Moreover, a difficulty for a sampling difference explanation is the fact

that full and identical feedback, of chosen and unchosen gamble outcomes, was presented to both actors and observers. However, sampling errors may occur at the level of attention rather than choice, where certain outcomes may be deemed to hold more personal relevance than others. The active nature of operant learning could also engage the actor and improve efficiency of learning (Cohn et al., 1994), although this would be predicted to occur across the full range of probabilities. In this article, we demonstrate a difference in value learning between acting and observation, an effect not previously GW786034 price reported to the best of our knowledge. These findings have important implications for how we apply learning theory to vicarious learning, either social or non-social, as classical models assign no differences to these alternative models of learning. This bias in learning indicates that action-outcome contingency learning depends on the manner through which it is learned, and indicates that actors TSA HDAC mw and observers implement different weightings for positive and negative experiences as they sample outcomes. As we are interested in the mechanisms underlying this effect, we excluded two important alternative explanations. In Experiment 2 we rule out a value-specific order effect on learning, while in Experiment 3 we show that this effect is driven by poor estimation

of value rather than of probability. This leaves open a possibility that the effect reflects an optimistic bias in observational learning leading one to underestimate the likelihood of experiencing negative events, as observed occurring to others, a bias not present in actors learning by direct experience as in trial-and-error. To provide a more precise account we believe selleck products requires additional experimentation. In particular, the fact that the effect is specific to the lowest value option of the choice set (i.e. only the 20% win

option) could indicate that this over-valuation is a non-linear effect of value learning present over-and-above a certain threshold. This non-linear effect may also be explained by a critical role of context in value learning, whereby observers’ over-valuation is only for options that are of low value relative to either the whole choice set (i.e. 20% win options were the lowest value in the choice set) or to the alternative option in the pair (i.e. 20% win options were the only option never paired with an option of an even lower value). Indeed such reference dependent effects on subjective representations of value are supported by an extensive psychological (e.g. Kahneman and Tversky, 1979 and Mellers, 2000) and neuroscience literature (e.g. Breiter et al., 2001, Elliott et al., 2008 and Tremblay and Schultz, 1999). This work was supported by a Wellcome Trust Programme Grant to RJD and a Brain Research Trust Prize Studentship to AN. We thank Jeffrey M.