Only one of twelve church-member studies supports the curvilinear religion-prejudice model

Evidence

If you look at those appropriate twelve studies, you find only one that produced a significant up-and-down curvilinear relationship (Hoge and Car-

roll, 1973). In contrast Rosenblum (1958) found Lows and Moderates were both significantly more prejudiced than High attenders. The remaining ten studies involve “two-point” comparisons that can say nothing about curvilinear trends. Five of these were significant in the predicted direction, four were nonsignificant (or not known to be significant), and one was significant in the “wrong” direction.

— Bob Altemeyer, The Authoritarian Specter, 1986