The Swiss Army Knives of Neural Circuits: Mixed Selectivity Neurons

Among the most prominent goals of UChicago’s Neuroscience Institute is to make a major contribution to the understanding of neural circuits—how information is processed by networks of neurons. But experimentation to understand neural circuitry is costly and complicated, in several ways.

First, one must hone a hypothesis to get at the concept to be tested. Then you need to create the experimental environment, perhaps an immersive space where a mouse can float on air or a computer game that a monkey can play. Next, you need to fit the animals with electrodes to record how their neurons are reacting in the experiment. Then comes the really labor-intensive part: teaching the animal how to perform the behaviors involved in the experiment. All this before one can actually collect experimental data.

Several years ago, David Freedman, PhD and MD/PhD student Chris Rishel, now an instructor at Stanford University, were studying a part of the brain known to play a key role in visual and spatial processing. They wanted to see if the area was also involved in categorization—the ability to identify stimuli as, say, moving toward you or away from you.  

Over a period of about a year, Rishel and Freedman went through all the steps above, designing and building an experiment to understand how categorization decisions were made in one area of the primate brain. They trained the monkeys to move a lever when a new stimulus matched the category of a sample shown earlier. They also taught the monkeys, in certain cases, to move their eyes in a particular direction after the categorization decision, independent of whether they had moved the lever or not. This secondary part of the experiment was designed to understand how multiple behavioral functions, such as cognition (putting things into categories) and movement planning (in this case, eye movements) are processed within a pool of neurons.

They found that not only were both functions processed in the same region, but also sometimes even by the same brain cells. This finding provided important clues about how the brain’s neural circuits can participate in multiple functions.

Several years later, Jeff Johnston, a newer graduate student in Freedman’s lab, became interested in how our brains maintain reliable perception and behavior despite the obvious variability in individual neurons. If all processing were linear—a superhighway from one specific neuron to another—then wouldn’t one neuron misfiring cause the whole system to fall apart?

Previous circuit research provided some clues that certain brain areas might process sensory, decision-making, and motor data using neurons which show “mixed selectivity.” These neurons can code information flexibly—even simultaneously—for multiple types of tasks. Johnston wondered whether these neurons might be a source of the brain’s incredible reliability and efficiency. He also knew that direct experiments to understand the computational benefits of mixed selectivity were challenging, as they depend on monitoring the activity of many neurons in a particular brain area all at once, something so far technically infeasible.

So Johnston decided to mine the experimental data Rishel and Freedman had so laboriously generated in their categorization/movement study. His goal: to generate a theoretical basis for reliable neural processing via a “mixed selectivity” coding scheme, in which these neurons might be operating as tools in both decision-making and motor circuits at once.

With a background in statistical analysis and computer science and support from physicist/neuroscientist Stephanie Palmer, Johnston set out to derive how this might work. His algebraic analysis of Freedman and Rishel’s data showed that circuits that used mixed selectivity neurons had decoding errors orders of magnitude fewer than linearly dynamic neurons—whether the processing was sensory, motor, or abstract. Using mathematics and the earlier data to build theory, Johnston’s analysis suggests that nonlinear mixed selectivity may be pervasive throughout the brain—and the key to getting things “right” despite occasional poor performance by individual neurons.

And this work took Johnston MUCH less time than Freedman and Rishel’s original experiment.

Building theory from the experimental data is not just faster, it can also be more powerful and efficient, just like mixed selectivity neurons. Johnston’s mathematical analysis now can help experimentalists overcome the technical difficulties cited above and zero in on practical laboratory work that could power leaps in understanding of how neural circuits work globally.

One experiment his theory suggests might be to test animals trained to alternate between two different types of neural processing (as Rishel and Freedman did with categorization and movement), but with both mixed and pure stimuli for each trial, in other subpopulations of neurons throughout the brain. That way one might see if mixed selectivity neurons are important to processing in other brain areas.

Or one might design an experiment in which where the animals are trained to report on two different categories at once, and figure out if some trials showed the monkeys making mistakes in BOTH categories during the same trial, suggesting that the same “mixed signal” neurons were involved in processing both.

With theory to guide them, experimentalists can design experiments with more powerful insights into how the brain operates. And with data exquisitely captured by lab scientists who can confidently engineer the environments and train and test the animals, theorists can generate the mathematical equations that describe how neuronal coding works.

Clearly, this symbiosis, using researchers’ disparate capabilities, backgrounds, and conceptual frameworks, only really works when everyone is together, communicating regularly and brainstorming fluidly, whether hanging out in each other’s labs or just eating lunch together. They have to learn to speak each other’s languages, become deeply immersed in the concepts that guide either approach.

This is the kind of environment Director John Maunsell, PhD is building at UChicago’s Neuroscience Institute—one unlike most academic neuroscience centers. By bringing everyone into the conversation, he and his faculty are building a shared culture that makes progress in understanding the circuits of the brain a practical possibility. 

Elise Wachspress is a senior communications strategist for University of Chicago Medicine & Biological Sciences Development.