To Misspecify Is Common, to Probe Misspecification Scientific: Common “Confounds” in Pornography Research May Actually Be Predictors

Abstract

AbstractThe ills of modeling variables substantively involved in a causal process as “controls” have been discussed extensively by social scientists who do not study media. Until recently, Slater was one of the few communication scientists to suggest that media effects scholars engage in overcontrol. Bushman and Anderson have now echoed this concern in the context of a broader treatise on research trends in the media violence literature. The present study responded to Wright’s recent discussion of control variable usage in the pornography literature. Specifically, using a national probability sample of approximately 1,900 U.S. adults, the present study assessed whether multiple demographic variables routinely modeled as controls in the pornography effects literature may be better conceptualized as initiating predictors. Results were inconsistent with the confounding approach but consistent with the hypothesis that individual differences predict cognitive response states that increase or decrease the likelihood of media effects.