TIMBRE JUDGMENTS OF JAVANESE GAMELAN INSTRUMENTS BY TRAINED AND UNTRAINED ADULTS

Psychomusicology, 14,137-153 1995 Psychomusicology TIMBRE JUDGMENTS OF JAVANESE GAMELAN INSTRUMENTS BY TRAINED AND UNTRAINED ADULTS Sandra Serafini University of British Columbia, Canada Timbre similarity judgments were collected from two groups of musicians, one with training in Javanese gamelan music (the gamelan group), and one without training in Javanese gamelan (the western group), to examine whether differences in timbre perception were evident when adults trained in another culture's music were compared with untrained listeners. Two contexts were tested: an isolated tone and a melodic context. Multidimensional Scaling (MDS) and multiple regression analysis revealed that the two subject groups did not differ in their timbre perception in the isolated tone context, where the initial 50 ms of the tone (the attack centroid) was emphasized equally. In the melodic context, the subject groups diverged in a statistically significant manner. The gamelan group focused their attention almost completely on the attack centroid while the western group divided their attention between the attack centroid and the middle portion of the amplitude envelope. These results indicate that timbre perception can be modified when a listener has received training in Javanese gamelan music as an adult. A musical context is needed for these modifications to become apparent, however; otherwise the effects of training do not result in significantly different timbre judgments. Introduction Ethnomusicology and Psychoacoustics At first glance, the disciplines of ethnomusicology and psychoacoustics seem to pursue mutually exclusive areas in music research, with ethnomusicology immersed in the humanities and psychoacoustics in the physical sciences. On closer inspection, however, they meet at a point where fundamental questions about music are raised (Tolbert, 1992). Ethnomusicology shows that musical meaning has distinct implications within a specific cultural context, but broader issues found in the various fields of human perception have been minimally explored. The relative lack of perceptual studies in ethnomusicology prevents an informed cross-cultural cognitive theory of music from being developed (Tolbert, 1992). Psychoacoustics contributes to cognitive theories of music, and is concerned with the relationship between physical properties of sound such as waveforms, frequency, spectrum, and intensity, and perceptual experiences such as pitch, timbre, and loudness (Parncutt, 1989). One of the aims of psychoacoustics is to discover which physical properties are responsible for perceptual responses and how they can be used to predict those responses. Music perception and cognition theories based solely on psychoacoustic experiments, however, have usually neglected to account for cultural influences. It is necessary, although perhaps not sufficient, to learn both musical and social rules in order to become a "cultural insider." The testing of psychomusicological processes that are involved in learning and internaliz- Serafini 137

ing these rules, however, is also necessary. Experimental evidence on whether perceptual changes occur in those who are trained as adults in another culture's music system is rather scant. The processes and changes that occur in those who are trained in another culture's music system can provide new perspectives on the inner workings of how culture or cultural knowledge affects the way we listen to music. By bringing together the humanistic and behavioral positions of ethnomusicology and the scientific positions of psychoacoustics and neuropsychology we can begin to work toward more complete models of music perception that take into account different levels of cultural knowledge. Vision and Speech Perception Studies At present, research on the neurological basis of visual and speech modalities is more advanced than research on the neurological basis of music perception, but questions and arguments formulated by analogy are still useful. Nobel Prize winners David Hubel and Torsten Wiesel showed conclusively that visual stimuli deprivation at critical stages of development produced anatomical changes in the structure of visual cortex (ocular dominance columns), resulting in perceptual deficits that remained throughout adult life (Hubel, 1988). It is clear that the visual cortex is modified early in life due to the visual input shaping synaptic functions between neurons. There is reason tc suspect that this type of plasticity exists in parts of the brain responsible for other sensory modalities; and as Hubel maintains, the degree of plasticity may vary region to region: The modifiability of different cortical regions and the age span over which the modifications can occur probably will turn out to vary widely from area to area in the brain, perhaps with the primary visual cortex at one extreme, with the least plasticity and confined to the least number of postnatal years. (Hubel, 1988, p. 217) In speech perception experiments, Kuhl, Williams, Lacerda, Stevens, and Lindblom (1992) and others showed that by the age of six months, exposure to a specific language altered infants' phonetic perception, indicating that plasticity does exist in modalities other than vision. Infants from the United States and Sweden were tested with vowel sounds in both native and foreign languages, and both groups "exhibited a language-specific pattern of phonetic perception" (Kuhl et al., 1992, p. 606). Similar to the studies of Hubel and Wiesel (showing that vision perception is shaped by visual experience), the ability to differentiate between certain speech sounds became reduced or streamlined as a result of exposure to a specific language by the age of six months. By adulthood, the effects on speech perception due to linguistic experience are strong, to the point that specific phonetic segments that can be differentiated by a speaker in one language cannot be differentiated by the speaker of another language (Goto, 1971;Miyawaki, 1975;Strange & Dittmann, 1984; Werker & Lalonde, 1988; Werker & Tees, 1984, cited in Kuhl etal., 1992). 138 Psychomusicology Spring/Fall 1995

