Sniff one smell and lick left. Sniff another and lick right. Now flip the rule depending on a whiff that came a moment earlier. That is the kind of small, ordinary flexibility your brain performs constantly, and it is harder than it sounds. The same stimulus has to trigger opposite actions depending on context, and the switch has to happen fast. A new study in mice pinpoints where a piece of that machinery sits and, more interestingly, how it works.
Jia Shen and colleagues, working across Columbia and Mount Sinai, trained mice on an olfactory task built around exactly this problem. A mouse first received a context odor. Then, after a delay, it received a test odor and had to lick left or right for a reward. The catch: which direction counted as correct depended on the earlier context. The context odors came from one set of smells, the test odors from another, so the animal could not simply memorize fixed pairs. It had to hold the context in mind and use it to interpret whatever came next.
Watching a decision assemble itself
The researchers focused on the anterior lateral motor cortex, or ALM, a region known to sit near the border between sensing and moving. Using two-photon calcium imaging, they recorded hundreds of neurons while the mice worked. Distinct groups of cells lit up for different things: some tracked the context odor, some the test odor, and some the animal's eventual choice. So ALM was not just relaying a motor command. It carried all the ingredients of the decision at once.
To test whether ALM was actually doing the computation rather than passively reflecting it, the team used optogenetics to briefly silence the region at specific moments. Shutting it down during the context and delay periods hurt performance. That timing matters. It suggests ALM is involved in setting up the correct rule before the test odor even arrives, not just in executing the lick afterward.
Contingency neurons and the logic of the switch
The most striking finding concerns how the brain represents rules that mean the same thing. Two different context odors might both instruct the "lick left for odor X" mapping. You might expect the brain to collapse them into a single shared code, an abstract stand-in for the rule. It did not. The two context odors were represented by separate populations of neurons. Their paths only converged later, at the cells that encode the actual choice.
Among those choice cells, the researchers found a subgroup carrying dual information: they were selective for both the context and the upcoming decision. The authors call these "contingency neurons." They seem to be the junction where a sensory cue gets routed to the right motor output. Rather than the cortex building a tidy, reusable symbol for each rule, it appears to rewire its own local connections on the fly, steering input toward the appropriate response.
That distinction is worth sitting with. One popular idea is that flexible behavior depends on the brain abstracting away messy details to find a clean underlying rule. Here the data point elsewhere. The circuit keeps the messy, cue-specific detail and solves the problem downstream, by dynamically reconfiguring which neurons talk to which. Flexibility comes from reconfiguration, not abstraction.
What the experiment can and cannot say
A few limits are worth naming plainly. This is a mouse study using one specifically designed olfactory task, so it is an open question how far the pattern extends to other senses, other brain regions, or more complex rules. Calcium imaging reports neural activity indirectly and on the slow side, which can blur fast dynamics. And silencing a region tells you it is necessary for the behavior without proving it is the sole author of the computation; other areas feed into ALM and could shape what the recordings show. The "contingency neuron" label describes a response property, not a proven circuit mechanism wired end to end.
Still, the appeal of this work is its concreteness. Context-dependent decision-making is a phrase that can float free of biology. Here it lands on identifiable cells, at identifiable moments, doing an identifiable job. If the reconfiguration picture holds up in other settings, it reframes a basic question about cognition: maybe the brain is less a filing system of abstract rules and more a switchboard that rewires itself, cue by cue, to connect what you sense to what you do.
Comments