Executing two actions at the same time usually results in performance costs. However, recent studies have also reported dual-action benefits: under certain circumstances, doing two things at once can yield improved performance in terms of lower reaction times and lower error rates. Here, we present a novel theoretical framework that can account for such effects by assuming that 1) performing only one of two possible actions may necessitate the inhibition of the initially activated, but unwarranted second action, leading to single-action costs and 2) dual-action execution may benefit from statistical facilitation based on redundant, response-related features of the stimulus (i.e., dual actions may be triggered as soon as either enough information for the execution of response 1 or enough information for the execution of response 2 has been extracted from the stimulus). The resulting model of multiple action control allows for context-dependent inhibitory coding (i.e., actions are sometimes cognitively represented in terms of what not to do) and handles inhibition via explicit cognitive control codes; it is complementary to existing theories of countermanding - in particular, the Pause-Then-Cancel framework by Diesburg and Wessel (2021) - but includes further testable specifications of processes (and their components) which were previously only treated implicitly.