International Symposium on Performance Science ISBN 978-2-9601378-0-4 The Author 2013, Published by the AEC All rights reserved The role of texture and musicians interpretation in understanding atonal music: Two behavioral studies Maurizio Giorgio 1, Michel Imberty 2, and Marta Olivetti Belardinelli 3 1 Department of Psychology, Sapienza University of Rome, Italy 2 Paris West University Nanterre la Défense, France 3 ECoNA, Interuniversity Centre for Research on Cognitive Processing in Natural and Artificial Systems, Sapienza University of Rome, Italy Two experiments aimed to inquire: (1) the role of performance and texture in the segmenting of a musical composition during the listening; (2) whether the structure perceived by the listeners depends on processes developing simultaneously with the listening or on an a posteriori synthesis; and (3) the role of expertise in segmentation. For each experiment 30 subjects were asked to attentively listen to two versions of an atonal composition, identify the architecture underlying the piece, and mark the boundaries between different segments by pressing the spacebar. The order of presentation of the two versions was balanced. In the first experiment the two performances differed in duration and in many dynamic aspects. In the second experiment the two performances differed only in duration. For both the first and the second experiments results showed a good number of coinciding segmentations in the two performances irrespective of the order of presentation. Musicians indicated a lower number of segmentations than not-musicians, even if many of the chosen boundaries remained the same. The results suggested that texture provided all of the necessary information for representing the pieces during the first listening. Keywords: Music cognition; grouping; segmentation; music representation; auditory perception In the 1980 s Deliège proposed a simple but effective theory: the Cue Abstraction Hypothesis (Deliège 1987, Deliège and El Ahmadi 1990, Mèlen and Deliège 1995, Deliège 1996, Deliège and Mèlen 1997). In listening to a piece
784 WWW.PERFORMANCESCIENCE.ORG some elements stand out, themselves becoming a guide to the processes of categorization and representation of the musical surface. These elements, which Deliège calls cues, allow the listener to form a mental plan of the piece by dividing it into chunks on multiple hierarchic levels. The cues represent prototypes of redundant elements; the music would then be analyzed with reference to similarities and differences between the chunks. Deliège's conception of cues is strictly related to the model of categorization elaborated by Eleanor Rosch (2002), which distinguishes two levels in categorization: the horizontal level, based on relations of similarity and contrast between elements; and the vertical level, related to hierarchic levels. Unfortunately, Deliège s model does not take into account the role of the musician s performance. To fill this gap, it is fist necessary to define the features of a musical performance depending on the performer. Gingras et al. (2011) showed that local tempo modulation is the main variable involved in characterizing the performer s expressive style. This point is in agreement with Palmer (1989). Gingras also found differences between performers in onset asynchrony. These variables (tempo modulation, onset asynchrony) coincide with what a musician can read in the dynamic indications of a score. We can hence define texture as the simple sequence of notes that a musician can find on the score of the piece, regardless of the dynamic indications. Another problem related to segmentation is the choice of composition: using tonal music can indeed introduce a bias related to cultural factors. In his model, Lerdahl (1989) suggests that the prolongational structure (strictly linked with dynamic features) has a main role in segmenting atonal music. A segmentation paradigm applied to atonal pieces, then, represents the best way to examine these ideas (Imberty 2000). Concerning expertise, differences in the number of segmentations done by musicians and non-musicians are examined not only in classic segmentation studies but also in analyzing the performance of the same composition by expert musicians (Ordoñana and Laucirica 2010). Koniari et al. (2001) found differences in segmentation accuracy with regard to groups of 10 to 11-year-old children with different degrees of training. Olivetti Belardinelli (1996) found differences in musical aptitude among genders with naïve but not with expert subjects. These experiments aim to inquire whether the segmentation of a composition mainly depends on texture or performance. Other variables examined include gender, expertise, first versus second listening, and serial order of presentation.
