Bayesian analysis of multimethod ego‐depletion studies favours the null hypothesis

Abstract

Ego-depletion refers to the purported decrease in performance on a task requiring self-control after engaging in a previous task involving self-control, with self-control proposed to be a limited resource. Despite many published studies consistent with this hypothesis, recurrent null findings within our laboratory and indications of publication bias have called into question the validity of the depletion effect. This project used three depletion protocols involved three different depleting initial tasks followed by three different self-control tasks as dependent measures (total n = 840). For each method, effect sizes were not significantly different from zero When data were aggregated across the three different methods and examined meta-analytically, the pooled effect size was not significantly different from zero (for all priors evaluated, Hedges' g = 0.10 with 95% credibility interval of [-0.05, 0.24]) and Bayes factors reflected strong support for the null hypothesis (Bayes factor > 25 for all priors evaluated).

Is the Letter Cancellation Task a Suitable Index of Ego Depletion?

“…Overall, the lack of an effect of the letter cancellation task partially fits recent findings, for example, the preregistered replication study (Hagger et al, 2016), Bayesian analyses of the effect (Etherton et al, 2018) and behavioural results with a large sample size that found a very small effect (Wimmer, Stirk, & Hancock, 2017). However, current results can only be directly compared to the latter study where the same ego-depletion task and Stroop task as outcome task pairing was used.…”

“…Second, Bayesian analyses that compared performance in three different pairs of previously widely used ego-depletion and outcome tasks (restricted writing, letter cancellation, Stroop task followed by handgrip duration, anagrams, and mental arithmetic) revealed evidence in favour of the null hypothesis. Third, a preregistered replication study involving 23 laboratories that used the letter cancellation task as ego-depletion measure and the multi-source interference task as outcome measure of inhibitory control (Carter, et al, 2015;Etherton, et al, 2018;Hagger et al, 2016) raised further doubts about the strength of the effect. However, interestingly the replication study showed great variations in task performance in accuracy both between ego depleted and control participants and across different laboratories (15%-44% of participants performing < 80% correct) and a great range of effect sizes on response time differences between ego depleted and control participants (95% CI on Cohen's d = -0.06, 0.36).…”

Has the evidence for moral licensing been inflated by publication bias?

“…A well-known example of this is the ego depletion literature (Baumeister, Bratslavsky, Muraven, & Tice, 1998), where a meta-analysis over almost 200 studies reported an effect size of d = 0.62 (Hagger, Wood, Stiff, & Chatzisarantis, 2010). However, a re-analysis indicated that the effect is not as robust as previously thought (Carter & McCullough, 2014; see below) and two recent highly powered replication projects found no evidence for an effect (Hagger et al, 2016;Etherton et al, 2018).…”

Time to Set a New Research Agenda for Ego Depletion and Self-Control

“…Similarly, studies that employed other depleting tasks also yielded results from which no definitive conclusion could be drawn, with some studies reporting substantive, non-zero effects (Bayer & Osher, 2018;Dang, Liu, Liu, & Mao, 2017;Garrison, Finley, & Schmeichel, 2019) while others reporting effects that were no different from zero (Etherton, Osborne, Stephenson, Grace, Jones, & De Nadai, 2018;Lurquin et al, 2016;Singh & Göritz, 2018).…”