We are surrounded by salient cues and actions in our everyday lives that predict reward in a constantly changing environment. A critical feature of goal-directed behavior is the ability to discriminate stimuli or actions that predict reward from those that do not, and further, to flexibly update our responses if predictions become inaccurate. This is referred to as behavioral flexibility, which measures our ability to re-evaluate previously learned associations that predicted reward and adjust our current responses following changes in the environment. Impairments in behavioral flexibility can lead to maladaptive behavior and impairments in decision-making, and is often associated with many neuropsychiatric disorders. The research presented here investigated the behavioral and neural basis of flexible learning under uncertainty by manipulating the associations and probability of reward outcomes in the environment, following alcohol experience and chemogenetic inhibition. Altogether, this approach establishes a causal link between brain and behavior and increases our understanding of both adaptive and maladaptive decision-making. We first tested the effect of prior alcohol exposure on flexible learning using a probabilistic, stimulus-based reversal learning task. Sex differences in alcohol consumption emerged, with females exhibiting heavier alcohol consumptions patterns than males. Furthermore, all alcohol-exposed animals, regardless of sex, were slower to learn, exhibited attentional deficits, and were less sensitive to negative feedback, compared to water-matched controls. Next, we compared the performance of rats on stimulus-based reversal learning in a rich (90/30) and poor (70/30) reward environment. All rats exhibited increased preservation during early reversal learning, resulting in slower learning. Additionally, a sex-dependent effect emerged on latencies to choose the better option, with females exhibiting longer latencies than males during reversal learning, an indication of slower processing speed. And finally, chemogenetic techniques were used to inhibit pyramidal neurons in the ventrolateral orbitofrontal cortex (vlOFC) and basolateral amygdala (BLA), on both a stimulus- and action-based reversal paradigm. Overall, the BLA seemed to be more necessary for learning action-based probabilistic reversals, as inhibition of this region resulted in poorer learning; whereas, the OFC was more necessary for detection of stimulus-based reversals. There was also evidence of sex-modulated learning flexibility, such that females with vlOFC inhibition were slower to learn during both deterministic and probabilistic action-based reversals compared to non-inhibited females; whereas, females with BLA inhibition were more impaired on probabilistic action-based reversals than non-inhibited females, this pattern was not observed in males. Altogether, these findings highlight the importance of including sex as a biological variable, as all of these studies found sex-dependent effects.
