When is category specific in Alzheimer's disease?

Mixed findings have emerged concerning whether category-specific disorders occur in Alzheimer's disease. Factors that may contribute to these inconsistencies include: ceiling effects / skewed distributions for control data in some studies; differences in the severity of cognitive deficit in patients; and differences in the type of analysis (in particular, if and how controls are used to analyse single case data). We examined picture naming in Alzheimer's patients and matched elderly healthy normal controls in three experiments. These experiments used stimuli that did and did not produce ceiling effects / skewed data in controls. In Experiment 1, we examined for category effects in individual DAT patients using commonly used analyses for single cases (chi(2) and z-scores). The different techniques produced quite different outcomes. In Experiment 2a, we used the same techniques on a different group of patients with similar outcomes. Finally, in Experiment 2b, we examined the same patients but (a) used stimuli that did not produce ceiling effects / skewed distributions in healthy controls, and (b) used statistical methods that did not treat the control sample as a population. We found that ceiling effects in controls may markedly inflate the incidence of dissociations in which living things are differentially impaired and seriously underestimate dissociations in the opposite direction. In addition, methods that treat the control sample as a population led to inflation in the overall number of dissociations detected. These findings have implications for the reliability of category effects previously reported both in Alzheimer patients and in other pathologies. In particular, they suggest that the greater proportion of living than nonliving deficits reported in the literature may be an artifact of the methods used.