Article
Reference
Emotional states generated by music: An exploratory study of music experts
SCHERER, Klaus R., ZENTNER, Marcel Robert, SCHACHT, Annekathrin
Abstract
A group of music experts (N = 98) were asked to report (in responding to a questionnaire) on their affective, cognitive, and physiological reactions to a piece of music they recently heard and that struck them as having produced an emotional response. In addition, participants were also asked to rate the relative importance of a list of musical and extramusical features that could have contributed to their reactions. A coding system was developed to organize and quantify the freely reported reactions. With respect to bodily symptoms, the most frequent reactions included semi-physiological variables such as tears and shivers, cardiovascular symptoms, as well as incitement to motor action such as jumping or dancing. With respect to subjective experiences or feelings, reports such as feeling nostalgic, charmed, moved, or aroused were more frequent than reports of “basic” emotions such as sadness, anger, joy, or fear. Musical structure was given the highest rating of the list of potential determinants, but technical, acoustical, and interpretational features also received high ratings. The authors discuss how these [...]
SCHERER, Klaus R., ZENTNER, Marcel Robert, SCHACHT, Annekathrin. Emotional states generated by music: An exploratory study of music experts. Musicae Scientiae , 2001, vol. 5, no. 1_suppl., p. 149-171
DOI : 10.1177/10298649020050S106
Available at:
http://archive-ouverte.unige.ch/unige:102080
Disclaimer: layout of this document may differ from the published version.
Music:ae Scientiae
Specal issue 2001-2002,149-171
CI 2002 by ESCOM European Society for the Cognitive Sciences of Music
Emotional states generated by music:
an exploratory study of music experts
KLAUS R. SCHERER, MARCEL R. ZENTNER * AND ANNEKATHRIN SCHACHT **
"Departmentof Psychology, University of Geneva, Switzerland
"Department of Psychology, Humboldt University Berlin, Germany
• ABSTRACT
Agroup of music experts(N = 98) were asked to report (in responding to a questionnaire) on their affective, cognitive, and physiological reactions to a piece of music they recently heard and that struck them as having produced an emotional response. In addition, participants were also asked to rate the relative importance of a list of musical and extramusical features that could have contributed to their reactions. A coding system was developed to organize and quantify the freely reported reactions. With respect to bodily symptoms, the most frequent reactions included semi-physiological variables such as tears and shivers, cardiovascular symptoms, as well as incitement to motoraction suchas jumping or dancing. With respect to subjective experiences or feelings, reports suchas feeling nostalgic, charmed, moved, or aroused were more frequent than reports of "basic"
emotions such as sadness, anger, joy, or fear. Musical structure was given the highest rating of the list of potential determinants, but technical, acoustical, and interpretational features also received high ratings. The authors discuss how these results and their conceptual elaboration can provide a guide for more systematic investigation of emotion induction via music.
It is commonly held that listening to music elicits emotion. Why should this be so? One possible explanation is a phylogenetic link between music and affect vocalizations in man and animals. Based on pioneering work by Darwin, Helmholtz, Wundt, and others, Scherer (1991, 1994, 1995) has suggested that emotional expression in speech and music may ultimately be based on nonverbal affect vocalizations as found in many socially living species of animals including man (see also contributions in Wallin and Merker, 2000).This suggests that the ability to decode the emotional messages in the vocal signals produced by significant others in a social situation is of prime importance for survival, well-being, and social status in the group. Itseems natural, then, that in the course of evolution there has been a strong pressure to rely on emotional information in vocal signals. This, in turn, has
149
produced a tendency to manipulate the emotional message in vocal signals for strategic purposes (see Hauser, 1996; Scherer, 1985, for reviews) requiring ever sharper sensitivity to the "real" emotional state underlying a particular type of vocalization. Without gening into the old controversy of whether speech originated out of music or vice versa (with prosody playing a major role in either case), one can reasonably assume that many of the evolutionarily based expression mechanisms found in pre-verbal affect bursts have been maintained and adaptedto the specific form of communication, that is, speech or music. In addition, it can be assumed that additional expression mechanisms have developed in both speech and music along with the specific segmental units and the suprasegmental organization(e.g.,prosody and melody).
Civen this strong link of music tothe phylogenetically continuous mechanism of affective vocal communication, the emotional effects of music have occupied composers and philosophers for centuries, with music often being called "the language of emotion" (Cooke, 1959). While there is philosophical argument (Robinson, 1994) and strong empirical evidence (see reviews in Juslin and Sloboda, 200 1) that listeners will very reliably ascribe emotional meaning tocertain types of music and often claimtoactually experience the respective emotional states, there is little empirical data on specific, nonverbally measured, emotional reactionstomusic (except for the soothing or arousing qualities of different types of music in terms of physiological activation; see Bartlett, 1999). One of the most influential theories of emotional effects of music has been Meyer's (1956) analysis of the affective reactions
to the violation or confirmation of expectations with respectto the unfolding of musical structure in the course of the execution of a piece of music. In this vein, an important area of empirical work in music psychology is concerned with the temporal variation in perceived affective tension as related to local variations in musical structure (Krumhansl, 1997; Narmour, 1991).
Another potential source of emotion elicitation through music consists of memory associations (Balch, Myer and Paporro, 1999). There are many anecdotal reports on how certain pieces of music facilitate the recall of emotional memories and thus stimulate the recurrence of related emotional experiences. So far, the underlying mechanism has not been elucidated.Itwould be important to identify which aspects of the acoustic characteristics of musical sound activate memory representations. In general, the role of memory in the elicitation of affective experience through music constitutes a very promising field of study.
