Costs and limits are assumed to be the major constraints on the evolution of phenotypic plasticity. However, despite their expected importance, they have been surprisingly hard to find in natural populations. It has therefore been argued that natural selection might have removed high-cost genotypes in all populations. However, if costs of plasticity are linked to the degree of plasticity expressed, then high costs of plasticity would only be present in populations where increased plasticity is under selection. We tested this hypothesis by investigating costs and limits of adaptive phenotypic plasticity in development time in a common garden study of island populations of the common frog Rana temporaria, which have varying levels of development time and phenotypic plasticity. Costs of plasticity were only found in populations with high-plastic genotypes, whereas the populations with the most canalized genotypes instead had a cost of canalization. Moreover, individuals displaying the most extreme phenotypes also were the most plastic ones, which mean we found no limits of plasticity. This suggests that costs of plasticity increase with increased level of plasticity in the populations, and therefore costs of plasticity might be more commonly found in high-plastic populations.