Comparing real-time and retrospective perceptions of segmentation in computer-generated music

We confirm our earlier observations (Bailes & Dean, 2005) that listeners can efficiently detect segmentation in computer-generated music which is primarily timbral and textural, atonal and non-metrical. We show that contextual aspects influence this detection, such as those in force in the ramp effect whereby an addition or increase of sound is more readily detected than an equivalent removal or reduction. Listeners are less efficient in detecting a change of segment after a short than after a long segment. They are more efficient in detecting change in real-time when allowed to repeat the detection task. Real-time and retrospective perceptions correlate well