Whether musical timbres of specific cultures are consistent with the timbres of their corresponding speech is a question yet to be answered. By analogy with Werker and Tees* and Kuhl et al.'s speech perception data, however, those who have been exposed to the timbres of their traditional music at a critical period of development may have had their timbral perception abilities developed in specific ways. Although adults learn to speak and understand foreign languages, and are trained to perform in and understand foreign music systems, there seem to be specific abilities possessed by the native speaker or musician that remain beyond the reach of the adulttrained foreigner. The Cultural Insider "Learning Curve" A larger, more complex, series of experiments involving field work would be necessary to compare perceptions of adult-trained musicians in one culture with native musicians in another culture. One can imagine a cultural insider "learning curve" that progresses from an untrained, nonnative listener at one end to a trained, non-native one, and finally to a trained, native listener at the other end. This study assumes that it is possible to advance on the "learning curve*' through the acquisition of musical knowledge, and proposes an experiment that seeks to explore differences in perception between untrained and trained non-native listeners. Cross-Cultural Music Perception Studies The balance that is sought in cross-cultural music perception studies is summarized by Dowling and Harwood (1986): "Describing our hearing music is partly how the auditory system processes the basic dimensions of sound and how our cultural learning and our familiarity with a particular style of music leads us to hear it in a certain way" (p. x). Harshberger (1992) made an attempt at achieving this balance by investigating whether listeners with and without training in Indonesian gamelan differed in their perceptions of Javanese and Balinese isolated gong tones as measured by timbre similarity judgments. His results indicated that the two groups did not significantly differ in their perceptions, which was inconsistent with his hypothesis. One may speculate that the isolated tones examined only lowerlevel perceptual processes, and failed to capture a context where cultural knowledge in the form of music training could be used by the subjects and measured by the experimenter. A different experiment by Vaughn (1992) indicated that ratings of consonance/dissonance and goodness-of-f it between North Indian Thats (scales) and three common drone tunings of the tambura (a long necked lute) differed for Western and Indian musicians. Western musicians' judgments of consonance depended on properties of the spectral content of the tambura, whereas the music-theoretical relation between the drone context and the Thats influenced the divergent ratings of the Indian musicians. These experiments are examples of investigations that attempt to correlate training or explicit knowledge of music theory to subject responses. Serafini 139

