Evaluating the degree of “unhappy randomization” in psychology: Are psychology experiments appropriately arranged to counter internal validity threats?
This study sought to evaluate whether the randomized controlled trials currently conducted in psychology are suitable for countering internal validity threats in terms of precluding alternative explanations. Experimental research primarily achieves the goal of drawing causal inferences through random assignment. As such, one of the threats to the internal validity of experiments is when random assignment fails to balance nuisance variables, thereby causing the treatment effect to be confounded by nuisance variables. The authors argue that the successful random assignment is largely decided by the sample size and the confounding effects that the researchers are willing to tolerate. In this study, one thousand and eighteen randomized controlled trials published in three leading psychology journals were investigated in order to estimate the probability that nuisance variables were balanced in these experiments. The results indicated that, even if researchers are willing to tolerate a medium effect size of confounding effects, there is still a high probability that over one-fifth (22.99%) of the psychological experiments investigated could be confounded by nuisance variables, and the situation would be even worse if a stricter criterion was set. Some suggestions are proposed for enhancing the internal validity of future experiments.