In case the articles included multiple studies, multiple samples, or multiple self-talk treatment conditions, the corresponding effect sizes were calculated. Ifmultiple dependent measures were obtained through a single manipulation (e.g. multiple indicators of task performance), the calculated effect sizes were averaged to produce one overall effect size for this particular treatment (De Coster, 2004). Outlying effects were identified, and these values were set at the next closest value in the data, following Hagger, Wood, Stiff, and Chatzisarantis’ (2010) procedures. The fail-safe statistic (Hedges & Olkin, 1985) was calculated to account for possible publication bias.