In many situations perception and production of time intervals seem to go hand-in-hand. For instance, standard deviations for synchronization in tapping are about the same magnitude as the estimated perceptual thresholds for timing perturbations in isochronous sequences.
For continuous tempo drift, however, there seem to be inconsistency between estimated perceptual thresholds and estimates of long-term changes in production. For instance, several studies have suggested the threshold of tempo drift to be about 2 % for nominal IOIs of about 500 ms, which is in the same range as the threshold for stepwise tempo changes. However, studies that have estimated linear tempo drift in production data report values much lower, about 0.2 %
In discrimination experiments Dahl and Granqvist recently have found listeners to be biased towards either increasing or decreasing tempo, meaning that some perceive a small increase in tempo as decreasing, and vice versa. The magnitude and sign of the perceptual bias differed between individuals, but intra-individual variations between test sessions were small. The perceptual threshold was about 0.3%, which corresponds well with the 0.17% for longer sequences reported by Madison and Merker (2004). These thresholds also agree well with values of drift in production data.
This paper will review the literature on perception and production of tempo drift and compare estimates of linear drifts from production data to reported threshold estimates.