Clustering in music: an appraisal of task factors

The pitch and temporal intervals of tune fragments, taken from folk music, were analyzed in terms of gestalt similarity principles and studied as determinants of music grouping structure. In experiment 1, the listeners were required to divide each tune into its 'natural' parts. Their ability to segment the tunes closely matched the boundaries generated by the clustering determinants under study, this consistency being particularly high for musicians. In experiment 2, a probe recognition task was used in which subjects were presented with a tune fragment followed by a true or false probe; true probes fell either within or across the boundaries studied in experiment 1. Since the probe was presented after the tune fragment, the differential ease of verifying within and across probes was taken to reflect the part structure of the tune held in memory. In order to study the part structure representation that is being built during music listening, a third experiment was devised by presenting the probe prior to the tune fragment and having subjects monitor the probe while listening to the fragment. Of the last two paradigms, only the latter was found to be sensitive to cluster boundaries, as defined along the temporal dimension. The entire set of results suggests that there does not seem to be a basic format of musical representation that is uniquely specified from the output of the clustering process.