Research on motivated reasoning suggests that partisans who discount unwelcome evidence also learn less from it. Yet persuasion experiments find opposing groups updating in parallel, and some studies report greater movement toward counter-attitudinal evidence. We reconcile these findings through a meta-reanalysis of 41 experiments (30,731 respondents) that measure both evaluations of political information and subsequent beliefs. Contrary to stronger accounts of motivated reasoning, we find that partisan bias is strong in how people judge political information but much smaller in how far their beliefs move in response (0.54 vs. 0.13 SD in designs least prone to measurement bias). We also find no evidence that partisans learn more from counter-attitudinal information. Instead, we show that this apparent pattern largely reflects regression to the mean arising from how belief change is measured. The results suggest that partisan divisions over evidence may pose less of a barrier to learning than is often assumed.