Expectations profoundly influence perception and emotion. Computational models of reinforcement learning provide fruitful descriptions of how expectations dynamically develop in response to rewards and punishments. Dr. Atlas' talk focused on the relationship between expectations, learning, and subjective experience. Part 1 focused on the relationship between cue-based expectations and pain. Expectancy effects on pain are highly stable, despite neural signals consistent with dynamic learning. This is likely due to the fact that knowledge and expectations can directly influence learning. However, most quantitative learning models consider reinforcement history alone, irrespective of knowledge and expectations. Part 2 described evidence that instructed knowledge dissociates the neural mechanisms underlying feedback-driven learning: the striatum and orbitofrontal cortex (OFC) learn from instructions, whereas the amygdala learns from reinforcement alone. Expectations appear to slow associative learning and reduce striatal responses. Collectively, this body of work indicates that emotional learning is highly sensitive to cognitive factors and underscores the complex relations between expectations, learning, and subjective experience.