Coordinating multiple actions across different bodily effector systems is central to most everyday behavior. Research on multiple-action control has traditionally focused on the execution of concurrent or sequential actions, often emphasizing the difficulties of performing two actions simultaneously compared to executing a single action. Accordingly, traditional models of multiple-action control have primarily been developed to explain dual-action costs. However, daily activities sometimes require the selective execution of one action while actively suppressing additional prepotent response tendencies. Recent studies have revealed paradoxical challenges in executing a single action relative to two (i.e., dual-action benefits), due to the erroneous co-activation of a second, unintended response. This has been evidenced by frequent false-positive co-execution of the to-be-inhibited action in single-action conditions. These findings have been interpreted within a response-selection crosstalk framework, which explains co-activation as "spill-over" of activation between overlapping response-relevant codes. The major aim of the present dissertation was to identify key determinants and further specify the mechanisms underlying such erroneous response co-activation during multiple-action control. To this end, five laboratory studies (seven experiments in total; total *N* = 285) were conducted in which participants controlled two response modalities: saccadic eye movements and manual key presses (Studies 1-4) or saccades and vocalizations of directional words (Study 5). Participants were instructed to make a spatial (left vs. right) response with either one modality, both together, or not at all (Study 4 only). The key dependent measure was false-positive execution errors in single-action (and no-action) conditions indicative of erroneous action co-activation. Such errors emerged consistently across all studies, underscoring the robustness of the phenomenon. Several modulating factors were identified. Across all studies, saccadic eye movements were consistently more prone to co-activation than manual or vocal responses, indicating inherent differences in effector-system prepotency. Study 1 demonstrated the impact of contextual factors by manipulating block structure. In pure blocks, participants performed the same response type (e.g., single manual) throughout the block, whereas in mixed blocks, response demands varied randomly. False-positive errors were more frequent in mixed blocks, indicating that general anticipation of action execution increases action prepotency. Moreover, false-positive executions in mixed blocks were more likely following a trial that engaged the same effector system, regardless of the specific direction of the previous response. Study 2 varied the preparation interval between a cue signaling response demand and an imperative stimulus specifying response direction. False-positive errors were substantially more likely with short (vs. long) preparation intervals, indicating that participants were able to proactively downregulate action prepotency at the level of the response modality when given sufficient time. Study 3 systematically varied stimulus-response translation ease by using different types of stimuli. This manipulation did not significantly affect false-positive error rates, suggesting that co-activation arises from the selection of another response, rather than being automatically triggered by stimulus characteristics. Study 4 further supported this interpretation of response-driven co-activation by showing higher false-positive error rates during active single-response execution than in no-action trials requiring complete suppression. Finally, Study 5 generalized the co-activation effect to oculomotor-vocal response combinations and showed that moderate practice significantly reduced false-positive errors, suggesting improved distinction between single-action and dual-action representations. Together, the present work identified key conditions under which the selective execution of a single action becomes challenging due to the automatic co-activation of additional action tendencies. Based on the present results, a crosstalk framework of integrated response selection was refined to incorporate different sources of action prepotency and a response-driven co-activation mechanism. Future studies of such co-activation effects offer a promising lens through which to study the (dis-)integration of complex action representations.
Jens Kürten (Thu,) studied this question.