Limited data are available regarding how response prompting procedures should be used in early childhood settings. The purpose of this study was to compare the efficiency of progressive time delay instruction presented via two trial arrangements: massed and embedded. During massed trial sessions, a short instructional session was conducted, followed by game play. During embedded trial sessions, trials were conducted during game play. Across 12 participants with and without disabilities, efficiency and preference data were mixed. Preference data suggested most participants preferred the instruction type that resulted in most efficient learning. Social validity data suggest inservice and preservice practitioners may be equally likely to use the procedures, but some respondents indicated they might be less willing to use embedded trial arrangements due to increased effort in materials preparation. Implications include the possibility that faster learners may learn better in embedded contexts and that children may prefer instructional contexts that are most efficient for them. (author abstract)