Brain states, defined by global activity changes across brain regions are arguably the most important determinant of behavioral output in mammals. Sleep is the longest state that most animals spend their lives in and comprises distinct sub-states crucial for different physiological and behavioral processes. Classification of different sleep stages is therefore essential for understanding and studying processes taking place throughout sleep in health and disease in both mice, the most common animal model in neuroscience, and humans. Sleep stage classification has typically been done through analysing multiple physiological measures including electroencephalography, electromyography and electrooculography. While these techniques have enabled numerous studies, they remained complex, expensive and tedious to implement. Other auxiliary measures such as heartbeat, breathing and skin conductance revealed body-wide changes across sleep and their correlation to the autonomic rhythms and neuromodulator levels. However, these signals have a too low time resolution to derive significant results from. We have taken inspiration from studies involving eye pupil tracking in wakefulness to predict brain states and translated them into sleep settings by merging them with novel infrared illumination methods for both mice and humans. In this thesis, I present findings on pupil tracking during sleep in mice and humans. We discovered a robust correlation between brain states and fluctuations in pupil size as well as eye movements. Our findings suggest a strong parasympathetic control of pupil size during sleep in both species and the involvement of sympathetic pathway in humans, indicating fragility and arousal periods in sleep. We furthermore show that pupil constrictions during sleep might serve a protective function to preserve the stability of deep sleep. Taken together, our research revealed a reliable relationship between pupillary dynamics in sleep and brain states, which has so far been hidden behind closed eyelids. It underlines the importance of developing novel methods for physiological tracking in sleep that can be used for assessments in clinical or research settings.