<p>In uncertain environments, intelligent decision-makers exploit actions that have been rewarding in the past, but also explore actions that could be better. Several studies link exploration to pupil size—a peripheral correlate of neuromodulatory tone and arousal. However, pupil size may only track variables that make exploration more likely, such as volatility or reward, without directly predicting exploration or its neural bases. Here, we simultaneously measured pupil size, exploration, and neural population activity in the prefrontal cortex while two male rhesus macaques explored and exploited in a dynamic environment. We find that pupil size under constant luminance specifically predicts the onset of exploration beyond effects of reward history. Pupil size also predicts disorganized patterns of prefrontal activity at the single neuron and population levels. Our results support a model in which pupil-linked mechanisms drive exploration by pushing prefrontal dynamics through a critical tipping point.</p>