image: Earl Miller speaking at MIT in 2022.
Credit: Faith Ninivaggi/MIT Picower Institute
Over 30 years in his lab at The Picower Institute for Learning and Memory at MIT, Picower Professor Earl K. Miller has studied how the brain’s cortex produces thought. On Nov. 15, in an invited presidential lecture at the Society for Neuroscience annual meeting, he will tell the audience what reams of experimental evidence have led him to propose: Cognition and consciousness emerge from the dynamic organization of the cortex produced by traveling brain waves performing analog computations.
Analog computing is an old idea that embraces how information in the real world is continuous, not chopped up into digital binary bits. Just two waves colliding with each other can smoothly represent any value from the negative to the positive sum of their amplitudes simply based on their phase. Miller says it’s no coincidence that the brain, where the coordinated electrical activity of millions of neurons produces large-scale oscillations across a broad range of frequencies all the time, evolved to exploit the information-rich, fast-propagating, and reliable efficiency that its waves provide. Sure enough, numerous experiments by his lab and others over decades provide evidence that brain waves sculpt the flow of information in the cortex and enable volitional, goal-directed control of thought.
“The brain uses these oscillatory waves to organize itself,” says Miller, a faculty member in MIT’s Department of Brain and Cognitive Sciences. “Cognition is large-scale neural self-organization. The brain has got to organize itself to perform complex behaviors. Brain waves are the patterns of excitation and inhibition that organize the brain, and this leads to consciousness because consciousness is this organized knitting together of the cortex.”
Organizing neural computation
The nature of consciousness is one of the most hotly debated topics in neuroscience (and other fields such as philosophy), but many biologically based theories (e.g. Global Neuronal Workspace Theory or Integrated Information Theory) agree that a unified awareness of thought and experience requires a cortex unified by information exchange among at least some of its regions. Miller says what his evidence adds to such ideas is that brain waves not only unify the cortex, but also organize it with analog computation to control and accomplish information processing.
In 2001, Miller co-authored a seminal review paper, one of the most cited in all of neuroscience, arguing that the prefrontal cortex implements executive control of the brain by actively maintaining goal-directed activity patterns that bias the functions of other regions around the brain. Around the same time, culminating in a 2013 paper, he and colleagues showed that many neurons in the prefrontal cortex are not dedicated to specific tasks, like gears in a machine, but are instead multi-functional. They can participate in multiple networks at once.
But what organizes these neural networks to implement widespread, goal-directed function? The answer began to emerge in 2007, when Miller’s lab published the first in a long series of papers—continuing through the present day—that show how finely patterns of neural oscillations can sculpt information flow across the cortex. “Top-down” goal-directed signals (the brain’s internal sense of the rules), are encoded in relatively slow alpha and beta frequency waves (15-35 Hz) while incoming sensory information is encoded in higher frequency gamma waves (35-60Hz). Subsequent studies have shown that in cognitive functions such as working memory and making predictions, the beta waves constrain the power of the gamma waves, essentially imposing the brain’s goals on information processing. Notably, this regime enables volitional control of thought: When you want to retrieve information committed to working memory (like today’s specials at a restaurant), your cortex can relax beta power to let gamma fetch that for you.
Miller’s work has shown that this organization extends across the cortex. In a paper in 2020, his lab showed that the peak frequency within each band increases continuously from the back of the brain to the front. And last year he and co-authors showed that across multiple mammalian species, including humans, the cortex showed a consistent motif in which deeper layers of cortex produced peak power in the beta range while superficial layers produced peak power in the gamma band.
Many other studies in Miller’s lab have shown that brain waves physically travel within the cortex, either linearly or in rotations. Miller has associated properties of those travels with specific functions, including working memory, because they may provide computational advantages such as constantly maintaining information and providing a clock for computation.
In 2023, Miller and collaborators combined these notions of the spatial interplay of different wave frequencies that represent different kinds of information to produce a new theory of cognition in the cortex: Spatial Computing. The theory proposes that the brain sculpts neural networks by applying this brainwave interplay to neurons within physical patches in the cortex. Beta waves essentially act like stencils that constrain gamma activity in those patches. When goals and rules (represented by beta) dictate that information represented by the neurons in a patch are needed, gamma wave power can rise to enable retrieval and processing of that information.
Other advantages of waves, Miller notes, include their speed. Neurons famously rewire their network connections, or synapses, all the time but that process is much too slow to provide the flexibility needed for cognitive tasks, he said. Waves, and underlying electric fields, can propagate across long distances of cortex much faster. Miller’s lab has also shown that electric fields representing whole ensembles of neurons also represent information more reliably than individual neurons do.
Waves and consciousness
To Miller, “consciousness is the tip of the iceberg of cognition.” Brain waves and their analog computations do a lot of cognitive work without your explicit intervention, but they also enable volitional control. For instance, you don’t always consciously direct your brain to keep track of potential changes in your environment (e.g. a sudden peal of thunder), but it can surface those changes when it spots them to enable you to consciously think about what to do (going back inside to get a raincoat).
“Consciousness is there for planning behavior before you engage in it, and for countermanding ongoing decisions that are going to be stupid,” Miller says. “In that second mode, it’s almost as if consciousness is the story your brain makes up to explain what it just did…It’s there to keep tabs on itself and plan the future. Otherwise consciousness is hanging out there just following along.”
In numerous studies over the last several years, often in collaboration with Picower Institute colleague and anesthesiologist Emery N. Brown, Edward Hood Taplin Professor of Computational Neuroscience and Medical Engineering, Miller has connected consciousness to the brain wave dynamics he has studied by looking at how general anesthesia affects them. Anesthesia, after all, provides a clean dividing line between consciousness and unconsciousness. The studies have shown that anesthesia disrupts the power of waves in different frequencies, knocking them out of the normal beta-gamma balance needed to organize cognition. They have shown that anesthesia disrupts propagation of waves linking sensory and higher order regions of the cortex that provides unification and organization. They have shown that anesthesia alters the travels of traveling waves. And in a study earlier this year, the lab also showed that distinct anesthetic drugs push brain waves out of phase with each other, which would disrupt their ability to engage in analog computations.
Miller’s proposal is by no means fully proved, but work in his lab is ongoing. By presenting it at SfN, he hopes that more colleagues will become, well, conscious of it.