Forty percent of U.S. citizens believe in creationism, and thirteen percent of people globally believe vaccines are more dangerous than the diseases they prevent. How do these and other misconceptions arise, and what can be done to correct them? In my talk, I will describe two lines of research examining the formation and correction of scientific misconceptions. First, I show that information-processing constraints can bias the output of explanatory reasoning in ways we can detect in how adults and children make sense of unfamiliar scientific observations. Strikingly, this bias can account for a systematic pattern of theory change in the history of science. Second, I illustrate two strategies for countering scientific misconceptions: by leading people to revise the auxiliary beliefs that prompt and support their misconceptions, and by extracting features of effective arguments mined from online communities. Together, this work reveals a mechanism underlying certain scientific misconceptions and new methods for correcting these beliefs.