It's not been a good month for the theory of ego-depletion - the idea that self-control is a limited resource that can be depleted by overuse. Two weeks ago, researchers reported evidence of bias in the published literature examining the question of whether glucose can reverse ego-depletion. Now, the very existence of the ego-depletion phenomenon has been questioned by an international collaboration of psychologists who conducted a preregistered replication attempt (RRR). The results have just been published in Perspectives on Psychological Science (although they've been circulating for a while.) 23 labs participated in the replication effort, with a total final sample size of 2,141 participants. The project was co-ordinated on the Open Science Framework (OSF) and the paper was written by Martin Hagger and Nikos Chatzisarantis. Methodologically, the replicators did not attempt to reproduce the original version of the ego-depletion task, saying that "the tasks used in the original experiments were deemed too elaborate or complex to be appropriate for a multilab replication." Instead they focussed on the procedures used by a recent paper, Sripada et al. (2014), which were fully computer-administered, and thus more portable. Importantly, Hagger and Chatzisarantis say that "The decision to use these tasks was based on the recommendation of Roy Baumeister"; Baumeister being the man who introduced the concept of ego-depletion back in 1998... but see below. When asked to predict the outcome of the experiment, only one lab predicted a null result. The other 22 predicted seeing at least a small effect, perhaps because there have been dozens of published studies reporting positive evidence of ego-depletion. Contrary to expectations, the replication gave a null result: there was no statistically significant ego-depletion effect, combined across all labs (Cohen's d = 0.04, 95% CI [−0.07, 0.15]). In two out of the 23 labs, a significant ego depletion effect was seen, but in a third lab, a significant reverse effect occurred. Restricting the analysis to English-speaking participants produced a slightly higher effect size, but it was still non-significant (d = 0.14, 95% CI [−0.02, 0.30]).