The aim of this study was to introduce a modified version of the covariation detection task to the meta-reasoning framework. This task has been used to assess scientific reasoning through the evaluation of fictitious experiment outcomes and hypothesis testing. The traditional covariation detection task was modified to include only the magnitude versus ratio-bias. The participants' task was to evaluate the effectiveness of an experimental manipulation in a series of fictitious experiments. Experiment 1 (N = 61) consisted of twenty covariation detection tasks. In half of the tasks, normative and heuristic responses were congruent, and for the other half they were incongruent. Experiment 2 (N = 48) had the same experimental design, however, the fictitious data was modified to increase the relative strength of the normative response. After each trial participants provided a judgment of confidence. Results confirmed that the main manipulation of congruence was successful. Participants were more accurate, faster and more confident in the congruent condition. The manipulation from Experiment 2 had a larger impact on response times than on confidence judgments and accuracy. Correct responses were faster in Experiment 2 when compared to Experiment 1, with higher confidence for correct congruent responses. Analyses by response type revealed large individual differences in the relative strength of the processes which generate normative and biased responses. Participants were faster and more confident when rationalizing in favour of their dominant response while they were slower and less confident when decoupling from that dominant response. The covariation detection task provides new valuable insight into metareasoning processes.
