Parncutt, R., Raekallio, M. & Sloboda, J.A. (1997). Talking fingers: an interview study of pianists views on fingerings. Musicae Scientiae, I(1), Faure, A. (2000). Des sons aux mots, comment parle-t-on du timbre musical? Doctoral thesis, Ecole des Hautes Etudes en Sciences Sociales, Paris, France. Fritz, C., Blackwell, A.F., Cross, I., Moore, B.C.J. & Woodhouse, J. (2008). Investigating English violin timbre descriptors. In Proceedings of ICMPC10: The 10 th International Conference in Music Perception and Cognition. Sapporo, Japan. Goebl, W. (2003). The Role of Timing and Intensity in the Production and Perception of Melody in Expressive Piano Performance. Doctoral thesis, Institut für Musikwissenschaft, Karl-Franzens-Universität, Graz, Austria. Goebl, W., Bresin, R. & Galembo, A. (2005). Touch and temporal behaviour of grand piano actions. Journal of the Acoustical Society of America, 118, Goebl, W., Flossmann, S. & Widmer, G. (2009). Computational investigations into between-hand synchronization in piano playing: Magaloff s complete Chopin. In Proceedings of SMC 2009: The 6 th Sound and Music Computing Conference (pp ). Porto, Portugal. Grey, J.M. (1977). Multidimensional perceptual scaling of musical timbres. Journal of the Acoustical Society of America, 61, Hajda, J.M., Kendall, R.A., Carterette, E.C. & Harshberger, M.L. (1997). Methodological issues in timbre research. In I. Deliège & J.A. Sloboda (Eds.), Perception and Cognition of Music (pp ). London, UK: Psychology Press. Kochevitsky, G. (1967). The Art of Piano Playing, a scientific approach. Secaucus, NJ: Summy-Birchard Music. Marozeau, J. (2004). L effet de la fréquence fondamentale sur le timbre. Doctoral thesis, Université Pierre et Marie Curie, Paris VI, France. Ortmann O. (1962). The Physiological Mechanics of Piano Technique. New York, NY: Dutton. (Original edition published 1929). Parncutt, R., Sloboda, J.A. & Clarke, E.F. (1999). Interdependence of right and left hands in sight-read, written, and rehearsed fingerings of parallel melodic piano music. Australian Journal of Psychology, 51(3), Parncutt, R. & Troup, M. (2002). Piano. In R. Parncutt & G.E. McPherson (Eds.), The science and psychology of music performance (pp ). New York, NY: Oxford University Press. Repp, B.H. (1996). Patterns of note onset asynchronies in expressive piano performance. Journal of the Acoustical Society of America, 100(6), Repp, B.H. (1997). Acoustics, perception, and production of legato articulation on a computer-controlled grand piano. Journal of the Acoustical Society of America, 102(3), Seashore, C.E. (1936). Objective Analysis of Music Performance. Iowa City, IA: University of Iowa Press. Seashore, C.E. (1937). Piano touch. Scientific Monthly, 45, Stepanek, J. (2006). Musical sound timbre: verbal description and dimensions. In Proceedings of DAFx-06: The 9 th International Conference on Digital Audio Effects (pp ). Montreal, Canada. Thompson, M.R. & Luck, G. (2008). Effect of pianists expressive intention on amount and type of body movement. In Proceedings of ICMPC10: The 10 th International Conference in Music Perception and Cognition (pp ). Sapporo, Japan. Traube, C. (2004). An Interdisciplinary Study of the Timbre of the Classical Guitar. Doctoral thesis, McGill University, Montreal, Canada. Matthay, T. (1932). The Visible & Invisible in Pianoforte Technique. Oxford, UK: Oxford University Press. McAdams, S., Winsberg, S., Donnadieu, S., De Soete, G. & Krimphoff, J. (1995). Perceptual scaling of synthesized musical timbres: Common dimensions, specificities, and latent subject classes. Psychological Research, 58, Neuhaus, H. (1971) L'art du piano [Translated from Russian to French by O. Pavlov & P. Kalinine]. Luynes, France: Van de Velde. (Original edition published 1958). 6