Are effect sizes reported in highly cited emotion research overestimated relative to larger studies and meta-analyses addressing the same questions?
We assessed whether the most highly-cited studies in emotion research report larger effect sizes compared with meta-analyses and the largest studies on the same questions. We screened all reports with at least 1000 citations and identified matching meta-analyses for 40 highly-cited observational and 25 highly-cited experimental studies. Observational studies had on average 1.42-fold (95% CI 1.09 to 1.87) larger effects than meta-analyses and 1.99-fold (95% CI 1.33 to 2.99) larger effects than largest studies on the same questions. Experimental studies had fold-increases of 1.29 (95% CI 1.01 to 1.63) versus meta-analyses and 2.02 (95% CI 1.60 to 2.57) versus largest studies. There was substantial between-topic heterogeneity, more prominently for observational studies. Highly-cited studies were uncommonly (12/65 topics, 18%) the largest ones, but they were frequently (31/65 topics, 48%) the earliest published on the topic. Highly-cited studies may offer, on average, exaggerated estimates of effects in both observational and experimental designs.