Introduction
Music psychological literature has only recently started to consider the teaching and learning of expressiveness in music performance through empirical and detailed analyses. There is empirical evidence that searching to communicate emotions through music is common at any level of learning (e.g., Juslin & Laukka, Reference JUSLIN and LAUKKA2004), that the intention of professional musicians is often to express emotions (Lindström et al., Reference LINDSTRÖM, JUSLIN, BRESSIN and WILLIAMON2003), and that expert musicians are able to express specific emotions through the manipulation of music elements such as tempo, articulation, dynamics and timbre (Juslin et al., Reference JUSLIN, KARLSSON, LINDSTRÖM, FRIBERG and SCHOONDERWALDT2006; Viellard, Roy & Peretz, Reference VIELLARD, ROY and PERETZ2012; Timmers & Sadakata, Reference TIMMERS, SADAKATA, Fabian, Timmers and Schubert2014). As stated by Berman (Reference BERMAN and Berman2000), ‘the pianist needs to know what physical actions influence sound and in what way’, as he thinks that every musician, in general should know what sounds are linked with specific emotions.
The ability of communicating emotions, as well as other forms of expressiveness, in music performance is sometimes wrongly assumed as natural and not susceptible to be improved with training (Broomhead, Reference BROOMHEAD2006; Lindström et al., Reference LINDSTRÖM, JUSLIN, BRESSIN and WILLIAMON2003; Williamson, Reference WILLIAMSON, Fabian, Timmers and Schubert2014). This assumption may be related to the implicit nature of emotional communication, which complicates its transmission or training (Sloboda, Reference SLOBODA and Ericsson1996). According to the congruence of emotion theory (Foa & Kozak, Reference FOA and KOZAK1986), the emotional state active at a given time would guide the cognitive structures to be used, favouring aspects such as recall of information, attention to stimuli or production of responses, also congruent with the active mood. Likewise, according to Bower (Reference BOWER1981), the ease in activating a concrete emotion (e.g., fear) would depend on the type and amount of concepts related with that emotion in long-term memory and on the strength of the connection between those concepts. Thus, emotions involved in music are strongly determined by our previous learning, and appear automatically, and information related to emotional content is more easily stored and less likely to be forgotten (Talmi et al., Reference TALMI, SCHIMMACK, PATERSON and MOSCOVITCH2007). As stated by Chaffin, Logan, and Begosh (Reference CHAFFIN, LOGAN, BEGOSH, Hallam, Cross and Thaut2009), this involves that
the performer's visceral response to the music contributes to musical memory. We have observed that musicians find it difficult to play from memory when asked to perform without expression and surmise that playing without expression eliminates emotional cues that normally contributes to the retrieval of the music from memory (p. 356).
This line of reasoning suggests that when expressiveness is taught to arise ‘from within’ the efficacy of a performer's expression is determined by the play of emotional interactions that are involuntarily activated, influenced by what happened before, what could happened after, who is listening to the performance, and so on. In support of this idea, Minassian, Gayford and Sloboda (Reference MINASSIAN, GAYFORD and SLOBODA2003) found that up to 60% of public performances of a group of young high-level musicians were considered as inadequate by themselves, despite very high pre- and post- expectations according to their preparation. Also Van Zijl and Sloboda (Reference VAN ZIJL and SLOBODA2010) found that the role of emotions of performers changed when they achieved an expressive performance: ‘the conveyance of the previously constructed interpretation towards an audience seemed to take centre stage – rather than the performer's own emotional experience’. Likewise, the verbal and nonverbal behaviours of both students and teachers play a relevant role in their personal interactions and in the pedagogical outcomes of music learning (Zhukov, Reference ZHUKOV2012; Ivaldi, Reference IVALDI2014). Explicit consideration of affect and emotion questions could therefore facilitate general learning, not only expressivity skills. This is for example evident in the effect of drama training programmes, where explicit consideration is given to emotion and affect in communicative expression in public (e.g., Nicholson, Reference NICHOLSON and Nicholson2000). Notably, training in acting techniques and collaborating with actors improved self-confidence, creativity and presence in the performance of musicians (Rea, Reference REA2015).
