Interpersonal timing is important for creating a cohesive music performance in small ensembles without a conductor. Wing et al. (2014) proposed that phase correction underlies ensemble timing in classical string quartets. Correction gains for the members of two quartets were close-to-optimal, as they minimised group asynchrony variance. To investigate whether adjustments change depending on other players’ phase correction and leadership roles, we simulated a quartet (two violins, a viola, and a cello) performing a homophonic Haydn excerpt. By means of a linear phase-correction model, the notes of three virtual players were synchronised to each other and to the participant tapping on a MIDI drum pad. Participants produced either the notes of the first or second violin part. In different trials, the correction parameter for the other violin was set to either under- or over-correct to the participant, and the timekeeper noise parameter was set to either more or less than the other virtual players. Results show that the participant’s timing correction changes according to the correction parameter of the virtual violinist, but not according to the level of noise. Participants show higher dependency when the other violin corrects less (and vice versa), congruently with the hypothesis of close-to-optimal group asynchrony variance minimization.