"All this to say that when studying human gender traits, gender identity, or sexual orientation, essential conditions for inferring cause and effect—the manipulation of one factor (sex) and the control of others (social as well as biological)—cannot be met. It means that 'all data on sex differences, no matter what research method is used, are correlational data,' and as every introductory social science student learns, you cannot draw firm conclusions about causality from merely correlational data."
Mary Stewart van Leeuwen, "Neurohormonal wars: old questions and dubious debates in the psychology of gender," Books and culture (September/October 2012): 12, small caps mine. This cuts, of course, both ways. But still. . . .