In music curricula, a systematic approach to the teaching and learning of expressivity seems lacking, and consequently its training depends on conceptions, beliefs or opinions about its relevance, and about the best method of training, of each teacher (Karlsson & Juslin, Reference KARLSSON and JUSLIN2008). A recent study about conceptions and practices of music teaching and learning showed a strong focus on transmission of musical and technical knowledge rather than on a promotion of musical understanding and expressiveness using a constructivist approach (De López-Íñiguez, Pozo & De Dios, Reference DE LÓPEZ–ÍÑIGUEZ, POZO and DE DIOS2013). Even when the curricula are based on constructivist principles, as in the UK, in practice a behaviourist approach is still often employed (Garnett, Reference GARNETT2013). Teaching in music classes seems to be mostly focused on learning specific skills instead of on constructing concepts and developing understanding (Ofsted, 2009). Investigations related to the teaching of expressivity similarly suggest that a behavioural approach is the dominant practice (Karlsson & Juslin, Reference KARLSSON and JUSLIN2008). In a small sample study, Meissner (Reference MEISSNERin press) found a variety of approaches, including auditory modelling, the use of metaphors, movements and gestures, and teacher's enquiry and discussion. Whatever teaching approach is employed, its success may depend on the conceptions and beliefs about expressiveness held by the teachers and students of music. In that sense, knowing the conceptions of expressiveness is a first step in developing and enhancing a teaching method. Little empirical work is available that has analysed this issue.
Currently, expression is understood as a multidimensional construct including emotional communication, but also stylishness, quality and expressiveness within particular musical parameters (Schubert & Fabian, Reference SCHUBERT, FABIAN, Fabian, Timmers and Schubert2014). As defined in the Grove Music Online (Kovalef Baker, Paddison & Scruton, Reference KOVALEF BAKER, PADDISON and SCRUTON2016): ‘In its simplest sense, the term “expression” is applied to those elements of a musical performance that depend on personal response and that vary between different interpretations’. In this study, hereafter, the main focus will be on emotional expression, the way in which music communicates emotions to listeners.
Different factors have been considered in the literature as potentially affecting expressivity in music and its teaching and learning. For the goals of the present work, we will focus on those in which there is no consensus, where a heterogeneity of responses as well as differences between teachers and students can be expected. Thus, there is still some controversy regarding to what extent (emotional) expressivity can be learned, whether it depends on innate talent, and if it can or cannot be explicitly addressed in class. Whereas music achievement and talent for music have been thought to depend partially on genetic and epigenetic factors (Chaffin & Lemieux, Reference CHAFFIN, LEMIEUX and Williamon2004), it is clear that early experiences, preferences, opportunities, habits, training, and practice are paramount for musical development (Howe, Davidson & Sloboda, Reference HOWE, DAVIDSON and SLOBODA1998). Developing musical expressivity is not an exception, given the evidence that indicates that explicit instruction improves expressivity (Juslin et al., Reference JUSLIN, KARLSSON, LINDSTRÖM, FRIBERG and SCHOONDERWALDT2006; Woody, Reference WOODY1999, Meissner, Reference MEISSNERin press), but it seems that this is not a generally shared belief. Other factors relevant to the teaching-learning of expression in music include the idea that a) expression in music is just or mainly a matter of technique (Karlsson & Juslin, Reference KARLSSON and JUSLIN2008); b) musical elements (tempo, timbre, etc.) are associated in specific ways or not with the production of specific emotional outcomes (see Juslin & Timmers, Reference JUSLIN, TIMMERS, Juslin and Sloboda2010 for a review of the evidence on this topic); c) different stylistic periods are associated with different expressive elements, and a specific piece should be studied on that basis (e.g., Daynes, Reference DAYNES2010); d) the age of the student affects the way of expressively studying a piece, and language for expressive terms should be adapted (Tan, Pfordresher & Harré, Reference TAN, PFORDRESHER, HARRÉ, Tan, Pfordresher and Harré2010); e) and that playing by memory affects the expression achieved (Chaffin, Logan & Begosh, Reference CHAFFIN, LOGAN, BEGOSH, Hallam, Cross and Thaut2009).