INTERNATIONAL SYMPOSIUM ON PERFORMANCE SCIENCE 785 Participants FIRST EXPERIMENT 30 subjects (M=12, F=18, mean age=27.46, SD=8.86) with normal hearing volunteered. All of the subjects were right-handed and had never received formal musical training. Materials Two versions of Sequenza VI per viola solo by Luciano Berio, performed by Christophe Desjardins (1998) and Garth Knox (2006), were used The performances differed in duration (12 13 versus 13 14 ) and in dynamic aspects. Procedure Subjects were asked to attentively listen to each version of Berio s piece to capture the structure and press the spacebar in order to mark boundaries between different parts. The order of presentation was balanced across subjects. Metaphor was used to explain the task to the non-musicians. A tutorial trial, with a different and shorter piece, was administered before starting the test. A graphic interface was realized using the software Max/msp (Cycling74) in order to administer the instructions, a tutorial piece, and the test and to collect the data. Analyses were performed using R-packages. RESULTS The effect of independent variables on the number of segmentations was examined through MC analysis. Empiric p-values were calculated using the formula p=(r+1)/(n+1) (Davison and Hinkley 1997). Results showed an effect of gender (NM=568, NF=1153, p<0.001) and a decrease from the first to the second listening (NI=999 and NII=722, p<0.001). Each version of Berio s Sequenza VI was divided into 24 classes of equal width (average width for performance: A=30.512s, B=34.519s). A new MC simulation was used to study the behavior of a randomized distribution of 852 (performance A) and 869 (performance B) answers in 24 classes. In a macro-analysis, we identified the classes collecting the highest numbers of answers (right tail of the distribution: p<0.1). Analysis showed 3 main segmentation areas (MSAs) for performance A (classes 4, 5, and 7) and 3 for performance B (classes 4, 5, and 8). In order to examine the possible overlapping of the segmentations in the two performances, we temporally mapped the score on the basis of the performances. Then, we could mark the relevant segmentations obtained with the
786 WWW.PERFORMANCESCIENCE.ORG two versions of the piece in the overlapping areas of the MSAs. Analysis showed 15 peaks in version A and 14 in version B, with 12 common pivots (82.75%). For the correlation analysis each performance was divided into 100 temporal classes. The analysis focused on the number of segmentations per class compared to the whole sample. Concerning the two performances, data showed a significant correlation in the zero/shift point (r=0.251, two-tailed, p<0.05), and other significant correlations with the LAGs surrounding the zero point. A stronger correlation (r=0.805, two-tailed, p<0.001) was found between the first and the second listening. DISCUSSION The cross-correlation between the two performances, along with the overlapping found both in the macro and the microanalyses, suggest that segmentation depends on texture more than on dynamics. The very strong correlation between the first and the second listening suggests that the knowledge of the piece plays a weak role in structuring the mental plan of the composition. The only difference found relates to the number of answers in each hearing, in particular to the false alarms, which decrease from the first to the second listening. The difference related to gender could be explained following Olivetti (1996), considering that none of the subjects were professional musicians. SECOND EXPERIMENT To better investigate the effect of gender a second experiment was carried out with participants of varying experience. Two performances by the same singer were used to focus on the role of duration and to attenuate the influence of dynamic variables. The procedure remained the same as above. Participants 30 right-handed subjects (M=15, F=15, mean age=34.16, SD=10.64) with normal hearing volunteered. Ten were professionals musicians and twenty were non-musicians. Mean years of formal musical training for musicians was 5.4 (SD=4.35). Materials Two versions of Sequenza III per voce solo by Luciano Berio, both performed by Cathy Berberian (1966, 1969), were used. The performances differed in duration (8 48 versus 6 55 ).
