Composite models never (well, hardly ever) compromise: reply to Schooler and Tanaka (1991)