Age-related prefrontal impairments implicate deficient prediction of future reward in older adults.

Foresighted decision-making depends on the ability to learn the value of future outcomes and the sequential choices necessary to achieve them. Using a 3-stage Markov decision task and functional magnetic resonance imaging, we investigated age differences in the ability to extract state transition structures while learning to predict future reward. In younger adults learning was associated with enhanced activity in the prefrontal cortex (PFC). In older adults (OA) we found no evidence for PFC recruitment. However, high-performing OA showed enhanced striatal activity, suggesting that they may engage in a model-free (experience-based) learning strategy. Change point analyses revealed that in younger adults learning was characterized by distinct and abrupt shifts in PFC activity, which were predictive of behavioral change points. In OA PFC activity was less pronounced and not predictive of behavior. Our findings suggest that age-related impairments in learning future reward value can be attributed to a deficit in extracting sequential state transition structures. This deficit may lead to myopic decisions in OA if contextual information has to be temporally integrated.