In the field of emotion research, it seems increasingly accepted that most emotions are elicited and differentiated through a process of cognitive appraisal (see Scherer, 1999; Scherer, Schorr and Johnstone, 200 I, for a review). While appraisal theorists have generally excluded emotion elicitation by stimuli like music from their explanatory efforts (Ellsworth, 1994), it does not seem unreasonable to assume that at least in some casesstandard appraisal processes may playa role, possibly combined
150
Emotional states lIeneratedbymusic KLAUSR.SCHERER, MARCELR.ZENTNER ANDANNEJ<ATHRINSCHACHT
with the effect of one or more of the other routes. In the simplest of circumstances, this may occur in the case of being exposed to music in a situation where one would rather not listen tomusic. While one could argue that this has nothing to do with the intrinsic power of music to elicit emotion, it could be the case that the type of music is evaluated in terms of standard appraisal criteria such as one's capacityto adjust to the situation or its normative implications. But even apart from issues concerning the presence or absence of music, the appropriateness of certain types of music to current goals and aims might be evaluated through standard appraisal dimensions (for example, it would be goal conducive to hear soothing music when one wants torelax, whereas it would be goal obstructiveto have tolisten to highly arousing music). Finally, it is possible that the effects of music due toother routes, e.g., memory associations, indirectly affect the criteria underlying the appraisal process. Thus, a triumphant marching rune might foster an appraisal of high power and high coping potential (the abilityto manage or adjust toan emergency situation), whereas a melancholy melody might make an appraisal of low power and low coping potential more likely. In any case, appraisal theorists might find it of interesttoextend their theoretical concerns to music and other esthetic stimuli.
While the literature on the topic of emotion and music is growing steadily, experimental psychologists interested in the emotional effects of music are often deploring the lack of attention to the nature of the musical and extra-musical determinants of emotional reactions. Furthermore, there are few theoretical models that invite cumulative research efforts (and allow falsification of hypotheses). In an attempt toremedy this situation, the first author has suggested a system of "rules", in the sense of lawful associations, that could serve as a theoretical basis for further research in this area (Scherer, 2000c). One set of rules consists ofperception rules, that is, concrete predictions asto which musical structures and performances will produce the perception of specific emotional qualities.Aspointed out above, it is also a pervasive belief that music can, at times, actuallyinduceemotion in listeners, assuming that there is a lawful mechanism underlying such effects. Here we focus onproduction rules, that is, concrete predictions as to what musical elements are likely to produce which types of emotional states in listeners and present a preliminary empirical effort to determine the relative importance of the various determinants in these rule systems and to link them to subjectively reported emotional reactions.
Since these rules have been derived in more detail elsewhere (Scherer and Zen mer, 2001), we provide only a brief summary. As in the case of perception rules. we suggest that an emotion that is actually experienced by a listener while listening to music, is shaped by a multiplicative function consisting of many different determinants (or groups of different musical and exrramusical features):
151
Exp"itmced Emotion '" Structuralftatures x hrftrmance ftatures x Listener ftatures, wh"e:
Structuralftatures '" WI (Segmmtalftatures) x W2 (SuprflSegmmtal ftatures),
hrftrmance ftatures=W3 (Perftrmer [idmtity x skills x state]) x W4 (Performance context), Listener ftatures=W5 (Musical expertise) x W6 (Stable dispositions) x W7(Currmimotiuationalstate) x WB (Currtmr moodstate)
Box: Production rules foremotioninductionbymusicasoutlined in Scherer and Zentner (2007)
The presentation of these rules in terms of quasi-mathematical notation reflects our belief that future work in this area should attempt to model the process whereby music elicits emotion by quantitatively measuring the determinants and statistically modeling their interactions and relative contribution to the variance. To this effect, the determinants, described above as groups of features, need to be represented by operarionalized variables that can be objectively measured. We suggest multiplicative rather than additive functions for these rules since, at least in some cases, ir seems unlikely thar constituent factors (such as listener personality)in and of themselves, in the absence of any emotion signaling music features (such as specific melodic structures or quality of timbre), can induce an emotional state.
Even more importantly, ir seems intuitively reasonable to assume that some listener features
v«..
musical expertise) will strongly interact (in the statistical sense of conveying special strength or significance) with specific stimulus features (for example, only experienced and somewhat expert listeners may be able to notice more subtle expectancy violations).The authors do not underestimate the difficulties involved in defining and measuring such variables in the case of musical structure, performance characteristics, etc. However, many different approaches can be brought to bear, from musical analysis to expert judgments. In some cases it might be impossible to conduct measurement operations on interval or ordinal scales and one might have to settle for nominal measurement, using qualitative operations and nonparametric techniques. Clearly, many details remain to be worked out. Yet, we intend the presentation of these rules in the form of quasi-mathematical models as a plea to start defining and measuring a large number of possible determinants and to examine their interaction and quantitative effects in an empirical fashion.
We have proposed (see Scherer and Zen mer, 200 I, for further detail) that the induction of emotional experience through music occurs via a number of different routes, based on different mechanisms: (1) central routes (i.e., implicating the central nervous system - CNS - in emotion generation) and (2) peripheral routes (based on direct effects on the peripheral. i.e., somatic and autonomic, nervous systems - SNS and ANS - with ensuing proprioceptive feedback to central areas).
152
Emotional states generatedbymusic Kl.AUSR.SCHERER,MARCaR.ZENTNER AND ANNEKATHRIN SCHACHT
In this paper we will not deal with these mechanisms nor the detailed predictions that can be generated. Rather, we will discuss research strategies that can fruitfully be employed to investigate the hypothetical predictions derived from the rule systems proposed above.
RESEARCH PROCEDURES
How can these predictions generated by perception and production rules be tested empirically? We believe that tests which are minimally acceptable from the point of view of psychological methodology require that emotional inferences from or reactions to musical stimuli are studied for which at least a certain number of segmental, suprasegrnental, or performance cues have been (a) objectively measured, (b) systematically manipulated, or (c) produced during purposeful emotion portrayal, Most of the work on emotional effects of music using one of these methodological approaches is summarized in Juslin and Sloboda (2001). In consequence, we can limit ourselves here to pointing out the major issues.
(a) The objective measurement approach requires idenrifying passages in specific pieces of music that have particularly strong emotional effects (on listener judgment and/or on observable emotional reactions) and determining the respective musical and acoustical parameters (using musical analysis, expert assessment, and acoustic analysis). While this seems to be relatively straightforward, there are only few examples for this kind of approach to be found in the literature (Frohlich and Wallbott, 2000; Krumhansl, 1997; Sloboda, 1991). Two reasons in particular seem
to be responsible for this state of affairs: (1) the difficulty of choosing the pieces of music to use on an a priori basis, and (2) the difficulty of deciding which parameterstomeasure and howtoconstruct appropriate measurement tools.
