Summary: A new computational model clarifies how individuals with different neural wiring can nevertheless agree on the identity and similarity of odors, provided they share even a small overlap of odor experience.
In a new study, researchers at Columbia University developed a computational model that explains why people can consistently identify and generalize odors even though the neural wiring in the olfactory cortex varies substantially between individuals. Their results show that two brains do not need identical odor histories; as long as there is any small overlap in past odor experience, both brains can correctly associate new, similar smells.
This study was published in Neuron.
Why identical perception emerges from variable neural wiring
The olfactory pathway is complex. When an odorant enters the nasal cavity, specialized olfactory receptor proteins detect chemical features and send signals to the olfactory bulb. In seminal work from the 1990s, Richard Axel and colleagues identified the large family of genes that encode these receptors, a discovery that later earned a Nobel Prize.
From the olfactory bulb, odor information projects to the piriform cortex, a region considered essential for processing and recognizing smells. Because no two odor encounters are exactly the same, the brain must generalize: it must link similar but not identical odor inputs into a common perceptual category (for example, recognizing different coffees as “coffee” or different flowers as “floral”).
Despite this need for reliable classification, previous experiments found two puzzling features of the piriform cortex. First, neural activity there appears disordered and variable; the particular neurons that respond to a given scent differ across individuals and even between the two hemispheres of one brain. Second, the piriform cortex contains many more neurons than seemed necessary to perform simple odor discrimination tasks. Together, these observations raised questions about how consistent odor perception is achieved and why the piriform cortex is so large.
The computational solution: distributed, redundant representations
To address these paradoxes, the Columbia team built a mathematical model and tested it against neural recordings, including data from fruit flies. Their model shows that randomness in connectivity does not preclude consistent perception. Instead, when the piriform cortex samples odor information across a very large population of neurons, the combined activity preserves similarity relationships between odors.
Think of it like crowdsourcing: each neuron contributes a partial, noisy piece of the odor’s identity. Individually those pieces may look random, but when pooled across many neurons they reconstruct a robust similarity structure. This redundancy allows different brains—each with its own unique wiring and experiences—to converge on the same odor categories as long as there is even a modest shared experience of odors.
Crucially, the model demonstrates that tiny amounts of shared odor history are sufficient to align the internal representations across individuals. Familiarity with a scent in one context (for example, the memory of a rose) can serve as a bridge that helps different people agree about the similarity of other odors (such as different types of coffee). Thus, a shared reference point need not be the same exact scent; any overlap in odor experience helps realign representations.
Implications for perception and cortical organization
The model reconciles why the piriform cortex looks both random and oversized: while straightforward tasks like discriminating two odors or assigning them to a category may require far fewer neurons, consistent generalization across varied and novel odor encounters benefits from the full, large population of piriform neurons. That expansive population provides the redundancy needed to preserve correlations and similarity relationships between odors despite variability in specific neural responses.
By explaining how generalization emerges from distributed, seemingly random activity, this work reveals an elegant principle underlying olfactory perception: consistency does not require identical circuitry or identical experiences—just a small shared history and a large, distributed representation.
Funding and disclosures
Funding: Supported by the Gatsby Charitable Foundation, the National Science Foundation NeuroNex Award (DBI-1707398), the Simons Collaboration on the Global Brain, and the Howard Hughes Medical Institute.
The authors reported no financial or other conflicts of interest.
Publication and research details
Source: Zuckerman Institute, Columbia University. Published in Neuron. Original research: “Odor Perception on the Two Sides of the Brain: Consistency Despite Randomness” by Evan S. Schaffer, Dan D. Stettler, Daniel Kato, Gloria B. Choi, Richard Axel, and L.F. Abbott. DOI: 10.1016/j.neuron.2018.04.004.
Key points:
- Random connectivity from the olfactory bulb to piriform cortex still preserves odor correlations when many neurons are involved.
- Distributed representations support consistent odor quality judgments across different individuals.
- Robust generalization across similar odors may depend on the full complement of piriform neurons, even if discrimination requires fewer cells.