This paper is concerned with the perception of small-scale timing changes in musical sequences. The control and expressive function of these have been studied quite extensively from a production perspective, but not much is known about listeners' ability to detect them. A pilot study and two experiments are reported which investigate the detectability of different amounts of timing change in different sequential positions, different pitch contexts, and against the background of both metronomic and expressive comparisons. The results show that listeners are able to perceive as little as 20 ms lengthening in the context of notes lasting between 100 and 400 ms, and that this threshold appears not to be a function of base duration in this range. Sequential position and pitch structure influence the detectability of timing changes to a limited extent, for which some possible explanations are offered. A case is made for regarding timing in music as both a medium to convey structure and an object in its own right, suggesting that it may be perceptually organized in two different ways--as the consequence of a structural interpretation and as a directly registered quantity.