Fig. 8.

Illustration of unlearning and the effect of cue–outcome order in Experiment 2. In this illustrative example, cues are the colours of the squares: red, blue or yellow; outcomes are the circle colours: black or grey. Arrow thickness indicates (very approximately) connection strength. Discriminative order (left). Cues precede outcomes, so they can compete for connection strength. Top panel: red and blue predict the black outcome and gain connection strength. Middle panel: blue and yellow gain connection strength to the grey outcome – simultaneously, they lose connection strength to the black outcome. Bottom panel: in a later trial, blue and red again occur with the black outcome. But the connection weight from blue has weakened (thinner arrow), because it occurred with a different outcome in a previous trial (i.e. in the middle panel). Blue is downweighted so that its connection strength becomes weak. Over time, blue is unlearned. In addition, the connection strength of red and blue to the grey outcome weakens, due to their occurrence with the black outcome on the current trial. Non-discriminative order (right). Outcomes precede cues, so there is no competition between cues for connection strength. Top panel: connection strength increases between co-occurring items. Middle and bottom panels: connection strength increases between co-occurring items and decreases between items that do not co-occur. Connection strength simply fluctuates up and down. No difference in the relative connection strength between cues and outcomes occurs. Only the conditional probabilities are learned. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)