In two experiments the hypothesis that caffeine reduces Stroop interference
was tested. In the first experiment interference was measured as the reduc
tion in performance when subjects had to indicate the numerosity of strings
of incongruent digits, relative to neutral-symbol strings. In the second e
xperiment the incongruent condition consisted of naming the color of words
referring to incongruent colors, and was compared to color naming of neutra
l strings. In the number-digit task 250 mg caffeine reduced interference at
the level of error rates, relative to placebo. In the color-word task inte
rference was reduced at the level of reaction times. These results were obt
ained with blocked presentations of incongruent and neutral conditions. The
color word experiment also contained a condition in which neutral and inco
ngruent trials were mixed within one sequence. In this mixed condition caff
eine still reduced overall reaction times, but no longer specifically inter
ference. It is argued that this dissociation reflects a caffeine-induced in
crease in flexibility. The results are discussed in relation to failed prev
ious attempts to demonstrate increased selectivity under caffeine using non
-Stroop tasks, the importance of including pretreatment sessions to detect
artificial effects, and the possible contribution of withdrawal effects. (C
) 1999 Elsevier Science Inc.