Synchronizing with a metronome that is steady except for minor perturbations can be accounted surprisingly well by a linear phase correction mechanism (Repp, 2001; Semjen, Schulze, & Vorberg, 2001). How van such a model be extended to account for two performers synchronizing with each other? I present an extension of the phase correction synchronization model (Vorberg & Schulze, 2002) to duet performance. Consider the general case of two subjects, tapping at a given rate and trying to keep synchronized with each other, which may be seen as paradigmatic for two musicians interacting with each other by adjusting the phase of the central timer. Surprisingly, predictions are analogous to those for synchronizing with a stable metronome. The model is tested with experiments in which a single performer taps in synchrony with an isochronous tone sequence produced by the computer. The computer simulates the other half of the duet, synchronizing its sounds with the registered taps by adjusting phase as dictated by the model. This guarantees that the model holds for at least one of the performers, and allows manipulating the gain parameter experimentally. I present results of duet synchronization in experiments that factorially varied tempo and gain.