In her review of the consideration of emotion in music education, Susan Hallam (Reference HALLAM, Juslin and Sloboda2010) highlighted the lack of research in this field and suggested several lines for further research in order to fill that gap, including pupils’ and teachers’ perceptions. As stated by Casa-Mas, Pozo and Montero (Reference CASAS–MAS, POZO and MONTERO2014), investigating the beliefs and conceptions of the different agents of musical education should be a central goal for research, which has not been the case yet for expressiveness. To address this gap, this work aimed to analyse the extent to which certain beliefs and concepts of expressivity were held by music performers and to explore sub-groupings of these beliefs and conceptions. Additionally, the aim was to compare these conceptions between students and professors participating in advanced instrumental courses in a conservatory. With this goal in mind, a questionnaire was developed consisting of questions related to potential beliefs of musicians about factors affecting expressiveness. Given the lack of previous studies comparing these groups this study has an exploratory character, and no hypotheses were formulated a priori.
Method
Sample
Participants were students and professors of advanced courses of instrumental music in the Higher Music Conservatory of Madrid, Spain. The final sample comprised 69 participants. 53 were students, 50.9% women (M age = 22.73, SD = 3.01, range: 17–29) and 49.1% men (M age = 22.73, SD = 3.01, range 18–28). 16 participants were professors, 68.8 % women (M age = 40.92, SD = 7.87, range: 30–55) and 31.2 % men (M age = 42.80, SD. = 12.19, range: 32–58). All students were involved in complementary piano lessons (piano as second instrument). Main instruments included 37.7% strings, 26.4% woodwind, 18.9% brass, and 17.0% keyboard instruments. Distribution of the main instrumental subjects for the teachers sample was: 37.5% piano complementary, 12.5% accompanying piano, 31.3% string, 12.5% woodwind, and 6.3% brass.
Measures
A questionnaire was developed in order to collect the relevant conceptions about the teaching-learning of emotional expression in music. Relevant dimensions were discussed among the investigators and a first version composed of 20 questions was completed for five professors and ten students, and contents and formulation of questions were discussed with both groups, in order to optimise content validity. Nine questions were considered as not relevant, or redundant, and were omitted. The final version of the questionnaire included 11 Likert-type items with four alternatives of response indicating the degree of agreement with each statement, ranging from ‘1’, Not at all to ‘4’, Very much. No specific instructions or additional explanations were provided. A verbal informed consent was obtained from the participants and total confidentiality and anonymity of the responses was guaranteed.
Statistical analyses
First, descriptive analyses (mean and standard deviation) were performed for all items in each group (students and teachers), and means were compared through independent-samples t-tests. Additionally, the effect size (Hedges’ g) was estimated for each comparison; g is an unbiased version of the known Cohen's d, and the usual guidelines for d (Cohen, Reference COHEN1988) were followed: lower than .20 is an irrelevant effect; between .20 and .50 a small effect, between .50 and .80 a medium effect, and over .80 a high effect.
Then the structure of the questionnaire was analysed with exploratory factor analysis techniques. A principal component factor analysis (PCA), as method for factor extraction, was used. Scree test and Kaiser rule (eigenvalue>1) were used to decide how many components should be retained. Cronbach's α coefficient was also used to assess the internal consistency for the total of scales. Despite of the small sample size, the degree of over-determination of the model (11 variables and 3 factors) and the moderate-to-high communalities of the variables indicate that the sample size was adequate according to theoretical and empirical (with Monte Carlo simulation) suggestions by MacCallum et al. (Reference MACCALLUM, WIDAMAN, ZHANG and HONG1999) and Arrindel and Van der Ende (Reference ARRINDELL and VAN DER ENDE1985).
