Fig. 1: Approach/avoidance conflict task.

a Top: Loss probability (indicated by the frame colour) and magnitude (number of red diamonds) are shown at trial start, while reward magnitude is always one token. The player (green triangle) is located at the safe position at the bottom corner of the grid and an inactive predator (grey circle) is located at the top corner. After a random time interval, a token appears either on the left or the right side, and disappears after another random time interval. Bottom: If the player leaves the safe position to approach the token, one of three outcomes ensues: a positive outcome P, if the player collects the token and safely returns to the initial position; a negative outcome N, if the predator wakes up and catches the player, causing the loss of a number of tokens; or a rare neutral outcome (not shown), in which the token disappears before the player can reach it. Alternatively, the player can decide to not perform an action and avoid the token. The trial ends 1000 ms after the disappearance of the token. b Behavioural results. Approach rate (left) and average approach latency (right) for each loss probability and magnitude; error bars represent the standard error of the mean (s.e.m.). See Table 1 for inference statistics.