Background. Extramusical elements (EMEs: emotions, memories, texts, narratives, images) are analogies to musical ideas that music teachers commonly adopt to help students understand and develop musical concepts (Meissner, 2016; Wolfe, 2018), as well as to enable expert musicians to maintain concentration while performing, seek inspiration, and further enrich their performances (Héroux, 2018).

Goals. In a set of previous studies, we explored how the addition of EMEs to the score triggers different expressive strategies —in terms of profiles of agogics, dynamics, and timbre. EMEs were also conceptualized in terms of intentions and interpretation strategies.

Methods. 90 guitarists participated in the study. A 30-second piece of music (Evocation I for solo guitar by Jimmie LeBlanc) was recorded under two conditions: (A) respecting only expressive indications written in the score, e.g., accelerando/rallentando, crescendo/diminuendo, dynamic accents; (B) following further textual evocations, related to emotion and association as extramusical elements. These were: (i) meditative, with tranquillity; (ii) untimely, with anger; (iii) a tempo, but wider, with exaltation. After the recording session, a questionnaire was administered to investigate adopted intentions and interpretation strategies. Half of the participants were exposed to condition A before condition B, while the remaining ones did the opposite. Signal processing techniques were used to extract profiles of timing, dynamics, timbral predictors, brightness, and roughness. Signal processing was carried out using the MIR Toolbox 1.7.2 (Lartillot, 2021) and a tool for detecting onsets and dynamic profiles based on the constant-Q time-frequency representation (CQT) (Schörkhuber et al., 2010).

Main outcomes. As a first step, a subset of audio recordings from 30 guitarists recorded under conditions A and B were compared to each other in terms of means and standard deviations, and their differences qualitatively related to the descriptions provided by the corresponding questionnaires analyzed with NVivo 13. 28 participants declared to have made use of EMEs. For 19 of them, this affected their playing, while 14 reported an effect on their mood. At that preliminary stage of data analysis, we highlighted two contrasting cases — the first stating that no EME was used but this participant behaved as if an EME had been used, and the second confirming the use of EMEs and deviating little from the reference performance. This study extends the previous ones in four directions. First, we now consider the full set of recordings from all 90 participants. Second, a new model was implemented, in which performance comparisons are no longer based on means and standard deviations across the different acoustic parameters, but rather on the peaks that emerge from profile differences. Third, we complemented the acoustic analysis with participants' self-reports, examining which textual evocations were the most inspiring and how reported performance strategies correspond to the observed acoustic peaks. Finally, such a more comprehensive analytical framework enables clustering interpretations into performance styles and the examination of relationships between performance strategies and EMEs' categorization and classification using multiple correspondence analysis.