The Use of Timbre in Javanese Gamelan The purpose of the present experiment was to explore whether differences existed in timbre judgments between musicians who were trained in Javanese music as adults (trained non-native listeners) and musicians who had not been previously exposed to Javanese music (untrained non-native listeners). Timbre was chosen because of its use in Javanese music to distinguish the layers of musical parts. Perceptually, timbre can serve to distinguish more simultaneous events than can loudness (Dowling & Harwood, 1986). Many studies provide evidence that stream segregation can be achieved through timbral differences (Bregman, 1990; Bregman and Campbell, 1971; Bregman and Pinker, 1978; Iverson, 1993; McAdams and Bregman, 1979; Singh, 1987; Wessel, 1979). Functionally, timbre is used as the prime carrier of melodic pitch material. The melody, or balungun, is an important part of identifying a piece and style of gamelan music. Sutton (1991), in his work detailing the traditions of gamelan music in different parts of Java, acknowledges the importance of recognizing the balungun by stating that"... more than any other element, it is the melodic line known as balungun (literally, "skeleton" or "outline") and usually sounded in single-octave form on saron and slenthem thaj; identifies a particular gendhing..." (p. 6). 1 Sutton essentially maintains that the sound quality, or timbre, of the instrument leads one to identify the melody (balungun), and in turn, the piece (gendhing). Not only is the timbre of the melody crucial for the proper identification of the piece, but the pitches of the melody determine what variation formulas other instruments will use to embellish and ornament it. Although the rhythmic patterns of the different instruments may vary slightly, it can be argued that it is the timbre of the instrument that distinguishes one mass of variations from another (Serafini, 1993). Hood (1988) gives the example of the gambang (a xylophone-type instrument) as providing the "only melodic sound of struck wood in the ensemble, so that it is clearly distinguishable from everything else" (p. 235). It is clear that "the sound of struck wood" refers to the timbre of the gambang. Timbre has also been favored over pitch since the late 1800s to mark the subdivisions of cyclic temporal patterns in Javanese gamelan music (Hoffmann, 1978). Hence, it is argued here that timbre serves as the most effective musical dimension in streaming musical lines in Javanese gamelan music. The special place of timbre in Javanese music as the carrier of the melody, and hence a major cue for what the remainder of the ensemble is to play, causes it to emerge as a good candidate for testing the cognitive responses of untrained and trained subject groups. Method Subjects Twenty musicians were divided into two groups of subjects. The westem group (/i = 10) were all enrolled in the School of Music at the University of British Columbia and had no familiarity with Javanese music, as determined by a pre-experiment questionnaire. The gamelan group (/i = 10) 140 Psychomusicology Spring/Fall 1995

were also extensively involved in musical activities, but in addition had received instruction and training from master Javanese musicians and possessed a minimum of two years* performance experience with the Vancouver Community Gamelan. Subjects were not paid. Stimuli As they are all important in the formal structure of a gamelan piece, possess the same pitch range, and the production of their sounds are minimally affected by the performer, the gambang, gender, demung, kenong, bonang, and siter were selected as the instruments for the experiment from numerous possibilities within the larger gamelan orchestra. Stimuli for the experiment were created using digital recordings of actual instruments. In the isolated tone context, tone 5 (the middle range) was used from each instrument to compare timbres. (See Table 1 for frequency and duration values.) In the melodic context, a seven note melody in pelog was used, specifically.323 6532 excerpted from a piece entitled Bubaran UDAN MAS. 2 Digital recordings of the instruments were made with a Sony TCD-D10 Pro Digital Audio Tape (DAT) recorder and a Fuji R-120 DAT tape, using an AKG 414 Condenser Microphone and a 48-kHz sampling rate. The stimuli from the DAT tape were downloaded onto a Macintosh Ilci, where SoundDesigner II for Protools (Version 2.3) was used to create and edit soundfiles that were later presented to subjects. Table 1 Frequency and Duration Values of one T< 5 in the Isolated Tone Context Instruments Dimension Gambang Gender Demung Kenong Bonang Siter Frequency 3 8 8 3 8 2 (Hz) Duration 4 0 7 4 0 7 (ms) 386 379 382 388 425 407 407 662 Because stimuli were not synthetically constructed, some compromises were made concerning the equalization of pitch and duration during sound editing. Differences in frequency from one instrument to another were minimal (the maximum difference was 9 Hz) and small enough not to distract subjects' attention away from rating only the timbres (as discussed in Krumhansl &Iverson, 1992,andP.Iverson [personal communication, September, 1993]). Compressing or expanding the sounds to equalize the duration resulted in large amounts of phase shift that severely distorted the timbres. It was decided that preserving the natural timbral character of the instruments would Serafini 141

