Fig. 1: Volatility task details.

A An example stimulus presentation screen (prior to choice), showing the two stimuli (abstract shapes, referred to as ‘A’ and ‘B’ throughout for convenience), a fixation cross in the centre of the screen, and the participant’s monetary total below the fixation cross. Their monetary total was initialised at £1.50. The two abstract stimuli presented were changed between blocks (i.e. after 80 trials, and after 160) and the same pair of stimuli was always presented within a block, though the side of the screen they appeared on was counterbalanced. Participants were encouraged to take breaks in between blocks. After participants made a choice, they could receive wins or losses. The stimulus-outcome associations for one stimulus, referred to here as stimulus A, are shown in panel (B). The associations were exactly reversed for the other stimulus presented in that block – i.e. for stimulus B, the probability of a win is 1 minus the probability of a win for shape A. If, on any given trial, ‘win’ feedback was shown for one stimulus, it could not be shown for the other. However, the summed probability of both outcomes for a given shape, i.e. win + loss, was not 1 as wins and losses were independent. Note that in block 1 (up to trial 80, marked with a dotted line) both win and loss outcomes are volatile, and change between a high probability (0.85) and a low probability (0.15). In blocks 2 and 3 (trials 81:160 and 161:240; order counterbalanced between participants), one outcome was stable (0.5 probability of that outcome resulting from choosing the relevant stimulus), and one was volatile (changing between 0.85 and 0.15 probability of outcome receipt). When outcomes were volatile, the probability of each outcome changed with a frequency of between 14 and 30 trials. The total probability within a block that wins and losses were associated with any given stimulus averaged 50%, such that the task did not systematically favour either of the shapes within a block. As can be noted from panel B, it is possible for a stimulus to be associated with both win and loss, neither win or loss, win only, or loss only. These four possible outcomes are shown as example trial sequences in panel (C). The box around shape B reflects the choice the participant made on that trial (so, in this instance, they selected shape B). Regardless of which stimulus was chosen, the outcomes associated with both stimuli were shown. From top to bottom, these are (i) win associated with shape A and loss associated with shape B, (ii) both win and loss associated with shape A, (iii) both associated with shape B, and (iv) win associated with shape B, and loss associated with shape A. Note that these example sequences show the win outcome being displayed before the loss outcome: in the actual task, the order of presentation of these two outcomes was counterbalanced (reflected by the dotted arrows), and there was a jittered delay of between 2 and 6 s before the other outcome was presented. A win outcome resulted in an addition of 15p to their monetary total; a loss resulted in 15p being deducted from their total. Notably, win and loss outcomes were independent, such that knowing the stimulus associated with the win was not informative about the location of the loss. Participants were required to learn over time which stimulus was associated with wins and losses and were asked to aim to maximise their wins and minimise their losses.