Standard response-selection bottleneck models of performance in psychological refractory period (PRP) tasks suggest that a Task-1 response should be unaffected by the Task-2 response in the same trial, because selection of the former finishes before selection of the latter begins. Contrary to this conception, we found backward response-level crosstalk effects in which Task-2 response force requirements influenced the force-time dynamics of Task-1 responses. In two PRP experiments, Task 2 required a hard or soft key press response. Task-1 responses were harder when the upcoming Task-2 response was to press hard rather than soft, suggesting some activation of Task-2 response parameters before Task-1 processing reached the final ballistic stage of motor output. A third experiment using a single-task flankers paradigm showed that this effect did not arise from automatic activation of responses by the stimuli associated with hard and soft responses. This backward response-level crosstalk extends previous findings of backward crosstalk in the PRP paradigm by showing that crosstalk can affect motor output as well as response time and that such crosstalk can arise even when the two tasks being performed are not semantically related.