Three children between the ages of 3 and 3 1/2 were asked to choose a colored object from an array of 5 colors in a baseline condition. After color preferences were established, stickers, small toys and praise were made contingent on choosing the least preferred color. After the first experimental condition resulted in consistent choosing of the least preferred color, a second experimental condition was implemented. At the beginning of each session a contingency-specifying stimulus (CSS) was presented, each CSS specifying a different color to be selected. Both contingency-shaping and CSS presentation resulted in stimulus control over responding. However, CSS presentation resulted in immediate redistributions of behavioral units across CSS sessions.