Learning at any rate: action–effect learning for stimulus-based actions

Prior implementations of the current experimental design used two independent criteria for participant selection (Elsner & Hommel, 2001, Exp. 3B; Hoffmann et al., 2009, Exp. 2). As first selection criterion, more than 10% erroneous responses in nogo-trials were considered as an indicator of response selection prior to go-signal onset. This behaviour would lead to reduced consistency effects as a prepared response is unlikely to be affected by tones accompanying the go-signal. This tendency was not present in the current data-set (r = 0.12 between the percentage of consistent choices and the number of nogo-errors). As second selection criterion, relative percentages of left and right reactions were compared and participants were excluded from data analysis when the distribution was not considered as even. This procedure is statistically warranted as an uneven distribution of response choices inevitably reduces the chance of finding a consistency effect that might be present in the data. This trend was also present in the current data, indicated by a correlation of r = −0.15 between the relative percentage of consistent key presses and the percentage of the preferred response. However, to obtain a conservative estimate of the consistency effect, we did not exclude any participant in the present study.