Constituency-level analyses of electoral turnout commonly uncover a correlation between the marginality of a seat and the level of electoral participation in the seat: the closer the local contest, the greater the rate of participation in the election. However, repeated efforts to assess the impact of constituency marginality on the propensity of individual electors to participate have met with failure. The 2001 British General Election was no exception to either result. This presence of an ecological aggregate-level relationship which is not replicated at the level of individual voters is paradoxical. However, the paradox can be resolved when two analytical steps are combined. First, nonvoters are classified into two groups according to their reasons for abstention: those who abstain on purpose ('voluntary abstainers'), and those who fail to vote for reasons largely beyond their control ('involuntary abstainers'). Second, attention is paid not only to actual marginality but also to perceived marginality. Individuals who think their constituency is competitive are less likely to abstain deliberately than individuals who think their constituency is safe.