The political distinctiveness of major American religious groupings is well-documented, but the role of church attendance in maintaining these cleavages has been unclear. Analysis of white respondents in five national surveys covering 16 years reveals that church attendance is significantly related to party identification and presidential vote in all but one instance. Application of Goodman's log-linear analysis of contingency tables shows that higher rates of church attendance are related to non-Democratic preferences among Protestants and Democratic preferences among Catholics. In several years, the relationship depends as well upon region of residence. The effects, particularly on the vote, seem to be changing systematically over time as the South becomes politically similar to the non-South.