Reversing arrows in the classic tri-variate X-M-Y mediation models as a test to check whether one mediation model is superior to another is inadmissible. Presenting evidence that one tri-variate mediation model yields a significant indirect effect, whereas one with some reversed arrows does not, is not proof or even evidence that one model should be preferred. In fact, the significance of the indirect or any other effect can never be used to infer whether one model should be preferred over another, if the models are in the same so-called equivalence class. The practice of running several mediation models with reversed arrows to decide which model to prefer should be abandoned. The only way to choose among equivalent models is through assumptions that are either fulfilled by design features or invoked based on theory. Similar arguments about reversing arrows in mediation models have been made before, but this current work is the first to derive this result analytically for the complete (Markovian) equivalence class of the tri-variate mediation model.A mediator M in the tri-variate X-M-Y mediation model cannot be statistically distinguished from a confounder (Fiedler, Schott, & Meiser, 2011;MacKinnon, Krull, & Lockwood, 2000). In fact, as we show, it cannot be distinguished from a common cause, common effect, or a mediator in the reverse direction using statistics alone. However, there is still a persistent belief that it can be helpful to test alternative mediation models by switching arrows and comparing size and significance of the indirect effects. Without wanting to single out specific publications or authors, this practice can be seen in published work; for example, Fredrickson, Tugade, Waugh, and Larkin (2003) used this technique to argue that one mediation model (the one with the significant effect) was more plausible than two alternative models (the ones without significant effects). Anecdotally, the author can also attest that it is not uncommon that reviewers (or editors) request that in addition to a presented mediation model, some arrows should be reversed, and the model tested again, presumably to gain insight into which model is preferred.The (erroneous) thought behind this practice is that if the mediated effect of one model is significant, but the other model has no significant mediated effect, then the mediation model with the significant indirect effect is more plausible. However, none of the six models that one could form by reversing the three directed arrows in the tri-variate mediation model can be preferred on the basis of the significance of any of the resulting effects. 1