The effects of two training procedures--(a) modeling and praise and (b) self-instruction, modeling, and praise--on complex gross-motor chain acquisition for preschool dance class students were evaluated. Six girls participated in the study. A multiple baseline design across six gross-motor chains with a secondary group comparison for treatment order effects was used. Both procedures were effective in teaching the gross-motor chains. Nevertheless, for 4 of the 6 participants, the self-instruction procedure produced a faster acquisition rate in at least two of the three comparable pairs of gross-motor chains. Furthermore, very early in gross-motor chain acquisition, for five of the six gross-motor chains, the self-instruction condition was associated with faster response acquisition.