(b) The manipulation approach requires either asking musicians tovary specific features systematically or to use digital synthesis techniques. While the conscious variation of specific musical features by performers is certainly possible and yields interesting results (Sundberg, Iwarsson and Hagegard, 1995), this method suffers from the shortcoming that both singers and instrumentalists have some difficulty in manipulating many musical parameters at the same time and in an orthogonal fashion. Asshown, among others, by Scherer and Oshinsky (1977) and Sundberg (1978), digital synthesis allowstomanipulate systematically many structural features within a systematic experimental design (although musical coherence can suffer in cases where strong interdependencies between different parameters exist and only one parameter is manipulated). This approach provides an almost ideal way
to examine links between emotion perception/reaction and musical structure systematically. Unfortunately, studies using synthesis are still rare (but see ]uslin, 1997; [uslin et al., this issue), which is all the more surprising given the tremendous advances in and the ready availability of digital synthesis techniques developed over the last 15 years (Risser and Wessel, 1999). On the other hand, as in the case of objective measurement, there is actually a dearth of knowledge about which musical
153
or acoustical parameters should be and could be most profitably manipulated in synthesis studies.
(c) Musicians can be instructedtovary the performance of a given piece of music in such a way as to communicate different emotions.It is then determined whether listeners are able to successfully decode the expressive intentions of the musicians and which acoustic and musical cues have been most affected by the expressive intentions of the performers (see Juslin, 2001). Although this work has produced very valuable information, our knowledge about the link between musical structure and perceived emotion is still somewhat tenuous. A5 in the case of objective measurement of different types of musical stimuli, one of the difficulties is to decide exactly which musical and acoustical parameters should be investigated.
The role of listener features, which is given an important place in perception and production rules. has been somewhat neglected in this work. although a few studies have looked at listener differences in emotion inference from music (see contributions in Juslin and Sloboda, 2001, for a review). This constitutes an interesting issue for further research, requiring the use of standardized manipulations of musical structure across studies and of a systematic inventory of individual difference characteristics relevant to music perception. Again, one of the problems wirh this kind of approach is the difficulty of deciding just which of the numerous types of listener factors deserve particular attention and should be included in studies of this kind.
Thus. while potentially promising research paradigms are available to investigate hypotheses suggested by prior research or by theoretically derived perception and production rules, given the extraordinary complexity of the effects and the need to keep experimental designs manageable, researchers encounter the problem of deciding which of the many musical and extrarnusical factors should be given priority. This is particularly true for the measurement and/or manipulation of the independent variables, that is, the musical/acoustical and listener/context parameters, but it also holds for the dependent variables. that is, the measurement of emotional inference and reaction. For example, it is not obvious that the "basic" emotion categories (obtained through self-report of subjective emotional experience) that are widely used in this research
kg.,
joy, sadness, fear, anger) really reflect the nature of emotional experience while listening to music (Zentner, Meylan and Scherer, 2000).Similarly, in many studies on self-reported physiological symptoms (Panksepp, 1995), the dimensions or categories used are derived ad hoc and have little basis in empirical observation. Even the studies that have used peripheral psychophysiological measurement procedures usually rely on physiological parameters as they are used in extrarnusical emotion research, and it is not certain that they are the most appropriate for the study of emotional reactions to music. One might argue that if musicislinked to phylogenerically evolved affect vocalizations, it should be expected that it expresses and elicits mostly "basic" emotions that are comparable to "normal"
affect expression. However, the assumption that our "normal" emotion expression is
154
Emotional states generatedbymusic KLA.US R. SCHERER,MARCEl.R. ZENTNER AND ANNEKATHRlNSCHACHT
organized around a few "basic" emorionsisbased exclusivelyona theoretical stance (discrete emotion theory), which is increasingly called into question, and is not supported by much actuarial evidence (see Scherer, 2000a)1.
In the empirical study reponed below, we have attempted to obtain a preliminary empirical basis for some of the following questions: What are the types of psychological and physiological reactions that seem most typical of intense emotional reactions to music? Which of the potential determinants of musically induced affect seem most important in terms of explaining the variance in listener experience? What is the potential impact of listener features (such as personality or current mood) and context effects
kg.,
environment, climate, etc.)?We decided that it would be useful to study emotional responses to music in a group of music experrsto get first answers to the question of the differential importance of the factors outlined above. The hope is that results concerning the potentially most promising factors may assist researchersinthis areatodecide on the priority of the independent variables to be measured or manipulated, given the restrictions in time and resources in any researcheffort, The experts were askedto
recall the last time they reacted in an obviously emotional manner to a piece of music and to provide their assessment, with the help of a structured questionnaire, of the relative importance of different musical and exrrarnusical factors, as derived from the rules described above. The population we chosetostudy these issues were experrs in music since we presumed that professional music scientists would have a high degree of insight into their reactions to music and their potential causes.
METHOD
• Participants. The musical experrs participating in this research consisted of the members of the audience in a plenary lecture on emotion and music during the 6rh meeting of the International Society for Music Perception and Cognition ar Keele University, UK, August 5-10. They were mostly psychologists and philosophers working on music, musicologists, music teachers, or performers. A total of 98 individuals took part in the study. To ensure the anonymity of the participants, no names nor information on gender, age, or musical experience were requested.
• Questionnaire. Participants were asked by means of a single page questionnaire (reproduced in Appendix) to recall the last time that they had been affected emotionally by a piece of music andtothen describe the symptoms that made them remember the instant and the type of music. In addition, they had to rate their impression of the relative importance of 8 musical/acoustic and extramusical factors
(1) The termactuarialrefers to research desigs in which archived documents are the main source of data (e.g., police records, medical records, letters).
155
(derived from the rules described above) for bringing about the effects they had experienced. Participants could also specify additional factors, not on the standard list, and freely describe their experiences.
• Procedure. Questionnaires were handed out as the partiCIpants entered the auditorium. They were asked to fill om the questionnaire during the talk and return the completed questionnaires when leaving the auditorium.
