The Stroop test comes in different variations, but all of them index performance on a basic task, like color or picture naming, when it is carried out by itself versus when it is performed in the presence of conflicting or incongruent stimuli. The present study examined the hypothesis that Stroop interference--the cost of performing one task in the presence of another--is a general as opposed to a test-specific index of cognitive flexibility. A second goal was to examine changes in Stroop test performance in old age. A group of 129 healthy older adults (> or = 65 years of age) were assessed on the color- and picture-word Stroop test, as well as on a battery of neuropsychological tests. Subjects' performance on each card of both Stroop tests, and various derived (differences and ratios) scores, were used to prepare age-group norms. The use of the norms is illustrated with findings from previous studies. Regression analyses showed age-changes in several aspects of Stroop test performance. Hierarchical multiple regression analyses, and causal modeling showed an age effect on Stroop interference only on the picture-word test but not on the color-word test. Exploratory factor analysis of the Stroop data and the neuropsychological test data revealed different factor loadings for the color- and picture-word test. The combined findings suggest that the color- and picture-word Stroop test measure different cognitive functions, at least in old age.