be the prime objective in the sound-editing stage and outweigh strict equalization in pitch, duration, and tempo of the melodies. Small differences in duration for the isolated tones and tempo for the melodies consequently were left intact. Room and microphone noise were eliminated up to the initial attack of the sound but not beyond this point, again because of the severe timbral changes that occurred when this was attempted. Apparatus Experimental sessions were conducted in a sound-attenuated booth specifically designed for auditory experiments. The DAT playback machine was controlled by the experimenter outside the booth, and sounds were presented binaurally through high quality headphones inside the booth. Procedure Subjects adjusted the loudness of the stimuli to a comfortable level. The six timbres (i.e, gambang, gender, demung, kenong, bonang, and siter), were presented twice at the beginning of each of the two contexts to familiarize the subjects with the stimuli, followed by ten practice trials. Pairs of the timbres were presented consecutively (one instrument at a time) in two contexts: (a) isolated tones (Tone 5), and (b) the seven note Javanese melody in pelog tuning. Half the subjects in each group received the isolated tone context first and the other half received the melodic context first, to ensure that the order of presentation did not bias systematically the structures calculated by the Multidimensional Scaling program. AH possible consecutive pairs of the six timbres were heard four times in each order, so that by the end of the experiment each combination was judged eight times by each subject. Same-instrument pairs were heard four times only, bringing the total number of experimental trials for each subject in each context to 144. At the start of each session, the experimenter explained that subjects were involved in a timbre perception study, and that they were to rate the similarities of six timbres to be presented consecutively in pairs. The rating scale was posted in the sound booth for reference, and subjects were instructed to write a number between 1 and 10 in the space provided on sheets of paper, where 1 represented very dissimilar timbres and 10 represented identical timbres. Subjects were encouraged to use the entire rating scale and to use the practice trials to adjust their rating strategies. Each subject was told that the recordings came from natural instruments, not synthesized tones, and that slight fluctuations may have been present in the frequency, tempo of the melodies, or the duration of the tones. They were instructed to ignore these small differences and to judge only the similarity of the timbres to one another. Western subjects were reminded that the timbres of the instruments and intervals between melody notes were unfamiliar to them and to resist making an aesthetic judgment, concentrating instead on judging the similarity of the timbres. Gamelan subjects were instructed not to make aesthetic judgments based on the performance of the tones or melodies. Instructions for the two subject groups differed only in 142 Psychomusicology Spring/Fall 1995

this respect. To avoid fatigue, short breaks were taken at every 36 trials until all 144 trials were completed. The duration of a complete session for each subject was typically 90 min, including instructions and subject feedback following the experiment. Multidimensional Scaling (MDS) The mathematical data analysis technique most widely used in timbre perception experiments, and also used in this study, is Multidimensional Scaling (MDS) (Grey, 1975,1977,1978; Grey & Moorer, 1977; Gordon & Grey, 1978). This tool is particularly useful for timbre studies because it provides a way to understand the relationships between objects, such as timbres of instruments, without having to identify beforehand what underlying physical or psychological dimensions influence subjects' responses in a particular investigation (Schiffman, Reynolds, & Young, 1981). Similarity judgments are used as the principal means of collecting responses on the 1 to 10 rating scale described earlier. Arranged in a matrix and signifying all combinations of pairs of similarities, an algorithm is used to systematize these sets of numbers. Then the experimenter is able to represent these similarities as geometric distances on a map with a minimum of bias. Objects judged to be dissimilar are more distant from one another than objects judged to be similar. The MDS map is represented by a number of dimensions that subsequently must be interpreted by the experimenter. The number of recoverable dimensions in an MDS solution depends on the number of stimuli and subjects involved in the experiment. Not all factors responsible for the ratings given by subjects may be recovered by an MDS solution and those that are recovered may correspond to a physical interpretation, such as an acoustic property, but can also correspond to a psychological factor, possibly rendering it uninterpretable by the experimenter. Individual Differences scaling, also called Subject Weights, provides a picture of perceptual-cognitive variation both within and across subject groups by assigning stimulus coordinates for each subject to a map. These coordinates show the relative importance a subject attaches to each of the recovered dimensions compared to other subjects, allowing group differences to be seen easily. The following section shows the MDS solution and subject weights in each context, followed by an attempt to interpret the recovered dimensions in each context. Independent f-tests were used in all cases where two sample means were compared (Moore & McCabe, 1989). These cases include comparisons of mean ratings for each instrument pair to determine the existence of order effects, and comparisons between two group means in the sections where subject weights are described. Serafini 143

