Performance variability on neuropsychological measures is not a unitary phenomenon, and different measures (consistency, dispersion, diversity) evaluate separate elements of variability. It has been suggested that increased variability may be a specific attribute of frontal lobe pathology. This hypothesis was tested in 2 matched groups of demented subjects, 8 with dementia of the Alzheimer type (DAT), 5 with frontal lobe dementia (FLD), compared with 10 elderly normal controls (ENC). A Stroop test and Reaction Time measures were administered weekly for 5 weeks to all subjects. Both measures contained three subtests varying in degree of complexity. The results from the Stroop task indicated that the FLD group showed significantly greater variability on measures of consistency (fluctuations over time) and diversity (between participant variability) regardless of the complexity of the subtest. For the Reaction Time subtests, measures of consistency and diversity showed significantly greater variability in FLD, but were affected in a different pattern. Greater variability in terms of consistency of performance was manifested only in the more attentionally demanding of the Reaction Time subtests (Choice Reaction Time, CRT). On the measure of diversity, variable performance was found to be greater on the Simple Reaction Time (SRT) subtest than on the more effortful CRT. Dispersion (within participant variability) was only assessed on the reaction time subtests. The results indicate no significant evidence for an increase in dispersion for the FLD patients. The hypothesis that variability will be increased in frontal lobe dementia is thus confirmed, and the independence of the three forms of variability measurement is demonstrated in dementia subjects. (JINS, 2002, 8, 360–372.)