Abstract: Why do false rumors spread? Why do otherwise sensible people believe them? Why are they sometimes impervious to correction? There are several answers. (a) Some false rumors gain traction because of their fit with prior convictions within particular groups and cultures. People are strongly motivated to accept certain beliefs, however groundless; they also have good reasons to accept some of those beliefs. Diverse groups will have diverse thresholds for accepting false rumors. It follows that particular rumors can have a tenacious hold on some groups and cultures while dying a rapid death in others; multiple equilibria are likely. (b) Informational cascades are often responsible for belief in false rumors. Such rumors typically spread as a result of such cascades; people believe them because they lack the information that would lead them to reject the signals given by the apparently shared beliefs of numerous others. The important point here is that with respect to many rumors, private signals are essentially nonexistent. (c) Reputational cascades help propagate false rumors. Sometimes people do not correct such rumors, and even endorse them, so as to curry favor or to avoid public opprobrium. Because of the role of early movers, multiple equilibria are (again) likely, as some groups come to believe rumors that other groups deem preposterous. (d) Group polarization accounts for the intensity with which people accept false rumors. Like-minded people, engaged in deliberation with one another, increase one another's confidence in rumors. Here too we see why false rumors are widely believed within some groups but widely rejected in others. As a result of group polarization, such rumors often become entrenched. (e) Biased assimilation can make false rumors exceedingly hard to correct. Because people with strong antecedent commitments process balanced information in a biased way, such information can strengthen people's commitment to false perceptions. That commitment can also be strengthened by corrections, which therefore turn out to be self-defeating. These points have significant implications for freedom of speech and the marketplace of ideas, especially in the age of the Internet; they demonstrate that the exchange of information may not produce convergence on truth and that damaging false reports will often be widely credited. A chilling effect on false rumors can be highly desirable; the goal should be to produce optimal chill, rather than no chill at all.