The ability to change our behavior is one that we frequently exert, although determining the mechanisms by which we do so is far from trivial. Task switching is a useful experimental paradigm for studying cognitive control functions. Switching between tasks is associated with a decrement in performance, or “switch-cost,” relative to repeating the same task. We have previously demonstrated that this cost is dependent on switching from performing one task to performing another; changing only our intended performance does not elicit the same performance deficit. Using event-related potentials (ERPs), we dissociated two electrophysiological indices mirroring this behavioral distinction [Astle, D. E., Jackson, G. M., & Swainson, R. Dissociating neural indices of dynamic cognitive control in advance task-set preparation: An ERP study of task switching. Brain Res, 1125 , 94–103, 2006]. However, what was unclear were the specific aspects of performance that were critical for triggering the neural mechanisms associated specifically with switching from a previously performed task. Two candidate aspects were: (i) that performance required a physical response and (ii) that the two tasks shared their responses (they had bivalent response mappings). The present study therefore compared three separate groups to explore the effects of these different aspects of performance. Each group completed the same basic task-switching paradigm, but with either an overt response or covert response, and either switching between tasks that shared their responses (bivalent response mappings) or had separate responses (univalent response mappings). When comparing precue-locked ERPs, we observed three separable components: one common to all three groups, one which primarily dissociated overt from covert responding, and one which primarily dissociated bivalent from univalent responding. We therefore concluded that changing our behavior engages at least three dissociable mechanisms. Interestingly, in the overt conditions, residual switch-costs were absent; in addition, therefore, we concluded that it is possible to engage cognitive control in advance, such that the new behavior is as efficient as were the subject to have repeated the old behavior.