We propose a theory of territorial captivity to explain the level of voter turnout in national elections. We start by showing that the consequences of voting in an election are limited to a clearly defined territory. For this reason, the expected return on the election results will be higher for electors who have high exit costs. According to the theory of rational voting, the expected return on the election results influences the level of turnout. If this is so, then we can argue that the more "territorially captive" voters are, the more likely they are to vote. We continue by describing the institutional, geographical and property-related nature of captivity. By testing our hypothesis in the context of the French parliamentary elections of 1997, we then demonstrate empirically that the constituencies in which individuals are most captive are also those with the highest turnout.