The study investigated biases in selective attention to emotional face stimuli in generalized anxiety disorder (GAD) and depressive disorder, using a modified probe detection task. There were 4 face types: threatening, sad, happy, and neutral. Measures of attentional bias included (a) the direction and latency of the initial eye movement in response to the faces and (b) manual reaction time (RT) to probes replacing the face stimuli 1,000 ms after their onset. Results showed that individuals with GAD (without depressive disorder) were more likely to look first toward threat faces rather than neutral faces compared with normal controls and those with depressive disorder. They also shifted their gaze more quickly toward threat faces, rather than away from them, relative to the other two groups. There were no significant findings from the manual RT data. Implications of the results for recent theories of clinical anxiety and depression are discussed.