Results Multidimensional Scaling Solution: Isolated Tones When all trials were completed, responses were organized into similarity matrices. Independent /-tests were carried out to compare mean ratings for instrument pairs. For example, the mean ratings of gambang-gender versus gender-gambang were compared, determining that significant order effects did not exist. Matrices were collapsed for each subject so the eight judgments were averaged into a single value for a particular instrument pair, and a total of 20 matrices were subjected to MDS analysis (Kruskal & Wish, 1978). The similarity ratings are represented in Figure 1 as distances in a Cartesian space of two dimensions. Instruments rated as more similar are represented closer together in the space and those rated as more dissimilar are more distant. Figure 1 illustrates where the instruments lie in relation to one another based on the similarity ratings of the isolated tones by all 20 subjects. 1 n~ r 1 1 i i r i r* L Gambang Gender c S Siter r Demung Kenong Bonang i i, i i -2.5-1.5-0.5 0.5 i i i i Dimension 1 Figure 1. Tone similarity: Stimulus configuration 1.5 2.5 Along Dimension 1 (x-axis), it can be seen that the gambang and gender (known as "soft" instruments) as well as the kenong and bonang (the two "kettle-gongs") are geometrically close, indicating that their timbres were rated as being very similar according to some property. The demung was 144 Psychomusicology Spring/Fall 1995

rated as more similar to the kenong and bonang than the gambang-gender pair, but still relatively dissimilar to both of these pairs; and the siter was rated as least similar to the other instruments on this dimension as seen from its location at the edge of the axis. Along Dimension 2 (y-axis), the gambang and gender are still geometrically close together, indicating they were ranked similarly on this dimension as well. The siter and demung also have little distance between them, as do the kenong and bonang. Subject Weights: Isolated Tones Of interest here is not only how similarly the instruments* timbres were rated, but how they were rated by each group. The MDS analysis allows an examination of individual subject weights in each dimension, and a representation of these weights in a Cartesian space gives an overall sense of whether the two groups had relatively similar or different perceptions of the stimuli. Figure 2 shows the weight each subject gave to each dimension in the isolated tone context. -0.05-0.17-0.39 0.61 Dimension 1 Figure 2. Tone similarities: Subject weights for gamelan (G) and western (W) groups. In this context, those in the gamelan group (represented by G), are assimilated with those from the western group (represented by W0. Thesolu- Serafini 145

tion suggests that the two groups did not perceive the similarity of the timbres in a distinct manner from one another. Most subjects gave the first dimension more weight than the second, as can be seen from the gathering of subjects towards the right and somewhat lower side of the space. Independent f-tests were used to compare the means of the gamelan and western weights in each dimension, and significant differences were not detected between the groups for either dimension (>.05), as shown in Table 2. Table 2 Independent t-test Statistics (Dimensions 1 and 2) for Gamelan-Western Similarities in the Isolated Tone Context Dimension 1 Gamelan Western Dimension 2 Gamelan Western *A/S(>.05). M SD df 0.627 0.695 0.645 0.460 0.214 0.287 0.204 0.289 18-0.60* 18 1.65* Multidimensional Scaling Solution: Melodies The same procedure to detect order differences was used for the seven note melody as for the isolated tones, and again, no order effect was found. Matrices were collapsed and subjected to MDS analysis, and again a two dimensional solution was recovered. Figure 3 shows that the stimulus configuration of the melodic context is different from the solution of the isolated tone context. In the melodic context, subjects still found the gambang and gender very similar to each other and found the siter to be different from the rest of the instruments on Dimension 1, the x-axis. The perceptual change on this dimension occurred between the other three instruments, as the demung appeared more timbrally similar to the kenong and bonang. The y-axis, Dimension 2, reveals a more drastic change among four of the six instruments compared to the isolated tones. The gambang and gender remained in almost the same position on this axis. The kenong and demung have switched places from that of the tones, indicating the kenong became perceptually more distant from the bonang. The rank ordering of the instruments was also affected by the new position of the siter, which separated from the other instruments on both dimensions and was at the edge of the space on the 146 Psychomusicology Spring/Fall 1995

