THE SELECTIVELY POLITICAL CITIZEN - MODELING ATTITUDES, NONATTITUDES,AND CHANGE IN 1950S PUBLIC-OPINION

Since Converse's pioneering work, social scientists have had to grappl e with a perplexing phenomenon: the apparent willingness of many citiz ens to express an opinion to survey researchers when they in fact have no corresponding attitude. New approaches to measuring latent structu res make it possible to address some of the controversies surrounding nonattitudes and to probe the empirical adequacy of Converse's account of mass political attitudes. This article extends the work of Duncan, Stenbeck, and Brody by analyzing Converse's Black-White model and alt ernative latent trail models when the American Panel Study data have n ot been recoded as dichotomies or trichotomies. The fit that Converse reported for his Black-White model is found to be the product of an in valid recoding scheme. Latent trait models provide a superior fit to t he data. Although latent trait models do not allow researchers to clas sify respondents into logical classes of attitude (or nonattitude) hol ders, modeling results indirectly suggest that nonattitude holders wer e not a majority on any issue. Respondents' tendency to choose ideolog ically consistent categories is found to be limited. The relevance of these findings to the idea of a ''selectively political citizen '' is discussed in conclusion.