Moreover, a borderline significant interaction between intention and block ( F 2.2,61.7 = 2.96, p = .056, η p 2 = .09) was driven by significantly better performance with implementation intentions compared with goal intentions on the first block-set ( F 1,29 = 6.27, p = .02, η p 2 = .18)—an effect that was absent in the later training block-sets (all p > .16, F < 2.0).
Paired t-tests revealed evidence for both: the group that used if-then planning showed a trend towards better performance on Go trials when intentions were introduced ( t 28 = –1.81, p = .08, d =–.34, 95% CI = [–.71, .04]), while goal-intentions had the opposite effect ( t 30 = 2.18, p = .04, d = .39, 95% CI = [.02, .75]).
Importantly, as in Study 1, the expected interaction between congruence and intention failed to reach significance ( F 1,29 = 2.26, p = .14, η p 2 = .07), although the Bayesian evidence against an interaction was inconclusive (BF excl = 1.04).