• Data Analysis. The ratings of the relative importance of the different determinants listed were obtained using 6-point Liken scales (see Appendix) and could be processed statistically without further operations. All other information was coded by a trained coder (author A.S.), following the procedures described below.
The type of music was coded (whenever sufficient information was provided to unambiguously identify the piece) into classical and non-classical music (jazz, pop, folk, big band). Classical pieces were further subdivided according to period (baroque, Viennese classical period and romantic period).
The symptoms and feelings that were obtained via free repon were classified according to the emotion components that are postulated by modern componential theories of emotion (see Scherer, 2000a): cognitive, expressive, physiological, behavioral, and subjective feeling components. A summary of the lisr of caregories and subcategories, including specific examples for each code, is presented in Table 1. Under "Cognitions" we included reponed memories, associations, or evaluations. The physiological symptoms were coded under the following categories:
Unspecific physiological arousal; shivers, cardiovascular symptoms, throat, stomach symptoms, piloerection, as well as respiration. Since only very few facial or vocal expressive behaviors were reponed, we only identified eye symptoms (e.g., crying, tears, moist eyes, close to crying). With respect to the motivational component, in particular action tendencies or behavioral readiness (see also Frijda, 1986), we grouped repons byproactive motivation, interrupt motivation,and motor movements.
The subjective feeling component (which Scherer, 1984,2001, conceptualizes as a reflection of all other emotional components in the sense of a monitoring system) was coded as follows:Unspecific fielings(no quality indicated: e.g., "strong feeling")2, specific fielings (indication of a feeling quality; e.g., melancholy), basic emotions (strictly as defined by discrete emotion theorists [see Scherer, 2000a]; specifically happiness/joy, fear/anxiety, sadness, anger/irritation; all other specific feelings, moods or emotions were classified as "specific feeling"), subjective arousal(this was coded when arousal or activation was mentioned in a non-physiological sense), fieling calm(the opposite of subjective arousal, without physiological connotations),
(2) It is, of course, possible that respondents who only indicated "strong feeling" did actually feel a specificemotion. However, this cannot be inferred from the response and so itseemsbest to treat it as an unspecific feeling.
156
Emotional states generatedbymusic KLAUSR.SCHERER, MARCELR.ZENTNER AND ANNEKATHRIN SCHACHT
positive valence and negative valence (as defined by dimensional theories of emotion, see Scherer, 2000a), and, finally,ambivalent ftelings3•
Table 1
Coding system used to organize and quantify the experts' freely reported reactions Major emotion
componenrs Cognition
Physiological symPtO/llli
Expressive behavior Motivation/
Action tendencies
Subjective feeling
Sub-categories
Unspecific physiological arousal Cardiovascular symptOms Throat/Stomach Piloerection Shivers Respiration Eye symptoms Proactivemotivation Interrupt motivation Motor movements Specific feeling Unspecific feeling Basic emotions ArousaUCalm Valence Ambivalent
Examples
"rememberedsituationsinme past"
"felt physiologically aroused"
"increasedheart rate", "heightened blood pressure"
"lump in me throat", "warmfeelingin me stomach"
"chicken skin", "goose bumps"
"shivers", "tickle•• "chills","thrills"
"breathed deeply"
"tears", "moist eyes", "(closeto)crying"
"felt energized", "focussed arrennon"
"music leads to interruption of ongoing activity",
"blocked other moughu"
"desire todance"I"urgeto move"
"nostalgia", "tenderness", "being moved"
"strong feeling"
"happiness","sadness", "fear", "anger"
"felt aroused", "felt powerful"; "get peaceful".
"soothing"
"felt good, up, high"
"bittersweet feeling"
• Reliability. One of the authors (A.S.) who participated in the development of the coding scheme, codedallof the free responses. Two additional coders, trained in the use of the coding scheme, independently coded 25% percent of the free responses and the results were compared with those of the main coder. Because of rare occurrence and low agreement, three categories were eliminated from the list presented in Table 1 (unspecific physiological arousal; muscle tension/relaxation, and ambivalent feeling). In addition the two categories of specific and unspecific motor movements were combined to a single category. The overall reliability computed over the remaining categories was acceptable (Cohen's kappa = .65).
(3) Giventhe wide divergence of opinion in the field of emotion psychology and the absence of an agreed-upon dassification scheme, especially with respect to types of emotions and feeling states, one could expect some disagreement as to the category system used. We attempted at least to specify the criteria used for our grouping of the free responses and to seek reliability in coding.
157
REsULTS
• Types of reported music. For 84 of the 98 participants it was possible to
determine the type of music that had elicited the emotional experience; 37 reported classical music and 47 non-classical music of different kinds. It is possible that this result was affected by a big band concert given the previous evening at the conference; 7 participants mentioned a piece from this concert. Of those reporting emotional experiences to classical pieces, 7 mentioned a Baroque piece, 7 a piece from the Vienna classical tradition, and 23 a piece from the Romantic period.
Participants were asked to report the duration of the emotional quality of the feeling; in 10.7% of all cases this wasless than 1 minute, in 54.8% of the cases between 1-15 minutes, in 20.2% of cases between 15-60 minutes and in 4.3% of the cases longer than an hour. Asshown in Table 2, the duration of the emotional feeling was significantly longer for classical pieces. It is possible that this is due to some of the non-classical pieces being generally shorter (e.g., songs). For 80 participants we could determine whether the piece was vocal or instrumental; in 42.5 % of these cases vocal pieces and in 57.5 % instrumental pieces were reported.
The percentage of vocal pieces was higher for non-classical (53.5%)than for classical pieces (29.7%). For 25 participants we could determine the duration of the emotional feeling in relation to the whole piece: in 20 % of the cases it lasted only for a specific moment in the piece, for 40 % it was the whole piece, and for another 40 % it was the whole piece as well as a period after the actual end of the piece.
Table 2
Duration of emotional feeling for two types of music
<15 min >15mins Tara!
Classical 18 19 37
48.6% 51.4% 100.0%
Non-classical 37 10 47
78.7% 21.3% 100.0%
Total 55 29 84
65.5% 34.5% 100.0%
~:Chi Square=8.28,P<.004; 4 participants did nor indicate the duration.
