To test preschoolers' ability to flexibly switch between abstract rules differing in difficulty, ninety-three 3-, 4-, and 5-year-olds were instructed to switch from an (easier) shape-sorting to a (harder) function-sorting rule, or vice versa. Children learned one rule, sorted four test sets, then learned the other rule, and sorted four more sets. In a control condition, seventy-two 3-5-year-old children learned one rule and were re-trained on that rule before the second test block. Half of each group received metacognitive reminders to ''think about'' the current rule before each test trial. The shape rule was easier: many 3-year-olds failed to follow the function rule, confirming findings of Deak et al. (2002). Switching rules did not reduce overall rule-following. However, reminders facilitated rule-following when rules were switched, but not when a rule was repeated (i.e., control condition). Reminders actually reduced rule-following by control children who got the easier (shape) rule. The results show (1) 4-year-olds readily switch between abstract rules, even if the second rule requires ignoring obvious, conflicting perceptual information (i.e., shape); (2) some rule-switching tasks do not impose performance costs on children, and (3) children's rule-following consistency and flexibility depend on the nature of available social support.