Seeing is believing, except when it isn’t. Optical illusions, in which the way we perceive objects is affected by their surroundings, highlight that the human visual system does not always portray reality accurately.

Scientists have long debated why. Some have suggested it is largely down to psychological processes, including previous knowledge (Witzel et al, 2011), while others have emphasised the role of “low-level” or early stages of neural processing in the eye.

A new study, published in PLOS Computational Biology, agrees with the latter. Jolyon Troscianko, a visual ecologist at the University of Exeter, and Daniel Osorio, a neuroscientist at the University of Sussex, developed a computational model based on principles of efficient coding to predict how colours appear to animals.

Our eyes send messages to the brain by making neurones fire within a range of speeds or bandwidth. Osorio and Troscianko took this bandwidth into account and assumed the brain uses its full range when processing information on contrast and patterns across different spatial scales. 

When they tested their model’s ability to account for 52 optical illusions, they found it to be highly accurate in predicting perception of colour in humans. They concluded limited neural bandwidth of the visual system to be key to explaining the role of human eyes in perception.

The model may partly explain in greater detail the mechanisms underlying some illusions.