Prompt fading methods are widely used to teach conditional discriminations to learners with developmental disabilities. To date, very little research has investigated the relative effectiveness of the various prompt fading methods described in the literature. Furthermore, learner preference for these instructional tactics has rarely been evaluated. The present study compared the relative effectiveness of three prompt fading methods (most-to-least with an embedded constant time delay, least-to-most, and progressive time-delay) to a control condition. Three boys with autism participated in the study. Two met mastery during the prompt fading assessment (one in progressive time-delay, and one in least-to-most prompting). In a subsequent concurrent chains preference assessment, both participants indicated a preference for the fading method that had been most effective. Considerations for practitioners when selecting prompt fading methods as well as procedural parameters warranting further investigation are discussed.