Visual disengagement has been hypothesized as an endophenotype for autism. In this study we used twin modelling to assess the role of genetics in basic measures of visual disengagement, and tested their putative association to autistic traits in the general population. We used the Gap Overlap task in a sample of 492 twins. Results showed that most of the covariance among eye movement latencies across conditions was shared and primarily genetic. Further, there were unique genetic contributions to the Gap condition, but not to the Overlap condition--i.e. the one theorized to capture visual disengagement. We found no phenotypic association between autistic traits and disengagement, thus not supporting the hypothesis of visual disengagement as an endophenotype for autistic traits.