Previous research has suggested that the response-independent delivery of preferred stimuli can increase compliance to low-probability instructions, although these results have been mixed. Two variables that might affect compliance during this procedure are the magnitude and duration of access to the stimuli. In the current study, we evaluated stimulus magnitude and duration of access during pretrial delivery of preferred items on compliance among children with autism. In Experiment 1, we compared high and low magnitude edible stimuli. In Experiment 2, we compared long and short durations of access to leisure stimuli. Results show that high magnitudes and long durations of access to preferred stimuli delivered once immediately before a low-probability instruction increased compliance more than low magnitudes and short durations of access to preferred stimuli. We discuss the implications and possible mechanisms responsible for these results.