Sensorimotor timing behaviors typically exhibit an elusive phenomenon known as the negative asynchrony. When synchronizing movements (e.g. finger taps) with an external sequence (e.g. a metronome), people's taps precede event onsets by a few tens of milliseconds. We recently reported that asynchrony is less negative in participants with lower asynchrony variability. This indicates an association between negative asynchrony and variability of timing. Here, in 24 metronome-synchronization data sets, we modeled asynchrony series using a sensorimotor synchronization model that accounts for serial dependence of asynchronies. The results showed that the modeling well captured the negative correlation between the mean and SD of asynchrony. The finding suggests that serial dependence in asynchronies is an essential mechanism of timing variability underlying the association between the mean and SD of asynchrony.