When tapping a desired frequency, subjects tend to drift away from this target frequency. This compromises the estimate of the correlation between inter-tap intervals (ITIs) as predicted by the two-level model of Wing and Kristofferson which consists of an internal timer ('clock') and motor delays. Whereas previous studies on the timing of rhythmic tapping attempted to eliminate drift, we compared the production of three constant frequencies (1.5, 2.0, and 2.5 Hz) to the production of tapping sequences with a linearly decreasing inter-tap interval (ITI) (corresponding to an increase in tapping frequency from 1.5 to 2.5 Hz). For all conditions a synchronization-continuation paradigm was used. Tapping forces and electromyograms of the index-finger flexor and extensor were recorded and ITIs were derived yielding interval variability and model parameters, i.e., clock and motor variances. Electromyographic recordings served to study the influence of tapping frequency on the peripheral part of the tap event. The condition with an increasing frequency was more difficult to perform, as evidenced by an increase in deviation from the intended ITIs. In general, tapping frequency affected force level, inter-tap variability, model parameters, and muscle co-activation. Parameters for the condition with a decreasing ITI were comparable to those found in the constant frequency conditions. That is, although tapping with an intentional drift is different from constant tapping and more difficult to perform, the timing properties of both forms of tapping are remarkably similar and described well by the Wing and Kristofferson model. © 2008 The Author(s).