The SNP rs2958182 was reported to be significantly associated with schizophrenia (SCZ) in Han Chinese. This study examined this SNP's associations with cognitive functions in 580 SCZ patients and 498 controls. Cognitive functions were assessed using the Wechsler Adult Intelligence Scale-Revised (WAIS-RC), the Attention Network Task (ANT), the Stroop task, the dot pattern expectancy (DPX), task and the N-back working memory task. Results showed significant or marginally significant interaction effects between genotype and diagnosis status on IQ (P ¼ 0.011) and attention-related tasks (ie, the forward digit span of WAIS-RC, P ¼ 0.005; the ANT conflict effect; P ¼ 0.020, and its ratios over mean reaction time (RT), P ¼ 0.036; the Stroop conflict effect, P ¼ 0.032, and its ratios over mean RT, P ¼ 0.062; and the DPX task's error rate under the BX condition, Po0.001, and the error rate of BX minus the error rate of AY (BX À AY), P ¼ 0.002). There were no such interaction effects on the measures of working memory (all P-values 40.05). Further analysis of the significant genotype-by-diagnosis interactions showed that the risk (T) allele was associated with better performance on cognitive tasks in patients but with worse performance in controls. These results seem to indicate that the association between this SNP and selected cognitive functions may be of an inverted U-shaped pattern. Future research is needed to replicate these results and to explore the biochemical mechanisms behind this association.