Few studies have examined the effects of the high-probability instructional sequence in the treatment of food selectivity, and results of these studies have been mixed (e.g., Dawson et al., 2003; Patel et al., 2007). The present study extended previous research on the high-probability instructional sequence by combining this procedure with low-probability demand fading with 2 boys with autism (9 and 10 years old) who had a history of food selectivity and engaged in active food refusal behaviors when presented with novel foods. Response requirements were faded gradually from responses the child would tolerate (e.g., touching the food) to the final requirement of chewing and swallowing the food. The antecedent-based intervention was implemented in the absence of escape extinction and was effective in increasing food consumption for both participants. Possible mechanisms responsible for the effectiveness of the intervention are discussed along with directions for future research. (Contains 2 figures.)