In many dual-task situations, responses to the second of two tasks are slowed when the time between tasks is short. The response-selection bottleneck model of dual-task performance accounts for this phenomenon by assuming that central processing of the second task is blocked by a bottleneck until central processing of Task 1 is complete. This assumption could be called into question if it could be demonstrated that the response to Task 2 affected the central processing of Task 1, a backward response compatibility effect. Such effects are well-established in younger adults. Backward compatibility effects in older (as well as younger) adults were explored in two experiments. The first experiment found clear backward response compatibility effects for younger adults but no evidence of them for older adults. The second experiment explored backward stimulus compatibility and found similar effects in both younger and older adults. Evidence possibly consistent with some pre-bottleneck processing of Task 2 central stages also was found in the second experiment in both age groups. For younger adults, the results provide further evidence falsifying the claim of an immutable response selection bottleneck. For older adults, the evidence suggested that Task 2 affects Task 1 when there is stimulus compatibility but not when there is response compatibility. Under most circumstances, people cannot perform two tasks at exactly the same time without interference, even when both tasks are very simple. As the temporal overlap between the two tasks increases, the time to perform the second task increases monotonically. Conversely, in many situations for each additional millisecond that the stimulus of the second task is delayed after the onset of the stimulus of the first task (the stimulus onset asynchrony or SOA), the reaction time (RT) to the second task drops by 1 ms. By contrast, the RT to the first task is largely unaffected by SOA. The slowing of the second task at short SOAs often is called the psychological refractory period (PRP) effect by analogy with the time after firing when a neuron is unresponsive to further input (Telford, 1931;Vince, 1948;Welford, 1952). The PRP effect has been replicated in a large number of experiments, thus deserving the status of one of the few laws of contemporary cognitive psychology. Pashler (1994Pashler ( , 1998 provides reviews and overviews of the research and theory (also see Spence, 2008;Tombu & Jolicoeur, 2005).The most successful account of the PRP effect is the response-selection bottleneck (RSB) model, the principal tenet of which is that central processing can be performed for one and only one task at any particular time. Central processing of one of the tasks must be delayed until central processing of the other task is complete. This leads to the prediction that each millisecond by which the arrival of the second task is delayed will be 1 ms that it will not have to wait for the central processing mechanism to become free. Central processing involves processes of mapping...