Redefining dissociations for a higher-powered neuropsychology: A methodological proposal

Dissociations are a core concept in neuropsychology, central to the effort to discover the organisation of the mind. Dissociations are usually investigated at the level of the ‘single-case’ (e.g. a brain-damaged patient). A range of statistical tests have now been developed for the comparison of single cases against control samples. Unfortunately, case-control comparisons of this sort are inherently low-powered. As Crawford, Garthwaite, & Ryan (2011) have argued, “anything that can increase statistical power to detect deficits or dissociations should be encouraged (provided that it does not achieve this at the cost of failing to control the Type I error rate)”. I propose that a simple adjustment could be made to currently-accepted statistical criteria, which would increase the power to detect neuropsychological dissociations, under a range of practically-useful conditions, whilst retaining appropriate control over Type I error. In this seminar, I will present my proposal and illustrate its consequences via Monte Carlo simulations. I hope to elicit critical discussion from HCN colleagues, to evaluate the idea, and explore possible problems.