• Relative frequency of specific symptoms. Asmentioned before, one of the major problems in this research is to determine the criteria that can be considered as reliable indicators of an emotional reaction. In this study we presumed that
158
Emotional states generatedbymusic KLAUSR.SCHERER, MARCELR.ZENTNER AND ANNEKATHRIN SCHACHT
participants might be able to identify the type of reactions that made them remember the specific instance of music listening and to choose it as a good example for their most recent affective reaction to music. Participants were specifically asked to indicate what made them decide that they had been emotionally affected and to describeallphysiological and expressive symptoms as well as the pertinent elements of their subjective experience. The free reports of these symptoms were coded as described above. Table 3 provides the total frequencies for the different categories in descending order as well as a breakdown by the type of music that generated the reaction.
Table 3
Frequency of symptoms and experiences reported (by type of music) Total Total % Non- Classical % Baroque % Vienna Romantic %
raw Classical % Cwsic%
N 98 49 38 7 7 24
Eye symptOm.l 27 27.55 24.49 39.47 42.86 42.86 37.50
Specific feding 23 23.47 30.61 21.05 42.86 28.57 12.50
Subjective arousal 23 23.47 28.57 23.68 14.29 0.00 33.33
Posirive valence 19 19.39 34.69 5.26 14.29 14.29 0.00
Unspecific feeling 18 18.37 18.37 23.68 28.57 14.29 25.00
Basic emotions 17 17.35 18.37 21.05 28.57 14.29 20.83
Shivers 13 13.27 14.29 15.79 14.29 28.57 12.50
Motormovements 13 13.27 9.18 10.53 0.00 1.02 3.06
Cardiovascular 12 12.24 16.33 10.53 0.00 14.29 12.50
symptoms
Proactive motivation 12 12.24 20.41 5.26 0.00 0.00 8.33
Cognitions 11 11.22 16.33 7.89 14.29 14.29 4.17
Feeling calm 8 8.16 10.20 7.89 14.29 0.00 8.33
Piloerection 5 5.10 10.20 0.00 0.00 0.00 0.00
Throat, stomach 4 4.08 4.08 5.26 14.29 0.00 4.17
symptoms
Respiration 2 2.04 2.04 2.63 0.00 0.00 4.17
Interrupt motivation 2 2.04 0.00 5.26 0.00 28.57 0.00
Negative valence 2 2.04 2.04 2.63 0.00 14.29 0.00
~:Percentages givenatecalculated as the percent of participants reporting a piece of music (N) from a particuJar category that mentioned the specific symptom.
The most frequently reported category was rytsymptoms, mostly related to moistness around the eyes or to tears. This result is in agreement with the work by Sloboda (1991, Table I, P: 112), which also suggested eye symptoms to be a frequent reaction. These symptoms are closely related to physiological, in particular parasympathetic, arousal. While there is little direct research on this reaction (but see some recent work by Martin and Labborr, 1991), it is generally seen as an indicator of an affect state best labeled as "being moved" (the term "tear jerker" for
"moving" movies provides a good illustration). It isquite significant that the second
159
most frequent category, specific feeLings, marks rather similar affective states, for instance, being moved, feeling nostalgic or melancholic. Basic emotioncategories such as happiness, sadness, anxiety or anger occur rather less frequently - on the whole only 17 times as compared to 93 times in which other categories of feeling report were coded. Subjective arousal, positive valence,and unspecific feeLingsare all mentioned more frequently than basic emotions. The next most frequent cate- gories areshivers, motor movements, and cardiovascularsymptoms, underlining the special interest that in particular phenomena like shivers or tickles have found in the music literature (e.g., Goldstein, 1980; Panksepp, 1995; Sloboda, 1991).
Interestingly, proactive motivation, illustrating the important effect of music on action tendencies was also fairly frequent. The remaining categories occur relatively rarely, with the possible exception offeeLing calm, exemplifying the potential relaxing effect of music.
In addition to the overall percentages, Table 3 also shows the relative frequencies for the different types of music. Given the low frequencies in the respective subcategories, it is impossible to run statistical tests to determine the significance of differences between the different categories of music. In the following we can only mention some tendencies that seem to warrant further discussion.Itseems that the moving, tear-inducing effect is more pronounced for classical music, whereas non- classical music produces more often feelings of subjective arousal, generally positive valence, proactive motivation, and specific motor movements. As to the differences between types of classical music, the differences are too small to warrant further discussion except possibly the absence of subjective arousal reportS for pieces from the Viennese classic repertoire and for positive valence for the Romantic repertoire.
• Relative importance of determinants. Table 4 shows the results of the relative importance of the different factors in the perception and production rules as rated by the participants. The means in the table are organized in descending order for the importance based on the Total column (classical and non-classical combined).
Columns 2 and 3 show the respective means for the rwo types of music separately;
t-tests of the difference between the two types of music yielded a significant p value onlyfor "musical structure" which is significantly more important for classical than for non-classical music. Interestingly, "other factors", that is, those factors not listed in the questionnaire but added by each participant, were consistently rated as the most important. Since these determinants vary enormously over participants one can assume that whether an emotional reaction occurs or not may be determined, to a large extent, by factors that are very specific for each individual4. Among the
(4) We examined the determinants mentioned under the "other" category with respect to the possibility of identifying similar factors that were repeatedly mentioned by respondents. However, while there was some overlap, there was not sufficient data to establish additional categories on the basis of these responses.
HiD
Emotional states generatedbymusic KLAUSR.SCHERER. MARCEl.R.ZENTNER AND ANNEKATHRlN SCHACHT
"standard factors" musical structure is clearly the most important for classical music, followed by acoustic features and interpretation, while the order is inverted for non- classical music. For both types of music the "listening determinants", context and mood, do playa less important role.
Table 4
Mean ratings of importance for different determinants of emotional reaction
Total Classical Non-classical
Other 4.54 4.38 4.67
Musical structure 3.88 4.53 3.36
Acoustic fearures 3.82 3.58 4.00
Inrerpreracion 3.45 3.47 3.44
Technical quality 3.08 3.34 2.88
Listener's mood 2.88 2.70 3.02
Affective involvement 2.88 2.92 2.85
Context factors 2.67 2.35 2.92
Personality 2.46 2.14 2.71
~:Significant difference (t= 10.73,P=.002) indicated in bold.