Finally, total scores for subscales were calculated adding scores for items associated with each factor. Items with negative loadings were reversed. t-tests were then performed comparing scores for each subscale between teachers and students, and effect sizes for each comparison were estimated. In order to control for family-wise type I errors, a multiple analysis of variance with covariates (MANCOVA) was run with scores in the previously constructed subscales as dependent variables, group and gender as independent variables, and age as covariate. Depending on the presence of omnibus statistically significant effects, post-hoc comparisons for group, age and gender, as well as for interactions, were performed. Box's M test was used as a way to test the adequacy of data for this type of analyses; effect sizes (square root of partial eta squared, which can be considered equivalent to r were calculated for each main effect (Rosenthal, Rosnow & Rubin (Reference ROSENTHAL, ROSNOW and RUBIN2000); .24 is considered a medium effect for r and .37 a large effect (Cohen, Reference COHEN1988). Bonferroni tests were used for the post-hoc tests for paired comparisons if homocedasticity could be assumed (i.e., statistical test indicated that variances were the same for both groups in that variable); Games-Howell tests were used when equality of variances could not be assumed.
Statistical analyses were performed with the software SPSS for Windows, version 19.0.
Results
Descriptive analyses for each item in both groups are shown in Table 1. Comparisons between groups showed statistically significant differences (p < .05) for four items, and effect sizes were in general medium.
Table 1. Means and standard deviations of responses from students and teachers per item and comparison between the responses using an independent samples T-test (t values in bold indicate responses that are significantly different)
g: Hedges’ g effect size
Chronbach's alpha for the whole scale was low (.515), indicating lack of unidimensionality. The determinant of the matrix was .01, KMO test had a value of 0.5, and Bartlett's statistic was 91.2 (df = 55; p = .002). Based on these results an exploratory factor analysis for the total sample was undertaken. According with the K-1 rule and the scree-test, three components were retained explaining 48.96% of variance. Distribution of eigenvalues is shown in Figure 1.
Figure 1. Scree-test for the questionnaire about conceptions on expressivity
Therefore, a three factors solution was retained (eigenvalues: 2.32, 1.63, and 1.36). Table 2 presents factor loadings for each item in the three factors, after oblimin rotation. Correlations between latent factors were low and non-significant (ranging between .03 and .14).
Table 2. Factor structure of items on conceptions about expressivity. Configuration matrix (direct oblimin rotation)
Note: Numbers in bold indicate in what factor each item presents the higher loading
For Factor 1, items 1, 4, and 7 had high regression weights. These items include questions related to the idea that expressivity should be worked on in the classroom and is part of the musical technique. Additionally, it associated with the idea that expressive elements are linked to specific stylistic periods. Seemingly contrasting, item 2 (‘expressivity is developed by oneself’) also had a high loading on this factor (.552), although its highest loading was on Factor 3. This suggests that participants believe that while expressivity should be worked on in class and is part of musical technique, it also develops by itself. We will see that this contrasts with Factor 3 which primarily emphasises that expressivity cannot be taught. We have labelled Factor 1 as ‘expressive technique’ (hereafter, ET).
The second factor had high factor loadings for items 3, 5, 6, 10, and 11. These items are associated with the factor either positively or negatively (item 6). These items include the ideas that expressivity is affected by playing by memory, concrete emotions are associated with specific music elements, teaching of expressivity has to be adapted to the age of the student, the expressive performance of a piece is not affected by stylistic period (negative factor loading), and expressivity is a gift necessary to be a good musician. The emphasis here seems to be on emotional expression that does not depend on style, but depends on age and performance context, and relates to the giftedness of students. We can think of this factor as ‘emotional expression’, ‘expressive talent’, or ‘expressive development’. We have labelled this second factor as ‘emotional expression’ (hereafter, EE).
Factor 3 includes high loadings of items 2, 8, and 9. Those items are associated with the idea that expressivity can be developed by oneself. Unlike the weight of item 2 on Factor 1, its combination with negative weights for items 8 and 9 indicate that participants belief that expressivity training does not need to be explicit – they do not need to be explicitly included in teaching programs (item 9), and previous knowledge of what (and how) musical elements influence expressivity is not needed (item 8). Given the positive loading of item 2 and the negative loadings for items 8 and 9, this factor could be labelled as ‘self-learning of expressivity’ (hereafter, SLE).
