A preliminary meta-analysis using the p-curve method (Simonsohn, Nelson, & Simmons, 2014) was performed on a subset of the learned categorical perception literature to explore the robustness of the phenomenon. Only studies using novel visual categories and behavioral measures were included. The results strongly suggest that the phenomenon is robust but that the studies are somewhat underpowered. We argue that this is problematic because it renders both statistically significant and nonsignificant results very difficult to interpret, which impedes progress in understanding the learned CP phenomenon, for example, why expansion vs. compression is observed, or boundary vs. dimensional effects. Fortunately, there is a clear solution: conduct studies with greater statistical power.