Data from a national survey were used to test a model representing the dynamics of newspaper reading behaviors. It was hypothesized that three types of constraints (structural constraints, transitional constraints, and self-constraints) would be present to various degrees in four types of newspaper readers: regular readers, regular nonreaders, past readers who had quit reading (droppers), and past nonreaders who had begun reading (adders). These four types of readers were present in a sample of 1,201 people surveyed during both the 1974 and the 1976 election campaigns. Operational indicators included education and income for structural constraints; age and residential, marital, and occupational changes for transitional constraints; and campaign interest, campaign activity, attention to public affairs, and political activity outside the campaign context for self-constraints. Overall, the constraints model held up well under testing. As predicted, structural constraints accounted for stable nonreading. Transitional constraints, particularly those associated with youth and mobility, predicted changes in readership status. Directional changes, as represented by the adders and droppers, were associated mainly with indicators of self-constraints. The data were also consistent with expectations that self-constraints would be a function of structural factors, the strongest indicators of readership status. (RL)