On the Tyranny of the Normal Distribution

The normal curve has long occupied a privileged place across scientific disciplines, and psychology was among its early adopters, in part because of its success in the analysis of the physical world. Yet historical influence and analytic convenience do not, by themselves, justify its continued use as a default model for psychological phenomena. The present article argues that the adoption of the normal, or Gaussian, distribution in psychology reflects sociocultural and philosophical forces more than purely scientific or mathematical ones. More specifically, the article advances three claims. First, the prevalence of positivist thought in 19th-century Europe was a primary reason this distribution found its way into psychology. Second, in principle, psychological data are difficult to model as normally distributed because the kinds of processes that generate human judgment, behavior, and measurement outcomes often do not satisfy the conditions under which normality naturally arises. Third, statistical methods in psychology implicitly assume that most data are, at the very least, elliptically contoured; otherwise, the covariance matrix is not an effective summary of multivariate associations. The article closes by illustrating, through selected examples and simulations, how these hidden assumptions can distort common statistical practices and their resulting inferences.