Relative and absolute attentional biases in response times and error rates to social and health threat words were investigated in a sample of female undergraduates with high and low social anxiety (n = 63; mean age = 20.22) and high and low depression (n = 63; mean age = 20.30). A dot-probe paradigm measured attention to (1) social threat versus neutral words, (2) social threat versus health threat words, and (3) health threat versus neutral words. Participants with low social anxiety displayed an absolute bias in response times but not error rates away from social threat words. In contrast, participants with high social anxiety displayed an absolute bias in error rates but not response times toward social threat words. Findings suggest that attention toward social threat may not be unique to social anxiety, and that individuals with high social anxiety may lack a protective bias away from social threat.