Panic disorder patients, obsessive-compulsive disorder (OCD) patients, and normal control subjects performed a computerized Stroop color-naming task in which they named the colors of panic-threat words (e.g. Collapse), general-threat words (e.g. Infectious), positive words related to panic (e.g. Relaxed), and neutral words (e.g. Sleepy). Idiographic stimulus selection ensured their personal emotional relevance for each subject. In accordance with prediction, panic patients, but not OCD patients, exhibited greater interference for panic-threat words than for positive words related to panic and for neutral words. Panic patients did not respond differentially to panic-threat and general-threat words. Complexities concerning attentional bias research in the anxiety disorders are discussed.