How do we learn? This is a question that has interested me since my first day of graduate school, and even before then, when I was a high school teacher. One of the most important ways that we learn is through feedback. You are mostly likely familiar with the old adage, "We learn from our mistakes". As it turns out, this is not entirely true. We learn when our expectations differ from outcomes. Formally, we call this a prediction error - the difference between an outcome and an expectation.
Consider the following example. You write a midterm exam. You expect to get an 80, but you actually get a 70. This would be a prediction error. Outcome = 70. Expectation = 80. Prediction Error = 70 - 80 = -10. Things are worse than expected! The reverse could also be true. You think you failed an exam and got a 40. But, you actually got 80. A massive prediction error of +40 - things are a lot better than expected. And of course, you may have no prediction error. Consider mini golf. You have an easy putt. You make it. Your expectation equals your outcome, no prediction error.
What is crucial here, is that these prediction errors drive learning. Positive prediction errors strengthen behaviours, negative prediction errors discourage behaviour. We call this reinforcement learning (or operant conditioning). You may be familiar with this if you have ever trained a pet or a young human.
What is cool, is that we can measure this in humans using electroencephalography, or EEG.
Consider the following example. You write a midterm exam. You expect to get an 80, but you actually get a 70. This would be a prediction error. Outcome = 70. Expectation = 80. Prediction Error = 70 - 80 = -10. Things are worse than expected! The reverse could also be true. You think you failed an exam and got a 40. But, you actually got 80. A massive prediction error of +40 - things are a lot better than expected. And of course, you may have no prediction error. Consider mini golf. You have an easy putt. You make it. Your expectation equals your outcome, no prediction error.
What is crucial here, is that these prediction errors drive learning. Positive prediction errors strengthen behaviours, negative prediction errors discourage behaviour. We call this reinforcement learning (or operant conditioning). You may be familiar with this if you have ever trained a pet or a young human.
What is cool, is that we can measure this in humans using electroencephalography, or EEG.
|
|
On the left, you can see an event-related brain potential, or ERP. It is basically the difference in the EEG signal between doing something right and doing something wrong. Your brain processing feedback, or prediction errors. We even know where in the brain this signal comes from, the Anterior Cingulate Cortex - you can see this on the right.
As a part of our research program, we study human learning, and in particular, reinforcement learning. We are interested in this signal, which we call the reward positivity, the neural systems that generate it, and the factors that influence it. We use a variety of tools to study it - behavioural data, EEG, fNIRS, fMRI, and computational modelling.
Representative publications in this area:
Holroyd, C. B., & Krigolson, O. E. (2007). Reward prediction error signals associated with a modified time estimation task. Psychophysiology, 44(6), 913-917.
Krigolson, O. E., Pierce, L., Tanaka, J., & Holroyd, C. B. (2009). Learning to become an expert: Reinforcement learning and the acquisition of perceptual expertise. Journal of Cognitive Neuroscience, 21(9), 1834-1841.
Krigolson, O. E., Hassall, C., & Handy, T. C. (2014). How we learn to make decisions: The rapid propagation of reinforcement learning prediction errors in humans. Journal of Cognitive Neuroscience, 26(3), 635-644.
Hammerstrom, M. R., Binsted, G., & Krigolson, O. E. (2025). Differential neural processing of reward and self-relevance in a social gambling paradigm. Cognitive, Affective, & Behavioural Neuroscience, 25(2), 377-386.
As a part of our research program, we study human learning, and in particular, reinforcement learning. We are interested in this signal, which we call the reward positivity, the neural systems that generate it, and the factors that influence it. We use a variety of tools to study it - behavioural data, EEG, fNIRS, fMRI, and computational modelling.
Representative publications in this area:
Holroyd, C. B., & Krigolson, O. E. (2007). Reward prediction error signals associated with a modified time estimation task. Psychophysiology, 44(6), 913-917.
Krigolson, O. E., Pierce, L., Tanaka, J., & Holroyd, C. B. (2009). Learning to become an expert: Reinforcement learning and the acquisition of perceptual expertise. Journal of Cognitive Neuroscience, 21(9), 1834-1841.
Krigolson, O. E., Hassall, C., & Handy, T. C. (2014). How we learn to make decisions: The rapid propagation of reinforcement learning prediction errors in humans. Journal of Cognitive Neuroscience, 26(3), 635-644.
Hammerstrom, M. R., Binsted, G., & Krigolson, O. E. (2025). Differential neural processing of reward and self-relevance in a social gambling paradigm. Cognitive, Affective, & Behavioural Neuroscience, 25(2), 377-386.