Two experiments were conducted to evaluate symbolic-deficit and memory-deficit hypotheses to account for the cognitive problems seen in children with autism. Experiment 1 tested imitation, in immediate and deferred conditions, of familiar actions with different sets of objects representing the developmental progression from functional to symbolic play. The results showed that the autism group and both their receptive language and nonverbal IQ-matched controls imitated familiar actions with realistic objects (evidence for functional play) and placeholder objects (evidence for symbolic play) after delays ranging from 24 hr to 3 weeks. Experiment 2 tested familiar three-step event sequences in which a placeholder object was substituted for the second step in half the events. The results showed that the autism group remembered as many of the actions with the placeholder objects as their language-matched controls and as many correctly ordered sequences, a finding that supports a symbolic-delay (rather than deficit) hypothesis. These results were obtained in highly structured test situations and sharply contrast with the impairments seen in children with autism who are observed in naturalistic settings. Two interpretations of these findings are offered. First, structured test settings minimize distractions that typically occur in naturalistic settings that may interfere or disrupt symbolic play in children with autism. Second, the results are consistent with an executive function deficit in that the autistic group demonstrated more knowledge in the test settings than they demonstrate spontaneously in naturalistic ones.