Subscales total scores were then calculated adding scores of the items representing each of the three retained factors (scores for items with negative factor loadings were reversed). Differences between students and professors in these three scores were tested using t-tests for independent samples. Statistically significant differences were found for EE (t(67)=3.03; p = .004; g = 0.88), with a higher score for professors (M = 14.5; SD = 2.1) than for students (M = 12.5; SD = 2.2). Likewise, there was a trend for higher scores of professors in SLE (t(67) = 1.81; p = .075; g = 0.53). Finally, there was no significant difference between students and professors for ET (t(67) = 0.34; p = .733; g = 0.10). These differences are graphically presented in Figure 2.
Figure 2. Mean scores for students and teachers in the three factors of conceptions about expressivity education
More restrictive analyses, statistically controlling for gender and age, produced similar results. Thus, in a MANCOVA 2x2 (group and gender) with age as a covariate and each composed subscale as dependent variablesFootnote 1 there was no significant effect in the omnibus analyses for gender (Wilk's lambda = 0.960, F(3,54) = 0.75; p = .526; r = .200), or age (Wilk's lambda = 0.950, F(3,54) = 0.96; p = .420; r = .224) or for the interaction gender*group (Wilk's lambda =0.929, F(3,54) = 1.38; p = .258; r = .266), or for any interaction with age, but the effect of group was statistically significant (Wilk's lambda = 0.841, F(3,54) = 3.39; p = .024; r = .399). When performing univariate analyses, it was found that differences between student and professors were only significant for Factor EE in the same direction that the previous reported t-tests showed (higher scores for teachers): F(1,56) = 6.15; p = .016; r = .315.
Summary and Discussion
This work aimed to analyse conceptions of music teachers and advanced students regarding factors associated with expressive performance and its teaching and learning. A questionnaire was developed including factors suggested in the literature as potentially affecting these processes and for which some lack of consensus was shown to exist, and heterogeneity in responses could be expected. The questionnaire including 11 items presented an acceptable factor structure with three components explaining around 49% of variance, which we named expressive technique (ET), emotional expression (EE) and self-learning of expressivity (SLE). Differences between teachers and students were only found for EE in univariate and multivariate analyses (controlling for gender and age), with significantly higher scores for teachers and a large effect size. Differences in specific items were also found: concretely, students agreed more with the statements that the manner of expressively performing a piece depends on the stylistic period, and that historical periods are linked to concrete expressive elements, whereas, teachers had higher scores on items that indicated that expressive elements should be explicitly included in the curriculum, and the age of the student influences what language should be used for expressivity.
Regarding the factor structure, there was a cross-loading in the results of the factor analysis, as item 2 (‘It is developed by oneself’) had high loadings in Factors 1 and 3. Further cross-validation of the questionnaire with larger samples of teachers and students should clarify the precise structure of the scale. The ambiguity of classification of item 2 contributing to two factors with contrasting meanings at the same time (i.e, susceptible to be taught or trained vs. developed by oneself) is a limitation of the present work. On the other hand, it highlights that it is possible to think at the same time and to different degrees that expressivity is developed by oneself and that it can be trained or taught, at least in the context of individual instrumental lessons.
Likewise, the difference in factor loadings between items 6 and 7 is of interest. They are similar in content and could be expected to have similar factor loadings. However a subtle difference in the content of the items could explain the difference. Item 7 includes the idea that concrete expressive elements differ with stylistic period, which is congruent with the first factor: expressivity is a matter of technique. Item 6 on the other hand is related to the stylistic dependence of expressive performance, which is negatively correlated with Factor 2, Emotional Expression. Participants either emphasise emotional or stylistic expression.
Interestingly there were no differences according to gender or interactions with gender, even when women have been traditionally thought to give more relevance to expression of emotions in music. Our results are in line with recent results that indicated no systematic differences in gender in the ability to classify musical emotions (e.g, Mohn, Argstatter & Wilhelm-Wilker, Reference MOHN, ARGSTATTER and WILKER2011). Men and women seem to share similar conceptions of expressivity. Age, or age*group interaction, were neither significantly related with any of the conceptions: differences between teachers and students seem related to their different roles in the teaching-learning processes rather than to age.
