Reward Prediction Error — Surprise as a Teaching Signal

Reward prediction error (RPE) is the difference between the reward you expected and the reward you actually got. In Wolfram Schultz’s classic work, dopamine neurons in the VTA fire when an unexpected juice reward appears; once the animal learns a cue predicts that juice, the dopamine spike moves to the cue. If the predicted juice fails to show up, dopamine activity dips below baseline.

Formally, RPE ≈ (actual reward – expected reward). Positive error (better than expected) strengthens the behaviours and cues that led there. Zero error (as expected) says “keep doing this, no need to re‑learn it.” Negative error (worse than expected) weakens the association, nudging you away from wasting effort next time.

This simple mechanism underpins a lot: why intermittent rewards (like variable social media feeds or gambling) are so sticky; how habits sharpen with practice; and how the brain learns to treat certain foods, people or apps as important even when they no longer feel especially good.

Why It Matters

Realising your brain is constantly comparing “what I thought would happen” with “what actually happened” explains both the thrill of surprise wins and the sting of disappointment — and why changing expectations can be as powerful as changing rewards.

Closing Line

Reward prediction error is your nervous system’s nudge to update the story: when the world deviates from the script, dopamine edits the next draft.