It is controversial whether the performance of visual search task in children with autism spectrum disorders (ASD) is better than that in typically developing children (TD) or not. One of the potential reasons is the difficulty levels of visual search task they employed in each study, i.e., adult version vs. child version. Through modulating the difficulty levels with the gap of director degrees between the target stimulus and the distractions during the visual search task, the study aim to 1) test whether the superior visual search performance in ASD could be modulated by the difficulties of visual search tasks; 2) and further study the underlying mechanisms of superior performance in ASD on visual search tasks using eye-movement data analysis if it is. The study adopted a mixed experimental design with two factors: group (ASD/ typically developing children, TD) and difficulty levels of visual search task (9 levels). Eighteen children with ASD (14M/4F) and 16 IQ and chronological age-matched TD (13M/3F) participated in the current study. The participants need to determine whether there was a distraction in the screen, which is just the same as the target stimulus as soon as possible. Among all distractions, there was a main stimulus which had the least differences of director degree compared with the target stimulus. The difficulty levels of visual search task was modulated by the gap of director degree between the target stimulus and the main stimulus (9 levels: 0°、4°、8°、12°、16°、20°、24°、28°、32°) (Figure 1). Participants’ behavioral responses were automatically recorded by the computer, and their eye-movement data were tracked by an eye tracker during the whole experiment. The behavioral responses indices, response accuracy and reaction time, during the task were analyzed by mixed design ANOVA. For eye-movement data, firstly several areas of interest (AOIs) in the stimulus picture were divided: Central AOI where the target stimulus was placed; Main AOI where the main stimulus was placed; Peripheral AOI where other distractions were placed. Then various eye-movement indices including the fixations, durations of fixation and saccade paths, intra and inter-AOI (s) were analyzed respectively between ASD and TD. The study found that only under the task difficulty level of 4° gap between target stimulus and main stimulus, the accuracy of visual search task in the ASD group was significantly higher than the TD group (t (1,32) = 2.59, p < .05, ηp2 = .17) (Fig. 3). Further eye-movement analysis suggested that less regression count on central AOI (t (1,26) = -3.77, p < .005, d = 1.28), less percentage of the fixation duration on left AOIs (t (1,26) = -5.49, p < .05, d = 2.03) and less fixation duration on main and peripheral AOIs (t (1,26) = -2.13, p < .05, d = .86) have been found in the ASD group during the visual search task (Table 1). Furthermore, ASD group tended to the right side lateralization fixation during the visual search task compared with TD group (χ2 (4) = 2.17, p < .005, d = .91). The results suggest that children with ASD only show superior performance in visual search tasks if the task difficulty is sufficiently high and that the superiority in ASD in the visual search task may come from enhanced perceptual sensitivity to interference stimulation during visual searching.