Fig. 1
Schematic overview of the training sessions. a Timeline study procedure. b Example of a trial on the Information Bias Learning Task (IBLT). At the beginning of the task, participants are provided with a start amount of £1.50. In each trial, a fixation cross flanked by two abstract stimuli is presented and the participant has to choose one of the stimuli via a button press. Once a stimulus is chosen, a win and a loss outcome are presented consecutively, with the order of their appearance (win first versus loss first) being randomised across trials. If the chosen stimulus is associated with a win the participant gains 10p, and if the chosen stimulus is associated with a loss the participant loses 10p. If the win and loss outcome both appear over the same stimulus, the participant does not win or lose any money irrespective of their choice. The win and loss outcomes are independent, meaning that the location of the win does not provide any information about the location of the loss. In this task, participants have to learn through experience which stimulus to choose in order to maximise total winnings. c Structure of the IBLT for negative training in Study 1. The task consisted of 5 blocks comprised of 80 trials each (vertical, dashed black lines separate the individual blocks). The x-axis represents the number of trials, with the y-axis indicating the probability p of an outcome appearing over stimulus ‘A’. The probability of the outcome appearing over stimulus ‘B’ can be calculated as 1 − p. The win outcomes are represented as continuous green lines, with the loss outcomes corresponding to the dashed red lines. The volatility of the win and loss outcomes is manipulated across the task blocks, with higher volatility being associated with a higher information content. In the first block, both the wins and losses are volatile (‘Both-volatile’ block), with the probability of an outcome appearing over stimulus ‘A’ switching between 20 and 80%. Here, both outcomes have a high information content, such that if the win/loss appears over shape ‘A’, it is more likely to be associated with shape ‘A’ than shape ‘B’ in the subsequent trials. In this block, participants are therefore expected to have high learning rates for both wins and losses. In blocks 2–4, on the other hand, volatility is manipulated so that losses are highly informative and wins are uninformative (‘Training’ blocks). Whereas the loss outcomes remain volatile, the association of shape ‘A’ with the win outcome is stable at 50%. Thus, the chance of the win appearing over either of the shapes remains equal across the trials, with its location on one trial providing no information about future trials. In these ‘Training’ blocks, it is expected that participants will have higher learning rates for loss than win outcomes. Finally, block 5 consists of another ‘Both-volatile’ block, in which both wins and losses are volatile. By comparing learning rates in block 5 with block 1, it is possible to quantify potential shifts in learning from win and loss outcomes following the ‘Training’ blocks. d Structure of the IBLT for positive training in Study 2. Similar to Study 1, the IBLT is comprised of a ‘Both-volatile’ block, three ‘Training’ blocks, and a final ‘Both-volatile block. However, volatility of the win and loss outcomes in the ‘Training’ blocks is reversed, such that win outcomes are highly informative (volatile) and loss outcomes are uninformative (stable). Therefore, contrary to Study 1, participants are expected to demonstrate higher learning rates for win than loss outcomes in the ‘Training’ blocks of Study 2 (Color figure online)