perimeter of both axes. A shift in the timbre perception of the demung and the siter occurred in relation to the other instruments when they were heard in a melodic context. 2 j 1,, r T 1 1 r Gender Gambang c g 0 s Demung- # Kenong Bonang -l r Siter -2-2.5-1.5-0.5 0.5 Dimension 1 1.5 2.5 Figure 3. Melody similarities: Stimulus configuration. Subject Weights: Melodies The subject weights configuration for the melodic context reveals a dramatic departure from that of the isolated tones. Here in the melodies the two groups split apart from each other, with most of the western group falling in one area of the diagram, and most of the gamelan group in a different area of the diagram. There is an insignificant amount of overlap in a small number of subjects, but as a whole the two groups perceived the melodic context timbres in distinctly different ways. The gamelan group is toward the bottom right edge of the space, indicating that Dimension 1 played a much more important role than Dimension 2 in their timbre perception. In contrast, the western group gave only slightly more weight to Dimension 1 than Dimension 2, as can be seen from their grouping toward the top left area of the space. Figure 4 illustrates the subject weights for the melodies. Serafini 147

-0.05-0,05-0.17-0.39 0.61 0.83 1.05 Dimension 1 Figure 4. Melody similarities: Subject weights for gamelan (G) and western (W) groups. Independent f-tests were performed and a significant difference (p <.05) was found between the gamelan and western groups in both dimensions (see Table 3). Table 3 Independent t-test Statistics (Dimensions 1 and 2) for Gamelan-Western Similarities in the Melodic Context Dimension 1 Gamelan Western Dimension 2 Gamelan Western */>(<05). M SD df 0.906 0.626 0.232 0.487 0.145 0.327 0.260 0.272 18 2.7* 18-2.1" 148 Psychomusicology Spring/Fall 1995

Interpreting the Dimensions Multiple linear regression was performed on several acoustic factors (specifically centroid 3, attack centroid, perceptual attack time, onset-offset synchrony, amplitude envelope, and peak amplitude synchrony) to determine how these factors may have correlated with the MDS solutions of the stimuli. Table 4 shows attack centroid and sustained amplitude envelope values for each instrument. Attack centroid, that is, the spectral centroid calculated for the first 50 msec of the tone, demonstrated the best correlation with Dimension 1 in the isolated tone context, R 2 = 0.845, p = 0.012, standard coefficient for Dimension 1 = - 0.952, Dimension 2 = - 0.01, as well as the melodic context,/? 2 = 0.987,p = 0.0001, standard coefficient for Dimension 1 = - 0.897, Dimension 2 = - 0.179. The sustained portion of the amplitude envelope was calculated as the mean of the second and fourth quartiles, and gave the best interpretation of Dimension 2 in the melodic context, R 2 = 0.739,/? as 0.039, standard coefficient for Dimension 1 a= - 0.003, Dimension 2 a= 0.919. The second dimension for the isolated tone context remained acoustically uninterpretable, implying there may have been an unspecified psychological or cognitive factor at work. Table 4 Attack Centroid and Sustained Amplitude Values of Tone 5 in the Isolated Tone Context Instrument Gambang Gender Demung Kenong Bonang Siter Attack Centroid (Hz) 458.33 432.28 1322.32 652.68 880.36 5968.83 Sustained Amplitude 7881.30 10117.45 1959.40 5,801.02 1,891.37 1,481.86 Discussion It should be noted that correlations between the MDS configuration of the melodic context and the acoustic factors mentioned above were made using values from Tone 5, that is, the isolated tone context. Tone 5 was included in the melody, and little variation in the spectra was detectable from tone to tone within one instrument's version of the melody. That no correlation was found between any of the acoustic factors and Dimension 2 in the isolated tone context may have been attributable to variations between the tones that went beyond those factors investigated. High correla- Serafini 149