Apart from the mean importance ratings it is interesting to explore the internal structure of the evaluations - are there determinants that tend to be evaluated in a similar manner and that seem to cluster together. The results of a factor analysis (following Varimax rotation) of the importance ratings are shown in Table 5 (next page). The principal components analysis yielded three factors with Eigenvalues> 1, explaining 67.3 % of the variance. The firstfactor can be called "performance features".Itis entirely determined by high loadings on the performer's interpretation and personality, technical quality, and affective involvement. The second factor can be called "listening features" with strong loadings on context factors and listener's mood. The third factor consists exclusively of "musical structure". Interestingly, the determinant "acoustic features" does not load on this factor and one may surmise that "musical structure"wasinterpreted by the participants in the sense of cognitive representation of musical architecture.
To determine whether participants focussing on particular types of determinants have a tendency to preferentially report certain types of symptoms or behaviors, individual factor scores were computed and symptoms reported by participants above and below the median were compared. No reliable differences were found.
161
Table 5
Factor structure of importance ratings for determinants Factors
Musical srructure Acoustic features Interpretation Technicalquality
Aff~cciv~involvement Personality
Context factors Listener's mood
-0.10 0.43 0.85 0.66 0.77 0.74 0.05 0.12
DISCUSSION
2 0.07 0.02 -0.11 -0.02 0.40 0.33 0.84 0.83
3 0.91 -0.05 0.11 0.50 -0.09 -0.11 -0.11 0.27
The aim of the present study was to obtain preliminary evidence on which types of determinants of emotional reactions to music are particularly salient and should be given priority in research conducted to test perception and production rule system predictions. One important finding was that a sizeable number of the participants reported other than the standard features and gave them a high priority rating. This does not mean that these "other features" are not distinct. Clearly, the large number of respondenrs in this category is due to our clustering many dilferenr features together. Still, our findings can be interpreted in the sense that contrary to what one might have expected, the standard factors, that generally come tomind. were not the most frequent responses. Sinceallthese other features are highly variable across individuals. we have a first indication that it will be difficult to predict a high proportion of the overall variance with models that contain a limited number of features or determinants. Apparently, the emotion-inducing power of music works in subtle and complex ways that vary appreciably over individuals and thus do not lend themselvestostatistical modeling.
This said, the relatively high convergence of the priority ratings for the standard features (listed in the rules described in the introduction) shows that quantitative statistical modeling using production rules with a limited set of predictors can be expected to have some success in explaining at least part of the variance. This seems to be true for both classical and non-classical music (with the potential exception that the feature "musical structure" does not seem to be as important in non- classical music as other features). One of the aims of this research was to determine whether there is any feature among those included in the rules described above that seems relatively unimportant and could be dropped from further modeling. While the personality of the performer comes in last, the distance to the preceding determinant does not seem sizeable enough to drop this feature from further modeling anemprs.
162
Emotionalstates generatedbymusic KLAUSR. SCHERER, MARCEl. R. ZE!'ITNER AND ANNEKATHRIN SCHACHT
We have argued that the symptoms, behaviors, and feelings reponed may have been markers or indicators of the emotional effect of a piece of music that allowed retrieving the listening event from memory. While in some cases there may have been recency effects(e.g., the participants having emotionally reacted to the big band concert on the preceding evening), in most cases this memory retrieval effect may have played an important role. If that is the case, one can also assume that the reponed symptoms or feelings constitute the strongest elements of the affective response and may characterize the type of response.
The data indicate that the most frequent affective reaction, especially to classical music, is one of being moved and feeling rather subtle types of affect(e.g., nostalgia or melancholy). These findings, based on coding of data obtained through free responses, converge well with our parallel work in progress on music-specific emotions, which is based on a different, predetermined response format (see Zentneret al.,2000, for preliminary findings)5.We take the convergence of results, which were obtained on the basis of two rather different methodologies and samples, as an encouraging sign for the validity and generality of these preliminary findings on musical affect.
Based on these findings, one could assume then, that music more often generates emotional states other than the standard "basic" or "fundamental" emotions which are generally elicited by the need to adapt to specific situations that are of high significance to the individual's well-being (see Scherer, 2000b, 2001). This is as one might expect given that music is usually listened to in situations in which there is little likelihood of events occurring that are of immediate relevance to an individual's vital needs or goals (unless the music or the listening situation itself has these characteristics). It is all the more surprising that a large portion of current research on the emotional effects of music seems tofocus on a small number of such
"basic" emotions, although there are some notable exceptions such as the work of Gabrielson (2001, this issue) and Sloboda (1991).
These remarks need to be qualified with respect to two limitations of the present study. One, we asked our respondents toreport exceptionally powerful incidences of being moved by music. Obviously, this could bias our data toward the exceptions rather than the rule. In other words, our data may say little about how music is normally listened to and what effect it has under those circumstances. Since our purpose was to make sure that some emotion had been elicited, we chose this approach and we believe that it is useful to understand what the relative importance of the different determinants might be. Clearly, it would be useful to complement
(5) In this parallel work,differentgroups of listeners were instructed to rate the degree of emotion felt in response to a great variety of classical and non-classical music excerpts based on a list of pre-selected emotion terms. This list, which is being continously shortened and refined, was systematically derived from an original list of 500 French affect terms. For preliminary data regarding this work see Zentner, Scherer and Meylan (2000) or contact the authors directly.
Hill
this approach with more actuarial studies that can help to determine how onen, under normal circumstances, listening ro music will evoke emotion, and what the relative contribution of different musical and extra-musical factors is. Two, we asked only music experts and it is quite possible that this population behaves differently from normal listeners. For example, given their knowledge about empirical data in this area and their own favorite theories, these may well have affected their repom.
