Double Dissociation: Understanding its Role in Cognitive Neuropsychology

The paper makes three points about the role of double dissociation in cognitive neuropsychology. First, arguments from double dissociation to separate modules work by inference to the best, not the only possible, explanation. Second, in the development of computational cognitive neuropsychology, the contribution of connectionist cognitive science has been to broaden the range of potential explanations of double dissociation. As a result, the competition between explanations, and the characteristic features of the assessment of theories against the criteria of probability and explanatory value, are more visible. Third, cognitive neuropsychology is a division of cognitive psychology but the practice of cognitive neuropsychology proceeds on assumptions that go beyond the subject matter of cognitive psychology. Given such assumptions, neuroscientific findings about lesion location may enhance the value of double dissociation in shifting the balance of support between cognitive theories. Cognitive neuropsychology uses patterns of impairment and sparing in patients following brain injury in order to constrain theories of normal cognitive structures and processes. It emerged as a distinctive research programme in the 1960s and by the end of the 1980s had its own journal (volume 1, 1984) and its own textbook (Ellis and Young, 1988/1996). In the practice of cognitive neuropsychology, the evidential value of double dissociation as support for claims about separate modules in the normal cognitive system has been highlighted. This highlighting is sometimes interpreted as the manifestation of a methodological assumption about a special logic of cognitive neuropsychology. For example, Karalyn Patterson and David Plaut say that ‘the gold standard was always a double dissociation’ (2009, p. 43) and describe ‘traditional cognitive neuropsychology logic’ as resting on an assumption that ‘the functional organization of cognition can be unequivocally revealed by dissociation’ (p. 44). I reject the claim that the practice of cognitive neuropsychology is based on assumptions about a special logic. Specifically, I reject the claim that cognitive neuropsychology relies on an assumption that the inference from a pattern of impairment and sparing to a structure of separate cognitive modules is underwritten by a special deductive rule of double dissociation inference. I am grateful to Anne Aimola Davies, Max Coltheart, Kim Plunkett, Nick Shea, Michael Smithson, and two anonymous referees, for comments on an earlier version of this paper. Max Coltheart was one of the founding editors of Mind & Language and has written with distinction and influence on the themes of this paper—the aims and assumptions of cognitive neuropsychology, double dissociation, computational modelling, and the relationship between cognitive psychology and neuroscience. My intellectual debts to him will be evident on every page and the paper is dedicated to him with gratitude. Address for correspondence: Corpus Christi College, Merton Street, Oxford OX1 4JF, UK. Email: [email protected] Mind & Language, Vol. 25, No. 5 November 2010, pp. 500–540. © 2010 Blackwell Publishing Ltd Double Dissociation 501 One reason to reject the claim is that it is difficult to locate texts in which cognitive neuropsychologists unambiguously claim that their approach has a distinctive logic with proprietary rules of inference.1 A further reason is that it is a familiar point about science in general that there is no logically valid deductive inference from evidence to explanatory theory. To be explanatory, a theory must go beyond a summary of the evidence and so cannot be entailed by the evidence. Within the narrower domain of psychology, it is well understood that there is no logically valid deductive inference from data, such as reaction time data, to an explanatory theory about cognitive structures and processes. It is very unlikely that cognitive neuropsychologists regard their research programme as being different from all the rest of empirical science and take themselves to have access to evidence with magical properties. There are good reasons to reject the claim that the practice of cognitive neuropsychology is based on assumptions about a special logic. Nevertheless, that claim seems to be part of the ‘received wisdom’ about cognitive neuropsychology, in the sense that many people believe—or many people believe that many people believe—that its practice does rest on such assumptions.2 This hypothesis about received wisdom helps to render intelligible the existence of influential papers with titles including, ‘What can we infer from double dissociations?’ (Dunn and Kirsner, 2003), ‘What do double dissociations prove?’ (Van Orden, Pennington and Stone, 2001), and ‘Why double dissociations don’t mean much’ (Juola and Plunkett, 2000). More precisely, the hypothesis helps to explain the existence of those papers, given the notable absence of papers called ‘What can we infer from evidence?’, ‘What does data prove?’, or ‘Reaction times don’t mean much’. If the practice of cognitive neuropsychology really had relied on a special logic with a special rule of double dissociation inference, this would have had serious consequences for the research programme. Over the last twenty years or so, developments in computational modelling, and particularly in connectionist cognitive science, have revealed a broad range of potential explanations of double dissociation. It has been demonstrated that the special rule of inference is logically invalid. If cognitive neuropsychology had been relying on a special logic including the special rule then these developments might have led, if not to the death of cognitive neuropsychology as originally conceived, at least to its entry into a persistent vegetative state (PVS).3 1 When Max Coltheart (1985, p. 17) speaks of ‘remorselessly pursuing the logic of cognitive neuropsychology’, he is not alluding to special inference rules. Rather, he is indicating the interest, for cognitive neuropsychology, of patients whose patterns of impairment and sparing would be readily explained in terms of particular proposals about separate modules in the normal cognitive system. 2 I am indebted to David Chalmers for discussion about the notion of received wisdom. 3 Here, I borrow the words of an anonymous referee. Following the publication of the Handbook of Cognitive Neuropsychology (Rapp, 2001), the journal Cognitive Neuropsychology published a review (Harley, 2004) under the challenging title, ‘Does cognitive neuropsychology have a future?’. © 2010 Blackwell Publishing Ltd