Introduction

Several authors have suggested that emotion scientists could fruitfully examine the underlying structure, or functional architecture, of human emotion systems using cognitive neuropsychological techniques (Davidson, 1992; Scherer, 1993; Lane et al., 2000). According to Bub (1994), ‘The general methodological problem for neuropsychology is how evidence, in the form of patterns of cognitive deficits, bears on theory, in the form of rival functional architectures’. Analogously, we can ask how evidence, in the form of patterns of emotion deficits, bears on theory, in the form of rival functional architectures of emotion (Mathews & MacLeod, 1994; Scherer & Peper, 2001).

In collecting evidence, neuropsychologists design experiments to elicit dissociations in task performance. Dissocations can be single (task A normal, task B abnormal), or multiple (tasks A, B normal, tasks C, D abnormal). Double dissociations are obtained when task A is normal, task B abnormal in patient P1; task B normal, task A abnormal in patient P2 (Shallice, 1988). Double dissociation evidence is regarded as the most compelling of all. This is because partial lesions allow dissociations to arise from resource artefacts, as one task demands more of some computational resource of a processing component than another (Shallice, 1988). Resource artefacts can be ruled out by double dissociations, and only systems that contain a high degree of functional specialization can produce strong double dissociations (Shallice, 1988). For this argument to hold, two relatively uncontroversial assumptions must be made:

1. That impaired processing is explicable in terms of the same model as normal processing, except that certain parameters of the model are changed (i.e. processes are not fundamentally reorganized following damage), and

2. […]