Regarding the questionnaire, given the lack of standardised measures for these purposes, it is proposed as a tool for assessing conceptions about expressivity both in educational and research settings. Development of items for each of the three factors can be used to validate the proposed dimensions. More research on its psychometric properties is clearly needed.
Differences between teachers and students in specific items, and in the different subscales deserve more attention. Whereas there is not clear consensus about the optimal way of improving expressivity, there is increasing evidence suggesting that explicit work with expressivity could be beneficial for students, and should be included in the curriculum. Analysing conceptions of teachers and students about expressivity gives relevant information, as they are directly relevant for the process of learning, and we should know how they differently cope with the question of expressivity in music. Particularly, results of this work suggest that teachers think more that expressivity can be explicitly trained and worked on, and is dependent on age and performing context, results for which empirical evidence has been previously found (Howe, Davidson & Sloboda, Reference HOWE, DAVIDSON and SLOBODA1998; Juslin et al., Reference JUSLIN, KARLSSON, LINDSTRÖM, FRIBERG and SCHOONDERWALDT2006; Tan et al., Reference TAN, PFORDRESHER, HARRÉ, Tan, Pfordresher and Harré2010). Students in contrast see expressivity more as dependent on a stylistic context, possibly reflecting a greater sense of historical performance practice among students. More explicit conceptions would be expected for teachers, and this idea is in agreement with the empirical evidence. The important point is that disparity in beliefs should be considered as well as their effects on teaching and learning, and possible implications for the curriculum, regardless the mere adequacy of the conceptions maintained for ones or others.
A clear limitation of this work is the small sample size, which could preclude obtaining statistically significant effects, even when effect sizes in some cases suggest that such effects could be present and be very relevant. Particularly, the sample of teachers was small, affecting statistical power, although the participating teachers can be considered representative of their educational context. Besides, for the exploratory factor analysis the whole sample was included and, as indicated before, the sample size could be considered as adequate.
Attitudes towards expressivity and its learning can be worked on and discussed in class, furthering students’ knowledge about how to develop expressivity in music performance. As Hallam (Reference HALLAM, Juslin and Sloboda2010) pointed out, direct support for professors is needed in order to develop adequate skills for practicing, particularly in middle and higher educational levels, and a gap has been reported between the extent to which teachers think that they are really teaching adequate strategies, and negative reports from the students (Jørgensen, Reference JØRGENSEN2000). Thus, although teachers in the present study were convinced that expressivity should be worked on in class and included in the syllabus, students did not recognise to the same extent the relevance of explicit work. Such a gap could be a barrier in the teaching of expressivity. Similarly, differences in beliefs and concepts of expressivity among students and teachers could be a source of miscommunication and frustration in teaching situations affecting the effectiveness of teaching (Madsen & Geringer, Reference MADSEN and GERINGER2014). On the other hand, it is also likely that beliefs and concepts are transferred between teachers and students, or that teachers are sensitive to and respond to students’ conceptions. More focused research is warranted here.
It is noteworthy that even in items in which there were statistically significant differences between teachers and students (e.g., ‘Expressive elements should be explicitly included in programs [curricula]’), mean scores for both groups were quite high. That is, overall all participants of the study seemed to consider expressivity as a relevant issue. It is not clear to what extent this is reflected in practice or how expressivity is considered in the contents, objectives and assessment criteria of education. Likewise, it is not clear what specific activities are developed for improving expressivity. The scant available evidence in previous research suggested a large variety of approaches and a lack of specific formal guides (Broomhead, Reference BROOMHEAD2006; Meissner, Reference MEISSNERin press). Besides, discrepancies have been found between teachers’ intentions and pedagogical practice as well as between teachers’ practices and students’ expectations (Carey & Grant, Reference CAREY and GRANT2015). Results of the present work suggest that both teachers and students think that expressivity is a relevant issue that should be explicitly considered in the syllabus. More research is needed on this topic, and music curricula could be critically considered taking into account varying beliefs and approaches towards enhancing expressiveness.
Acknowledgements
The preparation of this article was supported in part by a grant to the first author from the Universidad Autónoma de Madrid (UAM) for a research stay in the University of Sheffield. We would like to thank the Conservatorio Superior de Música de Madrid and the UAM for its collaboration.
