Three possible explanations for attentional bias effects in anxious subjects have been formulated: the threat-relatedness hypothesis, the emotionality hypothesis and the concern-relatedness hypothesis. In order to investigate these three hypotheses, an experiment was carried out with 33 obsessive compulsive (OC) patients and 29 normal controls. Both groups colour-named a Stroop card with 5 word sets: neutral words and 4 emotional word sets (a 2 × 2 matrix of words, related/unrelated to obsessive compulsive disorder and positively/negatively valenced). In line with previous studies, OC patients selectively attended to negative OC-related cues; this supports the threat-relatedness hypothesis. Although the set-up of the experiment was similar to the Mathews and Klug (1993, Behaviour Research and Therapy, 31, 57-62) study, no evidence was found for the concern-relatedness hypothesis, i.e. the OC patients did not show an attentional bias for positive OC-related words. Two possible reasons for these contradicting findings are discussed.