What Can we Infer from Double Dissociations?

The joint aim of cognitive psychology and cognitive neuropsychology is to describe the functional architecture of the human mind. That is, to identify, characterize, and enumerate the fundamental mental processes underlying human behaviour. In both of these fields, behaviour is measured in terms of performance on a wide range of tasks that are presumed to involve different combinations of mental functions. The two fields differ primarily in the nature of the conditions under which these tasks are performed. Specifically, whereas cognitive psychology generally involves systematic manipulation of variables in experiments involving non-brain damaged participants, cognitive neuropsychology is concerned with the impact of brain damage on task performance. Although they are, in a sense, the object of study, mental processes are not directly observable. Rather their existence and function must be inferred from the manner in which task performance changes from treatment to treatment, involving different levels of an experimental variable or different forms of brain damage. This is a primary methodological problem for both cognitive psychology and cognitive neuropsychology, and, in an attempt to solve this problem, researchers in both fields have placed increasing reliance on the logic of dissociations (Coltheart, 1985; Crowder, 1972; Shallice, 1988; Tulving, 1983; Vallar, 1999). Dissociations are used to infer the existence of separate mental processes. There are two main types, single and double. Let A and B be two tasks and let a and b be two manipulations, variables or factors. A single dissociation is observed if a affects performance on A but not on B. A double dissociation is observed if, in addition, b affects performance on B but not on A. In cognitive neuropsychology, manipulation a would usually correspond to a comparison between a patient or group of patients who are impaired on A but not B, and normal controls, who are unimpaired on both A and B. Similarly, manipulation b would correspond to a comparison between another patient or group of patients impaired on B but not A and normal controls. Both single and double dissociations invite the inference that there is an underlying mental function required by A but not by B. In addition, a double dissociation invites the converse inference, that there is an underlying mental function required by B ut not by A. Dissociations have long been used in neuroscience as a criterion for localizing mental function in the brain (Teuber, 1955; Young et al., 2000). In the late 1960’s, this logic was employed for the first time by cognitive psychologists