Abstract: This paper explores how an individual who assesses outcomes relative to a reference point may develop biased beliefs when learning from experience. We consider an agent whose utility depends on both the intrinsic value of an outcome and how that value compares to expectations. A misattributing agent neglects how the utility from a positive or negative surprise contributes to her overall experience, and wrongly attributes this sensation to the intrinsic value of an outcome. Our model provides an intuition for why first-hand experience influences beliefs differently than second-hand observations containing identical information and why losses can impact beliefs more than gains. Our main results study implications of misattribution in dynamic environments. First, a misattributor’s expectations are over-influenced by recent experiences and under-influenced by earlier ones. Second, long-run beliefs grow pessimistic and undervalue prospects in proportion to their variability, leading a decision maker to abandon some risky options that are optimal. Third, when outcomes are autocorrelated, a misattributor persistently forms extrapolative and volatile forecasts about future payoffs. Applying the model, we show that (i) uncertain availability of a good can increase its perceived value, (ii) a misattributor may over-invest to insure against undesirable outcomes, and (iii) a misattributing principal may overestimate the ability of a manipulative agent who initially suppresses expectations and beats them thereafter.