The following sound examples are the musical materials used in two recent experiments on the aesthetic quality of an "average" performance (Repp, 1997). The materials of Experiment 1, because of their length, are provided as MIDI files only, except for the average performance which is also provided as a sound file. The materials of Experiment 2 are provided both as sound files and as MIDI files. The sound files have been downsampled to conserve computer memory, so there is a slight loss of fidelity. (Sound files are in AIFF format and most are approximately 250K in size.) Playback of the MIDI files may result in an instrumental timbre different from the digital piano sound used in the original experiments.

Experiment 1 used 10 performances by piano graduate students of Robert Schumann's "Träumerei" (op. 15, No. 7) and an additional performance created by averaging the 10 student performances. The experimental stimuli preserve the expressive timing and dynamics of the original performances, but they have been regularized in other respects (onset asynchronies have been removed, pedaling has been standardized, all inaccuracies have been corrected) and have been reproduced on a Roland RD-250s digital piano rather than the Yamaha Disklavier on which they were orginally recorded.

Experiment 2 used the beginning of Chopin's Etude in E major (op. 10, No. 3) in performances by 9 graduate student pianists (on the digital piano) and 15 famous pianists (on commercial recordings). Only the original timing patterns were preserved in the experimental stimuli; expressive dynamics and other performance aspects were held constant. Separate and combined averages were formed, and three performances representing the first three (Varimax-rotated) components of a principal components analysis on the 15 expert performances were also included. All performances were synthesized on the Roland RD-250s digital piano.

The judges in the experiment gave the highest average quality rating to EA, followed by E11, S1, S3, S9, S2, and SA.

Together, the two experiments demonstrate that an average timing profile tends to sound better than average, even though it lacks individuality. Experiment 2 also indicates that, with certain exceptions such as E11, the timing profiles of acoustically recorded expert performances do not sound good when they are imposed onto a synthetic "carrier" with fixed dynamics. Timing may not be independent of dynamics and other aspects of expression. For details and discussion, see the reference listed below.

I would be interested in your thoughts and comments on these experiments. To send me an e-mail message, click below:

Send a message to Bruno (repp@haskins.yale.edu)

References

Repp, B. H. (1997). The aesthetic quality of a quantitatively average music performance: Two preliminary experiments. Music Perception, 14, 419-444.