Mediation models involve two regression equations, one each for the dependent and mediating variables, Y and M. Tests of theoretical hypotheses in strategic management focus on “betas,” that is, the estimated regression coefficients in the Y and M equations. A neglected point is that strategic management theory sometimes suggests a “hidden hypothesis” concerning the sign of the error correlation between the Y and M regressions. Systems-based estimation methods can be used to test “hidden hypotheses.” We show that sometimes “hidden hypotheses” offer the only hope of empirically distinguishing between competing theories.