Look again: Brain rhythms show what humans see.

Do rhythms in the brain serve a function or not? This question has been much debated by scientists. Now, a new study co-authored by ESI scientists shows how gamma-band oscillations in the visual system reflect what humans see. The results refute conclusions of a previous study and support the theory that gamma-oscillations do serve a function in information processing by the brain.

18 Jul 2019


The brain is full of rhythms that show quite something about what the brain’s owner is up to: Simple EEG recordings are enough to recognize readily if somebody is sleeping, daydreaming, or concentrating on a task. So the connection between these swinging brain rhythms and cognitive states is well established. However, the nature of this connection has remained cryptic: Are oscillations in the brain a mere by-product of neural activity? Or do they have a function in information processing within the cerebral cortex?

“If we look at brain oscillations, particularly in what we call the gamma-band between 30 and 80 Hertz, we have seen over and over that the firing thresholds for neurons change in correspondence with them – I can’t imagine that this doesn’t have a function.” explains Nicolas Brunet, Assistant Professor at Millsaps College in Jackson Mississippi and lead-author of a new study that shows, for the first time, a correlation between visual perception under natural stimulation and gamma-band activity in the human brain. The study puts a basic assumption to the test: Oscillations serve a function in how we perceive things. If they are important for example for seeing things, they should be present whenever there is something to see.

Oscillations: Always present when there's something to see?

As simple as that sounds, until recently we didn’t know if that’s actually the case. Whatever there is to see during an experiment usually doesn’t have much in common with things to see in real life. Research participants, humans as well as animals, usually get to look at checkerboards or gratings, in which contrast and colour are perfectly controlled. Many studies have shown that those simple laboratory stimuli are very potent in evoking strong gamma-band activity. But what happens when people look at more realistic images? This is something Nicolas Brunet explored a couple of years back in a study with macaque monkeys, as ESI Director Pascal Fries remembers: “Initially I was sceptical. I expected the natural and thus uncontrolled stimuli would produce ambiguous results. However, the data proved to have a very clear message: Every single of the 65 natural images we presented, elicited notable gamma-oscillations.”

Because of this, both Pascal Fries and Nicolas Brunet were very surprised to hear a team of scientists at Stanford, probing the phenomenon in a human subject, hadn’t been able to reproduce the findings. Due to severe epilepsy, the study’s subject was implanted with intracranial electrodes touching the surface of the visual cortex. The US scientists had the patient view black and white photographs of everyday snapshots like houses, cars or faces. In their report, only half of these images were accompanied by a significant increase gamma-band activity. As oscillations were absent for pictures that could be seen well, the authors concluded that the brain-rhythms do not play a functional role in perception.

Disagreements over theories is part of science, thinks Nicolas Brunet and explains: “It’s like a game of ping pong. You keep passing the ball until someone puts forward a match-winning, empirical argument.” For now Brunet is not ready to put the bat aside. He and Pascal Fries repeated the analysis of the human data themselves.

Revelation at second sight

Taking a closer look at the data they found gamma oscillations did in fact increase with every stimulus. This increase was easy to miss as it depended on the image structure. Images with little structure and larger areas of uniform grey resulted in weaker gamma-responses. Images with strong structures also produced strong gamma-band responses. Using an algorithm from computer vision Nicolas Brunet and Pascal Fries managed to quantify the correlation between image structure and rhythmic brain activity. It turned out that the amount of information contained in the gamma-band spectra was large enough to allow the scientists to tell from the oscillations alone, with 70% accuracy, which of two randomly drawn images had been presented.

More to see, more to swing – measuring image structure is a way to quantify how much information a picture contains. Cloudy grey sky like the one surrounding the “Founders Tower” on Millsaps College Campus (left) results in a low structure image (0.03 using DCT Energy measure operator). The image of the girl waving the college flag (right) is rich in details and thus has high image structure (0.31). (photographs: Nicolas Brunet)

More to see, more to swing – measuring image structure is a way to quantify how much information a picture contains. Cloudy grey sky like the one surrounding the “Founders Tower” on Millsaps College Campus (left) results in a low structure image (0.03 using DCT Energy measure operator). The image of the girl waving the college flag (right) is rich in details and thus has high image structure (0.31). (photographs: Nicolas Brunet)

Compelling results, but they come with a limitation: The data derives from a single subject. Surgery is a last resort of hope that is offered only to patients with severe epilepsy. Of those few patients who undergo surgery, even fewer require electrodes to be placed over the visual areas of the brain. Despite the single case report nature of the study, the significance of the results is not to be underestimated, believes Nicholas Brunet: “Most of our observations are made studying monkeys. And sure, humans are close to monkeys, but in the end monkeys will be monkeys.” Albeit limited to one subject, the human data does imply that observations made in macaque monkeys are transferable to people at least in this case. Whether monkey or human: The more there is to see, the more there is in swing. This shows that gamma-band oscillations are not limited to animals and artificial stimuli but do indeed play a role in how humans see the real world.