For example, the fact that we found relatively little mention of "basic" emotions could be due to the fact that music experrs tend to emphasize the artistic nature of music and the aesthetic feelings it provokes."Lay" listeners might well experience music differently. Yet, we felt that it was important to start with a group that would immediately understand our questions and be able to respond in a precise fashion (which may not be assured with "normal" music listeners). In general, there can be no doubt that our results are limited byallthe problems that surround verbal self-report - the possibility that listeners are not even conscious of the factors that moved them and that their responses may be biased by the factors mentioned above as well as by social desirability and many other factors. Unfortunately, so far there are few alternatives to self-report if one wants to obtain a general idea of which of the myriad of potentially important factors deserve to be systematically manipulated in an experimental study.Itwas in large part to this aim that our study wasdirected.
Thus, the present data, particularly with respect to the copious repom of expressive and physiological symptoms, suggests that investigations of emotion induction by music should measure indicators (t.g.,physiological recording, coding of nonverbal behavior) other than noncommittal verbal report which may reflect inferences of emotional meaning (or artempts at a "correct response" given sociocultural and historic stereotypes) rather than a true reactions. This is not to say that verbal report should be dropped from research on emotional effects of music.
On the contrary, subjective feeling, which so fat can only been measured by verbal report, isan important component of the overall emotional reaction and deserves precise measurement. However, given our data, it may not be advisabletocontinue with simple check lists using a very limited number of "basic" emotions. Since the affective feelings evoked by music seem very subtle and complex, the measurement of the verbal report of such feelings must be commensurate in subtlety. Concretely, rather than restricting judgment alternatives to a few "basic" emotion categories, more differentiated lists of affective state descriptors, possibly pretested for the use in music research, should be wed (see Zentner et al., 2000).
In closing, it may be importanttohighlight one particular feature of the results.
At least as far as classical music is concerned, there can be little doubt from our data that musical structure plays a central role in the mediation of these effects. This has been underlined in many mwicological and philosophical treatises on emotional effects of music (Cooke, 1959; Meyer, 1956; Narmour, 1991). However, in empirical studies in this area objectively defined structural characteristics of music
164
Emotional states generatedbymusic KLAUSR.SCHERER, MARCELR.ZENTNER AND ANNEKATHRIN SCHACHT
kg., tempo, melodic range), which could be used in a quantitative study to predict subjective response, are rarely measured (but see Frohlich and Willbott, 2000, for an exception). Clearly, any future attempts at identifying links between emotional reaction and musical structure could greatly benefit from a close collaboration between music theorists, performers, and psychologists - something that has been the exception rather than the rule so farG,7.
(6) This research was conducted with the help of financial support from the Swiss National Fund for Research. The authors thank John Sloboda and his collaborators who, as organizers of the 6th conference of the International Society for Music Perception and Cognition at Keele University, greatly facilitated data collection. The authors also acknowledge helpful suggestions from Ursula Scherer and three anonymous reviewers on the first draft of the manuscript.
(7) Address for correspondence:
Klaus R. Scherer
Department of Psychology University of Geneva 40 Boulevard du Pont d'Arve CH - 1205 Geneva
Switzerland
e-mail: [email protected]
'65
• REFERENCES
Balch, W., Myers, D. M., & Paporro, C. (1999). Dimensions of mood in mood-dependent memory.journal ofExp"immtal Psychology: Learning; M!mory,&Cognition, 25, 70- 83.
Barden, D.L.(J999). Physiological responses to music and sound stimuli.InD.A Hodges (ed.), Handbook of musicpsychology(2nd ed., pp. 343-85). San Antonio: IMR.
Cooke, D. (J 959). The language of music.London: Oxford University Press.
Ellsworth,P.C.(1994). Levels of thought and levels of emotion.InP.Ekman andR].Davidson (eds), Tbr nature of (motion: Fund8mmtal questions (pp. 192-96). New York: Oxford University Press.
Frijda, N. H. (1986). The emotions.Cambridge: Cambridge University Press.
Frohlich, P., & Wallbon, H. (2000). Uberprufung eines musikalischen Ausdrucksmodells von Freude und Trauer durch musikalische Inhaltsanalyse. [Examination of a musical expression model of happiness and sadness through musical content analysis.Acta MusicologicfJ.,72,107-19.
Gabric:lsson,A. (2001). Emotions in strong experiences wirh music. In P. N. ]uslin and ]. A. Sloboda (eds),Music and emotion: Theory and research(pp. 431-49). New York:
Oxford University Press.
Goldsrein, (1980). Thrills in response to music and other stimuli. Physiological Psychology, 8, 126-9.
Hauser, M. D. (1996). The evolution ofcommunication.Cambridge, MA: MIT Press.
]uslin, P. N. (1997). Perceived emotional expression in synthesized performances of a short mc:lody: Capturing the listener's judgment policy.Music« Scimtii£, I, 225-56.
]uslin, P.N. (2001). Communicating emotion in music performance: A review and a theoretical framework. In P. N. ]uslin and ]. A. Sloboda (eds),Music and emotion: Th!ory and research (pp, 309-37). New York: Oxford University Press.
[uslin,P.N., & Sloboda,].A. (eds), (2001). Music and emotion: Theory and research. New York:
Oxford University Press.
Kainz,F. (1962). Psychologi!rkr Sprach!. 1. Band. Grundlagm rkr al1gnn!inm Sprachpsychologi!.
(3ed ed.), Sturrgarr: Enke,
Kivy,P. (1989).Sound sentiment: An (SIay on th! musicalemotions.Philadelphia: Temple University Press.
Krumhansl, C. L. (1997). An exploratory study of musical emotions and psychophysiology.
Canadian journal ofExperimental Psychology,51, 336-52.
Martin,R B,& Labort, S. M. (991). Mood following emotional crying: Effecrs of the: siruation.
journal ofRmarch in Personality;25, 218-44.
Meyer,L.(1956). Emotion and m!aning in music.Chicago. Chicago University Press.
Narmour, E. (1991). The analysis and cognition of basic meiodic structures: The implication- realizetian moul. Chicago, IL: University of Chicago Press.
Panksepp (1995). The emotional sources of "chills" induced by music.MusicPerception; 13, 171- 207.
