Beliefs about climate change divide the U.S. public along party lines more distinctly than hot social issues. Research finds that better-educated or informed respondents are more likely to align with their parties on climate change. This information–elite polarization resembles a process of biased assimilation first described in psychological experiments. In nonexperimental settings, college graduates could be prone to biased assimilation if they more effectively acquire information that supports their beliefs. Recent national and statewide survey data show response patterns consistent with biased assimilation (and biased guessing) contributing to the correlation observed between climate beliefs and knowledge. The survey knowledge questions involve key, uncontroversial observations such as whether the area of late-summer Arctic sea ice has declined, increased, or declined and then recovered to what it was 30 years ago. Correct answers are predicted by education, and some wrong answers (e.g., more ice) have predictors that suggest lack of knowledge. Other wrong answers (e.g., ice recovered) are predicted by political and belief factors instead. Response patterns suggest causality in both directions: science information affecting climate beliefs, but also beliefs affecting the assimilation of science information.