tions were found between the MDS configuration and attack centroid (Dimension 1) as well as sustained amplitude (Dimension 2), so the procedure appears valid and useful. This study showed that isolated tones played from Javanese gamelan instruments were judged in a consistent manner by both groups (with and without training in Javanese music), while a musical context triggered different judgments between the two groups. Based on these results, timbre judgments seemed to be affected for those who were familiar with the musical/cultural context of the stimuli, but only in a musical context. Removed from a musical context, subjects judged the timbres in a manner that did not depend on familiarity with the stimuli. Acoustically, the attack centroid proved to be a significant factor in timbre organization, bearing approximately equal weight with an unidentified, perhaps psychological, factor in the isolated tone context. In contrast, a melodic context invoked a different timbral schema when the musical/ cultural context of the stimuli was familiar to the subjects. In a melodic context, the attack centroid was heavily emphasized by the gamelan group in their timbre organization, possibly due to the role of the attack in instrument identification, and the subsequent importance of instrument identification in following various temporal structures in Javanese gamelan music. The sustained amplitude level, measured by averaging the values of the second and fourth quartiles in the amplitude envelope, appears to be a second acoustic factor in timbre organization during the melodic context. The results described above indicate that amplitude envelope may play a more important role in timbre organization for those who are unfamiliar with a Javanese musical context than for those who are familiar. We may speculate that subjects who are familiar with a particular musical/cultural context may inhibit or suppress their attention to specific acoustic information to focus on what is meaningful in that context. Conversely, those who are not familiar with a musical/cultural context may not have learned to streamline or focus their attention on the attributes that give meaning to a particular context. Revisiting the Cultural Insider Learning Curve The work of Hubel (1988), Werker and Tees (1984), and Kuhl et al. (1992) shows that our visual and speech perception abilities are crucially affected, if not determined, by input stimuli at critical stages of development early in life, when the brain is "plastic'* enough to adapt physiologically to the demands of stimuli in the environment. It seems, however, that it is possible for adults to adjust and adapt their perceptions to foreign environments, to subtly alter their modes of organization by learning new sets of stimuli. In auditory perception, these stimuli could be the sounds of a foreign language or a foreign music. In revisiting our cultural insider learning curve, and keeping in mind the work of the researchers mentioned in this report, an important distinc- 150 Psychomusicology Spring/Fall 1995

tion needs to be maintained between the perceptions of someone who has become familiar with a musical or cultural context in adulthood and someone who is native to that culture and has been exposed to its auditory idiosyncrasies at critical stages of perceptual development. This study shows that training in a foreign musical/cultural context during adulthood is sufficient to invoke judgments of timbre similarity that are different from an untrained listener, but it is not possible to say whether the training processes these listeners use are necessarily ones that Javanese musicians would employ. Additional studies are needed to answer this question. It would be interesting to replicate this study in Java with local musicians and compare all three data sets and MDS maps. This type of investigation would give a more complete picture of the shape and dimensions of the cultural insider learning curve and help identify which attributes of timbre contribute to an overall process of timbre perception in the minds of Javanese musicians. The general methodology of similarity ratings in Multidimensional Scaling also could be used to look at other musical variables important in gamelan music, such as melodic contour and improvisation, or relationships between musical variables such as tempo and rhythm. Comparing judgments of Javanese musicians with non-javanese musicians who learn gamelan music as adults may help provide insight to the many unanswered questions associated with cross cultural music perception. References Bregman, A., & Campbell, J. (1971). Primary auditory stream segregation and perception of order in rapid sequences of tones. Journal of Experimental Psychology, 89, 244-249. Bregman, A., & Pinker, S. (1978). Auditory streaming and the building of timbre. Canadian Journal of Psychology, 2(1), 19-31. Bregman, A. (1990). Auditory scene analysis: The perceptual organization of sound. Cambridge, MA: MIT Press. Dowling, W. J., &Harwood, D. L. (1986). Music cognition. New York: Academic Press. Gordon, J. W., & Grey, J. (1978). Perception of spectral modifications on orchestral instrument tones. Computer Music Journal, 2( 1), 24-31. Grey, J., & Moorer, J. (1977). Perceptual evaluation of synthesized musical instrument tones. Journal of the Acoustical Society of America, 62,454-462. Grey, J. (1975). An exploration of musical timbre (Report No. STAN-M-2). Stanford, CA: Stanford University, Department of Music, Center for Computer Research in Music and Acoustics. Grey, J. (1977). Multidimensional scaling of musical timbres. Journal of the Acoustical Society of America, 61,1270-1277. Grey, J. (1978). Timbre discrimination in musical patterns. Journal of the Acoustical Society of America, 64,467-472. Harshberger, M.L. (1992). Acoustics and comparative psychoacoustics of Indonesian gong tones. Proceedings of the International Conference on Music Perception and Cognition, USA, 2, 37-38. Hoffmann, S. (1978). Epistemology and music: A Javanese example. Ethnomusicology, 22(1), 69-88. Serafini 151

