To explore the characteristics of visual search attention in children with Williams syndrome (WS). Twenty-two children with WS, 13 male and 9 female, aged 12 (6-16), 25 chronological age (CA) and mental age (MA)-matched children with Down's syndrome (DS), 45 biological age-matched children, and 43 mental age-matched children, aged 4 +/- 2 (2-8) underwent Peabody picture vocabulary test (PPVT). Visual search attention including selective, switch, and sustained attention, was tested using a set of computerized visual search tasks via a touch-screen. In the dual target tasks, participants were required to alternate between two different targets. Sustained attention was investigated with vigilance task test. (1) In the single-target searching task with no target similar distractor, the accurate response rate of the WS patients was 71% +/- 25%, significantly lower than that of the CA-matched children (87% +/- 14%, P = 0. 001). The searching time of the WS children was 5 s, significantly longer than that of the CA-matched children (3 s, P = 0.000). The distance for touch of the WS children was 25 relative units, significantly longer than that of the CA-matched children (18 relative units, P = 0.000). The mean reaction time of the WS children was 5 s, significantly longer than that of the children with Down's syndrome (3 s, P = 0.022). The shape error rate of the WS children was 15%, significantly higher than that of the CA-matched children (0, P = 0.000). When non-targets similar to the targets were added, the accurate response rate of the WS group was 39% +/- 20%, significantly lower than those of the CA-matched and MA-matched children (77% +/- 23% and 66% +/- 23% respectively, both P = 0.000); the mean searching time of the WS children was 13 s, significantly longer than those of the CA-matched and MA-matched children (3 s and 5 s respectively, both P = 0. 000); and the distance per touch of the WS children was 41 relative units, significantly longer than those of the CA-matched and MA-matched children (20 and 27 relative units, P = 0.000 and P = 0.004). (2) The results of the dual target tasks showed that the accurate response rate of the WS children was 52% +/- 28%, significantly lower than that of the CA-matched children (78% +/- 22%, P = 0. 000), the mean searching time of the WS children was 11 s, significantly longer than that of the CA-matched children (4 s, P = 0.000); and the distance per touch of the WS children was 54 +/- 27 relative units, significantly longer than that of the CA-matched children (31 +/- 13 relative units, P = 0.000). However, there were not significant differences in the accurate response rate, mean searching time, and distance per touch between the WS and DS children. The switching error rate of the WS children was 13%, significantly higher than those of the CA and MA-matched children (0 and 4% respectively, P = 0.000 and P = 0.004). (3) The vigilance task test showed that the accurate response rate of the WS children was 52% +/- 25%, significantly lower than that of the CA-matched children (80% +/- 21%, P = 0.000); the mean searching time of the WS children was 4 s +/- 1 s, significantly longer than that of the CA-matched children (3 s +/- 1 s, P = 0.000); and the error hit number of the WS children was 8, significantly more than that of the CA-matched children (3, P = 0.000). However, there were not significant differences in the accurate response rate, mean searching time, and error hit number between the WS and DS children. Conclusion Distinct visual search deficits exist in WS children.