Why is a significant part of consciousness, such as dreaming during sleep, disconnected from external reality? Alan Hobson’s ‘protoconsciousness’ hypothesis suggests REM sleep is essential for developing and maintaining consciousness. Karl Friston expanded this idea, proposing that REM sleep helps optimize the brain’s predictive processing. The theory implies that animals displaying REM sleep could potentially be conscious. Evidence shows that most animals, not just mammals and birds, experience distinct sleep stages, including REM sleep.

A team led by Bruno van Swinderen at The University of Queensland’s Brain Institute will explore REM sleep’s role in supporting consciousness across species by investigating neural mechanisms of consciousness transitions in the fruit fly brain. The project adopts a bottom-up approach, using the fly’s compact, fully-mapped brain to study how predictive processing varies across states of consciousness (wakefulness, sleep, anesthesia) and attention.

Two questions guide the research:

1. Does loss of consciousness represent a collapse to hardwired predictions?
2. Does sleep ensure prediction errors remain functionally flexible?

Four aims structure the project:

• Tracking prediction errors across general anesthesia induction and recovery by investigating how predictive processing in the fly brain changes during transitions into and out of anesthesia, comparing neural activity profiles (via electrophysiology and whole-brain imaging) between conscious and unconscious states.
• Tracking prediction errors across sleep and wake by analyzing how sleep stages (spontaneous and induced) modulate prediction error signals, focusing on differences between quiet sleep, active sleep, and wakefulness, using long-term electrophysiology and calcium imaging.
• Comparing prediction errors between attended and unattended stimuli by examining how selective attention and embodiment influence predictive coding in awake flies, using closed-loop virtual reality experiments and behavioral tracking to contrast neural responses to attended vs. ignored stimuli.
• Modelling predictive processing in the conscious vs unconscious fly brain by simulating visual sequential expectation paradigms on the Drosophila connectome to identify neural circuits involved in prediction error detection and test whether static wiring or dynamic oscillations explain conscious vs. unconscious processing profiles.