The dopamine reward prediction error signal

• Published in: Handbook of Behavioral Neuroscience Vol. 32; pp. 225 - 236
• Main Author: Schultz, Wolfram
• Format: Book Chapter
• Language: English
• Published: Elsevier Science & Technology 2025
• Subjects: Choice; Drug addiction; Learning; Neuroeconomics; Parkinson's disease; Prediction; Reinforcement; Subjective value; Utility; Learning; Drug addiction; Utility; Neuroeconomics; Choice; Parkinson's disease; Prediction; Reinforcement; Subjective value
• ISBN: 9780443298677; 044329867X
• ISSN: 1569-7339
• DOI: 10.1016/B978-0-443-29867-7.00027-X

The fastest and strongest dopamine response is the bidirectional reward prediction error (RPE) signal. The RPE response to reward-predicting stimuli and ultimate rewards reflects their subjective reward value relative to their predicted subjective reward value. With these conceptually defined properties, the dopamine RPE response is a suitable teaching signal for reinforcement learning. The RPE signal is preceded by a unidirectional, smaller, prediction-dependent response that unspecifically reflects the salience of any external event, including punishers, stimuli resembling rewards, and novel stimuli. The salience and RPE responses are distinguishable from later, slower, and lower dopamine excitations and inhibitions that occur with a wide range of sensory and motor events that generate behavioral activation. Distinction of these multiple dopamine signals requires behavioral techniques with good discrimination of behavioral events and recording techniques with suitable temporal resolution.