This study investigated the differential effects of two attention bias modification (ABM) with different stimulus durations. Seventy-two undergraduates with subclinical social anxiety were randomly assigned to one of four conditions: an ABM condition with either a 100-ms or a 500-ms stimulus duration (ABM-100/ ABM-500) or an attention placebo (AP) condition with either a 100-ms or a 500-ms stimulus duration (AP-100/ AP-500). Participants completed the pre-assessments, eight attentional training sessions, and post-assessments. A modified Posner paradigm was used to assess changes in attentional processing. After completion of attentional training, the ABM-100 group significantly speeded up their responses to 100-ms invalid trials, regardless of the word type. The ABM-100 group also exhibited significant reduced latencies to 500-ms invalid social threat trials and a marginally significant reduced latencies to 500-ms invalid neutral trials. The ABM-500 group showed significant reduced latencies to 500-ms invalid social threat trials. Both ABMs significantly reduced participants’ fear of negative evaluations and interactional anxiousness relative to their comparative AP. The effects on social anxiety did not differ between the two ABMs. This study suggests that although both ABMs using short and long stimulus durations reduce some aspects of social anxiety, they influence participants’ attentional disengagement in different ways.