Hood, M. (1988). The evolution of Javanese gamelan: Book III, Paragon of the roaring sea. Wilhelmshaven: Noetzel. Hubel, D. (1988). Eye, brain, and vision. Scientific American Library, Series 22. New York: Scientific American Library. Iverson, P. (1993). Auditory stream segregation by musical timbre. Unpublished doctoral dissertation, Cornell University, New York. Krumhansl, C, & Iverson, P. (1992). Perceptual interactions between musical pitch and timbre. Journal of Experimental Psychology: Human Perception and Performance, I8(3) 9 739-751. Kruskal,J.B & Wish, M. (1978). Multidimensional scaling. Beverly Hills: Sage Publications. Kuhl, P. K., Williams, K. A., Lacerda, F., Stevens, K. N., & Lindblom, B. (1992). Linguistic experience alters phonetic perception in infants by six months of age. Science, 255, 606-608. McAdams, S., & Bregman, A. (1979). Hearing musical streams. Computer Music Journal, 3, 26-43. Moore, D. S., & McCabe, G. P. (1989). Introduction to the practice of statistics. New York: W. H. Freeman and Company. Parncutt, R. (1989). Harmony: A psychoacoustical approach (Springer Series in Information Sciences 19). Heidelberg: Springer-Verlag Berlin. Sandell, G. (1991). Concurrent timbres in orchestration: A perceptual study of factors determining "blend." Unpublished doctoral dissertation, Northwestern University, Evanston, IL. Schiffman, S., Reynolds, M. L., & Young, F. M. (1981). Introduction to MDS theory, methods, and applications. New York: Academic Press. Serafini, S. (1993). Personal observations, Vancouver Community Gamelan Kyai Madu Sari rehearsals, Simon Fraser University, Vancouver, British Columbia, Canada. Singh, P. (1987). Perceptual organization of complex-tone sequences: A tradeoff between pitch and timbre? Journal of the Acoustical Society of America, 82, 886-899. Sutton, R. A. (1991). Traditions of gamelan music in Java musical pluralism and regional identity. Cambridge Studies in Ethnomusicology. Cambridge, MA: Cambridge University Press. Tolbert, E. (1992). Theories of meaning and music cognition: An ethnomusicological approach. In World of Music: Special Edition on Ethnomusicology and Music Cognition, 34, (3), 99-119. Vaughn, K. (1992). Influence of the tambura drone on proximity judgments among 10 northern Indian scale types. Proceedings of the International Conference on Music Perception and Cognition, USA, 2, 38. Werker, J. F., & Tees, R. C. (1984). Cross-language speech percepton: Evidence for perceptual reorganization during the first year of life. Infant Behavior and Development, 7,49-63. Wessel, D. (1979). Timbre space as a musical control structure. Computer Music Journal, 5,45-52. 152 Psychomusicology Spring/Fall 1995

Author Note The experiment reported in this paper was part of a Master's thesis accepted in October 1993 by the University of British Columbia School of Music. This research was supported by grant A9958 from the NSERC to L. M. Ward of the University of British Columbia. Correspondence concerning this article should be addressed to Sandra Serafini, University of Washington, Speech & Hearing Sciences, 1417 NE 42nd St., Seattle, WA, 98105-6246, e-mail serafini@u.washington.edu. The author wishes to thank the research subjects, particularly members of the Vancouver Community Gam el an Kyai Madu Sari, for participating in this project. Footnotes lr The saron and slenthem are instruments of the gamelan ensemble. A gendhing is a generic term for a piece of gamelan music. 2 The numbers of the melody refer to specific keys struck on an instrument, and dots signify rests. For example, the keys of a demung (a metallophone) in pelog tuning would be numbered one through seven (far left to far right), the pitch progressing from low to high. 3 See Sandell (1991) for the formula used to calculate a centroid with timedependent harmonics. Serafini 153