The Assessment of Basic Learning Abilities (ABLA) is an empirically validated clinical tool for assessing the learning ability of persons with intellectual disabilities and children with autism. An ABLA tester uses standardized prompting and reinforcement procedures to attempt to teach, individually, each of six tasks, called levels, to a testee, until a pass or a fail criterion is met on each level. The six levels are ordered in difficulty from Level 1 to Level 6. We examined the predictive validity of the ABLA performance of nine children with autism who passed ABLA Levels 2 or 3, and failed higher levels. We attempted to teach 20 criterion tasks to each child, using standardized prompting and reinforcement procedures, until each child met either the pass or the fail criterion of the ABLA on each task. We predicted that each child would pass the criterion tasks that corresponded to his/her previously passed ABLA levels, and would fail the criterion tasks that corresponded to his/her previously failed ABLA levels. A parent of each child was also asked to predict the child's pass-fail learning performance on the 20 criterion tasks. Ninety-two percent of the predictions based on the children's ABLA performance were confirmed, and the ABLA was significantly more accurate than the parents for predicting the children's performance on the criterion tasks.