Background: Although accumulating research demonstrates the association between attentional bias and social anxiety, the bias for positive stimuli has so far not been adequately studied. Aims: The aim is to investigate the time-course of attentional bias for positive social words in participants with high and low social anxiety. Method: In a modified dot-probe task, word-pairs of neutral and positive social words were randomly presented for 100, 500, and 1250 milliseconds in a nonclinical sample of students to test their attentional bias. Results: Non-significant interaction of Group × Exposure Duration was found. However, there was a significant main effect of group, with significantly different response latencies between the high social anxiety (HSA) and low social anxiety (LSA) groups in the 100 ms condition, without for 500 or 1250 ms. With respect to attentional bias, the LSA group showed enhanced preferential attention for positive social words to which the HSA group showed avoidance in the 100 ms condition. In the 500 ms condition, preferential attention to positive social words was at trend in the LSA group, relative to the HSA group. Neither group showed attentional bias in the 1250 ms condition. Conclusions: These findings extend recent research about the attention training program and add to the empirical literature suggesting that the initial avoidance of positive stimuli may contribute to maintaining social anxiety.