Simple dissociations for a higher-powered neuropsychology

Dissociations in cognitive neuropsychology are often investigated at the level of the single-case, and formal criteria exist for the detection of dissociations, and their sub-classification into 'classical' and 'strong' types. These criteria require a patient to show a frank deficit on one task (for a classical dissociation) or both tasks (for a strong dissociation), and a significantly extreme difference between tasks. I propose that only the significant between-task difference is logically necessary, and that if this simple criterion is met, the patient should be said to show a dissociation. Using Monte Carlo simulations, I show that this simplification increases the power to detect dissociations across a range of practically-relevant conditions, whilst retaining excellent control over Type I error. Additional testing for frank deficits on each task provides further qualifying information, but using these test outcomes to categorise dissociations as classical or strong may be too uncertain to guide theoretical inferences reliably. I suggest that we might instead characterise the strength of the dissociation using a continuous index, such as the effect size of the between-task difference.