• Aucun résultat trouvé

RESULTS 193 impact of each metaphorical instructions on the perception of the performance, we

Being moved: literally and metaphorically 1

F.3. RESULTS 193 impact of each metaphorical instructions on the perception of the performance, we

computed a generalized linear mixed model encompassing the interaction between the metaphors presented in the evaluation scales and the instructions given to the violinists when playing (2 expressive and 5 metaphorical levels). The random inter-cept effects modelled the variability induced by the participants, the musicians, the type of excerpts, and the excerpts. Such model outperformed a model without the interaction (χ2(24, N = 2040) = 1683.19,p < 0.001, Figure F.4, full statistics in Supplementary Material F.5.l). We reported than in both expressive cases (dead-pan and projected), the performance was rated with significantly lower Force and Wandering. For the metaphorical instructions, when violinists were asked to play with Flow, we observed that it was perceived with significantly less Force (p < 0.05, full statistics Supplementary Material F.5.m). Instructions related to Movement were perceived with higher Movement than Force (p < 0.01). Instructions of “Force”

were not represented by any significant difference in the rating of the metaphorical scales. Finally, when violinists were asked to play with the Interior or Wandering metaphors, participants evaluated such performance with significantly higher rat-ings in Flow and Interior compared to the other metaphors (p < 0.001). When we compared between instructions for each specific metaphor, we observed the instruc-tion of Movement induced a significantly stronger impression of Movement than the instruction of Interior or Wandering (p < 0.01). The instruction to play with Force was characterized by higher rating in Force than the instruction to play with Flow or Interior (p < 0.01). Finally, the instruction to play with the metaphor Interior resulted in higher ratings for Interior compared to all the other instructions except Wandering (p <0.01)

Similar to the liking and expressive intent models, we also computed general-ized linear mixed models with the principal components representing both acoustic and motion features of the performance. We also kept the random intercept ef-fects and looked at PCs as sole fixed effect to capture a general impact of these PCs across all instructions. We reported that the metaphor Flow and Interior did not seem to be impacted by the computed PCs. Movement and Wandering were similarly impacted with significantly positive relationship between the last three PCs (PCmode, PCcepst, and PCmotion) and the ratings. For the ratings relative to Movement, we report a positive slope with the last three PCs (χ2P Cmode(1, N = 2040) = 21.32,χ2P Ccepst(1, N = 2040) = 13.94,χ2P Cmotion(1, N = 2040) = 19.11, p < 0.01,FigureF.5,full statistics inSupplementary Material F.5.n). Similarly, for the ratings relative to Wandering, we report a positive slope with the last three PCs

Figure F.4 – Estimated odds ratio for GEMMES based on the instructions given to violinists. Each subgraph represents a different instruction. On top, the two expressive instructions. At the bottom, the five metaphorical instructions. All contrasts are FDR-corrected [*: p <0.05, **: p <0.01, ***: p <0.001].

2P Cmode(1, N = 2040) = 26.78,χ2P Ccepst(1, N = 2040) = 17.54,χ2P Cmotion(1, N = 2040) = 7.9, p < 0.01). Finally, Force was positively impacted by the first principal component, as well as the second and fourth (χ2P Cspect(1, N = 2040) = 14.51,χ2P Cmode(1, N = 2040) = 7.2,χ2P Cmotion(1, N = 2040) = 20.06, p <0.01).

F.4 Discussion

This study aimed to measure the impact of musical metaphors on the performance of musicians from both production and perception point of views. To investigate this question, we designed a series of experiments in two phases. In the first study, we recorded violinists playing metaphorically connotated music pieces following dif-ferent expressive and metaphorical exposures. We modelled the variations in both acoustic and motion features in respect of the different metaphorical exposures. In the second study, we asked participants of different musical backgrounds to evaluate how much they liked and how expressive were the recordings from the first study.

They also evaluated the metaphorical content of each recordings. We estimated the

F.4. DISCUSSION 195

Figure F.5 – Modulations of odds ration for GEMMES by the principal com-ponents extracted from acoustic and motion features. From top to bottom, flow, movement, force, interior, and wandering. All contrasts are FDR-corrected [*: p <0.05,

**: p <0.01, ***: p <0.001].

responses of the participants based on their different level of expertise, instructions given to the violinists in the first phase, and both acoustic and motion features.

The first study highlighted the potential of expressive but most importantly metaphorical instructions on both the acoustic and motion characteristics of a musi-cal performance. On the one hand, when asking violinists to play with no expression (deadpan condition), both motion and sound features seemed to be dampened. This reflects the difficulty for advanced musicians to produce expressionless performances resulting in reduced features of otherwise normal expressive performances (Gabriels-son, 1987; Palmer, 1997). On the other hand, when asking violinists to play following congruent or incongruent musical metaphors, we observed a differential impact on each excerpt. First, it appears that more technical and fast paced pieces (Par-tita, Bruch) resulted in less differences between metaphorical conditions. Second, principal components associated more with acoustic features (PCspect and PC-mode, associated spectral characteristics, and PCrough associated with roughness and meter) appeared to be less impacted than the component associated with mo-tion features (PCmomo-tion). We report more significant differences in momo-tion features resulting from the variety of metaphorical instructions in our excerpts. This com-ponent was representing kinetic energy, as well as derivatives of movement (speed, acceleration, and jerk). Such features have been reported to be impacted during ex-pressive play (J. W. Davidson, 2007; Wanderley, 2002). Different patterns of amount of movement, speed, and acceleration are associated with different performance con-dition, with more of these features being associated with expressive conditions (Van Zijl & Luck, 2013). Within each excerpt, we also observed that not all metaphors produced the same effect. Distinct metaphors seemed to be associated with cer-tain qualities of body movements, similarly to emotions (Castellano, Villalba, &

Camurri, 2007; Crane & Gross, 2007; Wallbott, 1998). One striking case was the opposite metaphors Movement and Interior. There were respectively as congruent and incongruent metaphors for the excerpts Partita and Reveuse. This resulted in opposite effects when comparing these two excerpts for every principal compo-nent. Consequently, this suggests that metaphors create patterns of variations in the musical performance that are consistent with the nature of the excerpts, i.e. the composer’s intention to induce some metaphorical representation. Moreover, oppo-site metaphors can produce oppooppo-site effects over multiple pieces. While metaphors emerge from embodied past experiences or image schemas (Lakoff & Johnson, 1980), our results seem to indicate that both the basic comprehension of metaphors and the embodied cognition of musical performance are linked by such bodily

experi-F.4. DISCUSSION 197