Protest is now central to politics in Western democracies, but it is known to citizens mainly through portrayals in the media. Yet the media cover only a small fraction of public protests, raising the possibility of selection bias. We study this problem by comparing police records of demonstrations in Washington, D.C. in 1982 and 1991 with media coverage of the events in The New York Times, The Washington Post, and on three national television networks. We model the consequences of demonstration form, context, and purpose on the likelihood of media coverage. The estimated size of a demonstration and its importance to the current media issue attention cycle are the strongest predictors of its coverage. Additional analyses support our claim that heightened media attention to an issue increases the likelihood that protests related to that issue will be covered. Comparing 1982 to 1991 suggests that television coverage of protests is increasingly subject to the impact of media issue attention cycles.