The present studies examined attentional bias for photographed faces and household objects among individuals diagnosed with generalised social phobia using a letter-probe computer task. Study 1 was conducted to replicate previous findings showing evidence of avoidance for faces among individuals diagnosed with generalised social phobia using a similar methodology as Mansell, W., Clark, D., Ehlers, A., and Chen, Y. P. (1999, Cognition and Emotion, 13(6), 673-690) on a clinical sample. Thirty-one clinical participants and 32 matched controls received identical threat instructions regarding an upcoming speech-task following the computer task. Contrary to previous findings, clinical participants demonstrated a preference for attending toward faces than household objects, regardless of facial expression and more so than controls. Study 2 was conducted on a different clinical sample in an attempt to replicate the findings of Study 1 with clearer instructions regarding the upcoming speech task. Thirty-one clinical participants and 32 matched controls were administered the same speech threat and task instructions as those used by Mansell, W., Clark, D., Ehlers, A., and Chen, Y. P. (1999, Cognition and Emotion, 13(6), 673-690). Findings revealed the same pattern found in Study 1. Hence, in contrast to previous findings, both studies demonstrated that under conditions of social threat individuals diagnosed with generalised social phobia are more vigilant of pictorial faces generally than non-anxious individuals. The current findings lend support to cognitive models predicting that anxious individuals prefer to attend more to social cues of threat, in this case, faces. Explanations for these contradictory findings on attentional processing in social phobia are discussed with reference to the possible impact of perceived social threat and altered levels of state anxiety. Suggestion is also made for clearer research to reconcile these seemingly opposing results.