“…Most studies also had low performance bias and detection bias, as both participants and personnel were blind to the experimental condition ( Blinding of participants and personnel ) in most cases, as were outcome assessors ( Blinding of outcome assessment ). Seven studies (Baker et al., 2019; Bancroft et al., 2017; Bratter et al., 2016; Churchill et al., 2018; Kost‐Smith et al., 2012; Peters et al., 2017; Rapa, 2016) had a high risk of attrition bias due to incomplete outcome data either overall or in one condition relative to another ( Incomplete outcome data ). Interestingly, these high‐risk studies yielded a small and nonsignificant average affirmation effect ( g = .05, 95% CI [−.14, .24], p = .626), while the low‐risk and the unclear‐risk studies had medium and significant average affirmation effect ( low : g = .11, 95% CI [.03, .20], p = .012; unclear : g = .25, 95% CI [.15, .34], p < .001).…”