The influence of audience presence in expert classical piano performance: Reliable reproduction with greater expressive variability

Data and analysis code accompanying a study of how the presence of an audience shapes expert classical piano performance. One pianist performed the same six-piece programme twice: alone in a private dress rehearsal, and three days later in a public concert, with venue, instrument and recording setup held constant. Performance was measured through body motion (frame-to-frame pixel change from video), beat-level timing, sound intensity (MIDI keystroke velocity) and sustain and soft pedal depth. A separate group of listeners provided continuous ratings of the concert recording on five affective dimensions: arousal, energy, tension, valence and expressivity. Analyses are organised around 160 epochs (sections of the score bounded by musical phrases, eight to sixteen bars long) each observed in both playing contexts. The archive contains: processed epoch- and beat-level data (data/), the two R scripts that reproduce every table, figure and analysis statistic in the paper and its supplementary material, and the resulting output (tables/, figures/) so a run can be checked against the published results. Not included: the raw audio, video and MIDI recordings, which identify the performer and cannot be anonymised; pitch has been removed from the note-level data for the same reason. Listener ratings are deposited only as means across participants; no per-participant or demographic data is included. Full documentation of every file is in README.txt, which is also uploaded separately for reading without downloading the archive