“…Finally, a model with a path from contact to prejudice ( ϭ Ϫ.49), 2 (42, N ϭ 769) ϭ 232.6, fit much better than one with a path from prejudice to contact ( ϭ Ϫ.42), 2 (42, N ϭ 769) ϭ 250.4; ⌬ 2 (1, N ϭ 769) ϭ 13.8, p Ͻ .01. Although these are crude tests, other investigators have reached the same conclusion using an array of different methods (e.g., Irish, 1952;Powers & Ellison, 1995;Wilson, 1996). …”