The Stroop task, in which subjects must name the color of letters that spell color words different than the color-to-be-named, provides an important experimental paradigm for the study of selective attention. Cerebral blood flow activation studies have not always demonstrated consistent activation patterns; inconsistent results may reflect nonspecific responses, such as arousal or anticipation, rather than cerebral networks specific to Stroop interference processing. In order to identify regions consistently implicated in Stroop interference processing, we undertook two experiments with a Stroop interference paradigm and contrasting lexical and nonlexical control conditions. In our first experiment, standard Stroop stimuli, e.g., the word "RED" displayed in a green font, were contrasted with color naming of the font of noncolor words and color naming of a false font. In our second experiment, we compared Stroop stimuli with colored symbols and a control condition designed to elicit nonspecific interference-taboo words displayed in color fonts. Only two brain regions showed a consistent CBF change in both experiments. Activation in the left inferior frontal gyrus reflected processing more specific to the Stroop task, while deactivation in the right superior temporal gyrus occurred for the Stroop and the taboo conditions, consistent with more nonspecific processing. Activation in the anterior cingulate cortex occurred in only one comparison in one experiment and may not reflect functions central to overcoming Stroop interference. Copyright 1997 Academic Press.