INTERNATIONAL SYMPOSIUM ON PERFORMANCE SCIENCE 787 RESULTS Data were analyzed with the same methods used in the first experiment. Analysis showed an effect of duration (NA=499, NB=408, p<0.01), order of presentation (NAB=524, NBA=383, p<0.001), expertise (NMUS=226, NNOMUS= 681, p<0.001), and first versus second hearing (NI=532, NII=375, p<0.001). No effect of gender was found. For non-musicians, women tended to indicate fewer segmentations than men (NF=351, NM=330, p<0.1), while the opposite occurred for musicians (NF=121, NM=105, p<0.001). Data showed 4 MSAs for performance A (classes 1, 3, 5, and 12) and 5 for performance B (classes 3, 5, 12, 15, and 20), with three common areas. We isolated 4 peaks in performance A and 5 in performance B, with three common pivots (67%). Crosscorrelation analyses between the performances showed significant results in LAG 0 and 2 (r=0.242 in LAG 0, two-tailed, p<0.05). As in the first experiment, we found a very strong correlation between first and second listening (r=0.773 LAG 0, two-tailed, p<0.001). DISCUSSION The second experiment supported the results of the first, with additional information. First, a longer work increases the number of segmentations if not merged into a wide set of different dynamic features. Second, differences due to gender only pertain to naïve listeners. GENERAL DISCUSSION We present here a quick summary of the main results of our study. The auditory segmentation of a composition mainly depends on the texture, while the dynamic features only have a weak influence on this task. Segmentation depends on the detection of patterns of similarity. The only difference between the first and the second listening is in the number of segments indicated by the subjects, not in their placement. Duration of the piece can influence the number of segmentations. Finally, expertise reduces the number of segmentations, perhaps because it allows the listeners to represent the composition to higher hierarchic levels. Expertise also modifies the differences between genders, nullifying the gap between the number of segmentations marked by males and females. Address for correspondence Maurizio Giorgio, Department of Psychology, Sapienza University of Rome, via dei Marsi 78, Rome 00185, Italy; Email: maurizio.giorgio@uniroma1.it
788 WWW.PERFORMANCESCIENCE.ORG References Davison A. C. and Hinkley D. V. (1997). Bootstrap Methods and Their Application. Cambridge: Cambridge University Press. Deliège I. (1987). Grouping conditions in listening to music: An approach to Lerdahl & Jackendoff's grouping preference rules. Music Perception, 4(4), pp. 325-359. Deliège I. (1996). Cue abstraction as a component of categorisation processes in music listening. Psychology of Music, 24(2), pp. 131-156. Deliège I. and El Ahmadi A. (1990). Mechanisms of cue extraction in musical groupings: A study of perception on Sequenza VI for viola solo by Luciano Berio. Psychology of Music, 18(1), pp. 18-44. Deliège I. and Mèlen M. (1997). Cue abstraction in the representation of musical form. In I. Deliège and J. Sloboda (eds.), Perception and Cognition of Music (pp. 359-382). Hove, UK: Psychology Press. Mèlen M. and Deliège I. (1995). Extraction of cues or underlying harmonic structure: Which guides recognition of familiar melodies? European Journal of Cognitive Psychology, 7(1), pp. 81-106. Gingras B., Lagrandeur-Ponce T., Giordano B. L., and McAdams S. (2011). Perceiving musical individuality: Performer identification is dependent on performer expertise and expressiveness, but not on listener expertise. Perception, 40(10), pp. 1206 1220. Imberty M. (2000). Prospettive cognitiviste nella psicologia musicale odierna - Cognitivistic perspectives in modern musical psychology. Rivista Italiana di Musicologia, 35(1-2), pp. 411-484. Koniari D., Predazzer S., and Mélen M. (2001). Categorization and schematization processes used in music perception by 10- to 11-year-old children. Music Perception, 18(3), pp. 297-324. Lerdahl F. (1989). Atonal prolongational structure. Contemporary Music Review, 4(1), pp. 65-87. Olivetti Belardinelli M. (1996). Musical Intelligence and Musical Behavior in Naif and Espert School Students. Paper presented at the International Conference of the European Council for High Ability Creativity and Culture. Vienna, Austria. Ordoñana J. A. and Laucirica A. (2010). Lerdahl and Jackendoff's grouping structure rules in the performance of a Hindemith sonata. The Spanish Journal of Psychology, 13(1), pp. 101-111. Palmer C. (1989). Mapping musical thought to musical performance. Journal of Experimental Psychology: Human Perception and Performance, 15, pp. 331-346. Rosch E. (2002). Principles of Categorization. Cambridge, Massachusetts, USA: MIT Press.