Risser.].-c.,&Wessc:l, D. (1999). Explorationof timbre by analysis and synthesis.InD. Deutsch (ed.), The Psychology ofMusic(2nd ed.), San Diego: Academic Press.
Robinson, ]. (1994). The expression and arousal of emotioninmusic.journal of'Aesthetics and Art Criticism,52, 13-22.
,,,
EmDtiDnal statesgeneratedbymusic KL\USR.SCHERER. MARCEL R. ZENTNER AND ANNEKATI1RIN SCHACHT
Scherer,K.R. (1984). On rhe nature and function of emotion: A component process approach.
In K.R. Scherer and P.Ekman (eels),Appraadm to emotion(pp. 293-317). Hillsdale, N]: Erlbaum.
Scherer, K R (1985). Vocal affect signalling: A comparative approach. In
J.
Rosenblatt, C. Beer, M.-C. Busnel andP.].B. Slater (eds),Aduances in tht:srudy ofbt:hallior, Vol.15(pp. 189- 244). New York: Academic Press.Scherer,K.R. (1991). Emotion expression in speech and music. In]. Sundberg,L.Nord and R Carlson (eds),Music. languagt:, spuch. and brain(pp, 146-56). London: Mac:millan.
Scherer,K.R (1994). Affect bursts. In S. van Goozen, N. E. van de Poll and
J.
A.Sergeant (eels), Emotions: EIsays on emotion tht:ory(pp. 161-96). Hillsdale, N]: Erlbaum.Scherer, K R (1995). Expression of emotion in voice and music.Journal ofVoict:,9, 235-48.
Scherer,K. R. (1999). Appraisal theories. In T. Dalgleish and M. Power (eds),Handbook of cognition and emotion(pp,637-63). Chichester: Wiley.
Scherer,K.R (2000a). Psychologicalmodels of emotion. In]. Borod (ed.),The nt:Uropsychology of emotion(pp. 137-62). New York: Oxford University Press.
Scherer,K.R (2000b). Emotional expression: A royal road for the study of behavior control. In A.Grab andW.Perrig (eels), Control ofhuman behavior; mentalprocesses, andawareness (pp, 227-44). Hillsdale, NJ: Lawrence Erlbaum Associates.
Scherer, K R. (2000c). Music and emotional meaning: Perception and production rules. Paper presented at the 6th conference of the International Society for Music Perception and Cognition, August 5-10, 2000, at Keele University, UK
Scherer,K.R. (2001). Appraisal considered as a process of multi-level sequential checking. In K R. Scherer,A.Schorr and T. Johnstone (eels),Appraisal prowst:s in emotion: Tht:ory.
mt:thods, rt:St:llrch(pp. 92-120). New York: Oxford University Press.
Scherer, K R., & Oshinsky, [, (1977). Cue utilization in emotion attribution from auditory stimuli. Motiuation and Emotion. 1,331-46.
Scherer, K R, Schorr,A.,& Johnstone,T.(eels). (2ooI). Appraisal prowst:s in emotion: Tht:ory.
Mt:thods. Rt:st:arch.New York: Oxford University Press.
Scherer, KR, & Zentner, M.R(2001). Emotional effects of music: Production rules. InP.N. Juslin and J. A. Sloboda (eds),Music and emotion: Tht:ory and rmarch(pp. 361-92). New York: Oxford University Press.
Sloboda, ]. A. (1991). Music structure and emotional response: Some empirical findings.
Psychology ofMusic, 19, 110-20.
Sloboda,]. A. (1992). Empirical studies of emotional responsetomusic. In M. R. Jones and S. Holleran (eds),Cognitillt: bam of musical communication(pp. 33-46). Washington, DC: American Psychological Association.
Sundberg, J. (1978). Synthesis of singing.SWt:dish JournalofMusicology,60, 107-12.
Sundberg. J., Iwarsson, J., & Hagegard, H. (1995). A singer's expression of emotion in sung performance. InO. Fujimura and M. Hirano (eds), Vocal fold physiology: Voiu quality control(pp. 217-29). San Diego, CA: Singular press.
Wallin, N.L.,& Merker, B. (eels). (2000). Tbe origins ofmusic.Cambridge: MIT Press.
Zentner, M.R,Meylan,S.,& Scherer, K R (2000).Exploring 'musical emotions' across fillt: gmm of music. Paper presented at the 6th International conference of Society for Music Perception and Cognition (ICMPC) August 5-10,2000, in Keele, United Kingdom.
167
Appendix
Questionnaire used in the study
ICPM 2000 -Survey on Emotional Effects of Music
Dear members of the audience,
Since you are attending this conference we assume you to be very knowledgeable about and interested in music. Our research group attempts to identify the factors that underlie the emotional effects of music and its effects.Therefore, we would appreciate if you could spend 3 minutes to fill out this sheet and leave it at the exit. You can also write further comments on the back of the sheet. Thank you very much. Your help is greatly appreciated!
Klaus Scherer& Marcel Zentner
Please recall the last time when listening to a piece of classical music produced a strong emotional state in you. Please don't read on until you have thought ofa specific listening situation.
What indicators made you decide that you had been affected emotionally by that piece of music (presumably the same factors that made you recall that specific situation just now)?
Please list all physiological and expressive symptems as well as the pertinent elements of your subjective experience.
Which piece of music did you hear?Tryto be as precise as possible with respect to the parr of the work that produced the emotional effect.
How long did the emotional quality of the feeling last?
Please indicate below how much different musical and/or extramusical factors have contributed to elicit your emotional reaction, i.e., how important you think they were to produce the affective tesponse. Please circle a number between 0 (nor at all important) to 5 (extremely important).
'68
Emotional states generatedbymusic KLAUSR.SCHERER, MARCELR.ZENTNER AND ANNEKATHRIN SCHACHT
Musicalstructure as written by the composer(i.t.,key, intervals, melody) Specific acoustic features (e.g.,timbre of an instrument or a singer) Inrerprerarion by the performer(s) (t.g., tempi)
Technical quality of the performance (by a soloist or an orchestra) Affective involvement of the performer(s)
Context factors (concert hall, special occasion, unexpectedness) Your own mood at the time
Other (please specify
169
Importance
