This article briefly examines two recent works (ed. Kuklinski 2001, Marcus, Neuman and MacKuen 2000) which raise some of the major current problems and hypotheses relating to political psychology, cognitive sciences and public opinion research. In particular, attention is paid to the controversy opposing the "combinatorial" perspective of Paul Sniderman et al. to the "constructionist" approach of John Zaller on opinion formation and change. We argue that Zaller's accent on situational frames and Sniderman's emphasis on the underlying political dispositions are much more complementary than mutually exclusive. Therefore, it seems difficult to deny "in toto" the plausibility of the "ambivalence" hypothesis, especially with regard to specific low-involvement segments of citizens and to various new non-obtrusive, not yet crystallized political issues.