How much can we learn from double dissociations?

It is sometimes assumed that double dissociations between performance in Tasks A and B indicate that different underlying cognitive mechanisms are implicated in each task. According to this viewpoint, double dissociations are a powerful tool for probing the modularity of the mind. The validity of the double dissociation inference, and related inferences has been widely discussed (e.g., Dunn and Kirsner, 1988; Shallice, 1988). Here we focus on a particularly simple type of example, which nonetheless challenges the general use of the inference from double dissociation to the modular structure of the mind. Suppose that the mechanisms underlying A and B overlap almost completely, except for one infinitesimally small component, which is specific to A; and one infinitesimally small component, specific to B. Sometimes damage to the system will leave these two special components intact − and perhaps just cause general degradation in performance for both tasks. Sometimes, damage will knock out both components, causing both tasks to be poorly performed. But, equally, damage that knocks out just one of the special components, or impairs it, will lead to dissociations between performance in one of the tasks. The double dissociation may be as extreme as one wishes. So it would seem wrong to conclude that Tasks A and B are subserved by different systems – instead they are subserved by the same system, except for the tiny specialized components. This seems to be more than a theoretical possibility. Suppose we knew nothing of the physiology of the digestive system from anatomical investigation, and instead attempted to infer it from ‘double dissociations’ in the behaviour of that system. Suppose that person 1 is allergic to prawns but can eat peanuts; and person 2 is allergic to peanuts but can eat prawns. We might conclude that prawns and peanuts are digested by separate systems. But of course they are digested by the same system – although there are presumably subtle chemical processes, presumably quite late in the process of digestion, that differ between the two. The point can be sharpened further. Even the two ‘specialized’ components need not be involved purely in one task but not the other. Instead, the two components may both be involved in each task, but to different degrees. Then an impairment to one component may leave it still able to function well enough to contribute to the performance of Task A, but not well enough to facilitate the successful completion of Task B; or vice versa. For example, consider two athletes, one who gets blisters from overzealous practice in throwing the javelin; another who gets blisters from overzealous