Local changes in action demands (e.g., due to an invalid cue) necessitate changes to pre-formulated action plans. Crucially, differential patterns of reaction-time costs associated with switches between distinct types of actions (e.g., single vs. dual) can shed light on the underlying cognitive representations. Here, we report on a series of experiments utilizing different types of stimuli in a precuing paradigm combining single actions, dual actions and no-go trials. In all studies, performance (expectedly) suffered when precue and signal were incongruent. Moreover, while there were dual-action costs in incongruent trials, congruent trials were associated with dual-action benefits. Depending on the type of stimulus, we also observed benefits for certain types of cue-to-signal changes: given a single stimulus lacking any dimensional overlap with the response (a colored circle), switching between single-left and single-right button presses (or vice versa) was faster than any other type of incongruency resolution. Given two separate, lateralized stimuli (geometric shapes) that spatially corresponded to the lateralized responses, on the other hand, direction-switch benefits in single-action trials were absent, but we measured no-go to dual-action switch benefits in dual-action trials. We believe that the combined findings of these experiments can be explained using a hierarchical model of action representation that is based on structured bundles of distinctive action features (the Task-Tier-Type-Token Framework).