The emotional Stroop task has been the most widely used task to examine attentional bias to emotionally salient stimuli. In one format of this task, words are presented to participants in a mixed randomized or quasi-randomized sequence. Using a mixed smoking Stroop task, we have previously demonstrated that smokers are slower to respond to words which follow smoking-related words than words which follow neutral words. Here we show that this carry-over effect is present in heroin addicts-but not control subjects-in a heroin Stroop task, and in normal subjects in a stress Stroop task. Thus, the effect generalizes to other populations. In addition, an examination of the studies that have collected data from both mixed and blocked formats provides converging evidence for the presence of carry-over effects. We discuss the implications of the carry-over effect for research using the emotional Stroop task.