We used an orthogonalized Go-NoGo task that contrasts a propensity to act, rather than not to do so, in context involving opportunity (‘win’) versus threat (‘avoid loss’). Participants were presented with four different abstract stimuli each of which had a constant, but unknown, association with a correct policy. The correct policy was either to emit (‘Go’) or to withhold (‘NoGo’) an action, here involving a button press. If the correct decision was made, the better of two outcomes was realised with probability 0.8. This better outcome was null (as opposed to a loss) for two stimuli and positive (as opposed to null) for the other two. The task closely followed a previously published paradigm, with some slight simplifications, unrelated to the core biases assessed by the task. These simplifications helped deliver it to a community sample, on a large scale and in the context of a multi-task battery. First, implementing the decision ‘Go’ was simpler, i.e. not dependent on any target features, unlike the original task in which the ‘Go’ action could be either ‘left’ or ‘right’ depending on the location of a target. This allowed trials to be shorter. However, time pressure from the remaining task battery (to be reported separately), meant that subjects performed a more restricted sample of 144 trials. Second, task clarity was improved by informing participants before performing the task that the outcome probabilities were 0.8 and 0.2. Third, motivation was made explicit by telling participants that they were playing for real money, that random performance would attract zero extra fee and excellent performance could be worth about five pounds sterling additional earnings. These changes were supported by piloting the whole battery in which the task was embedded, as we describe next.

We took precautions to ensure that the fact that the task was delivered as part of a battery did not affect the power for testing the hypotheses in question. The battery of which this task was part of consisted of 7 tasks and took over 2.5 hours to complete, whereas the task analysed here took about 23 minutes to complete, longer than the average in the battery. We first examined data from previous, longer versions of the task and performed a pilot of 15 participants. In addition to quantitative data, these participants were de-briefed in detail by trained research assistants (RAs) who interviewed them as to whether they found the tasks tiring, interesting or difficult. Although quantitative data were in line with the literature, qualitative data suggested that some participants might, subjectively, be affected by tiredness but most importantly some found the task hard to work out and felt discouraged by this. Therefore, first, the randomization of the order of tasks in the battery was constrained, so that this task took place within the first hour of testing. Second, research assistants were assertive in enforcing short breaks between tasks and emphasizing the importance of attending to the task. Third, they reminded participants that they were playing for real money and that all decisions counted approximately equally, in monetary terms, encouraging attention to each decision rather than assume that shorter tasks paid as much as longer ones. Fourth, participants were reassured that they should not be discouraged if the best answers were not clear to them as the task progressed, but on the contrary they should proceed by trial and error and the best answers were then likely to gradually ‘sink in’. This is consistent with the Rescorla-Wagner model that we used to analyse the data. After the first 50 participants were tested under careful RA supervision, an interim analysis of the whole battery and more limited feedback from RAs was reviewed. This gave no cause for concern with respect to the present task and the quantitative parameters extracted were reassuringly compatible with those from historical laboratory samples (as well as the Results reported here).

Participants were thoroughly informed about the task, including a veridical performance-related pay component.