Researchers at Rockefeller University found that 26 human volunteers could reliably distinguish between complex odor mixtures that shared most of their ingredients, a result the team extrapolated to a conservative lower bound exceeding one trillion discriminable olfactory stimuli. That figure, published in Science in 2014, shattered the long-standing textbook claim that people can detect roughly 10,000 distinct smells. But a subsequent critique argued the trillion-odor estimate rests on fragile assumptions, and no independent replication has settled the dispute.
A trillion-scent claim and why it still divides smell science
For decades, introductory biology courses taught that the human nose could pick out about 10,000 odors. The number was never tested rigorously; it traced back to rough estimates from the 1920s and was repeated without serious challenge. When neuroscientists Andreas Keller and Leslie Vosshall at Rockefeller University designed a psychophysics experiment to actually measure discrimination ability, the answer they got was orders of magnitude larger.
Their study asked 26 subjects to sniff vials containing mixtures of 10, 20, or 30 odor components drawn from a library of 128 distinct odorant molecules. Each trial presented three vials: two identical and one different. Subjects had to identify the odd one out. Even when two mixtures overlapped by more than half their components, most participants could still tell them apart. The team then used the discrimination thresholds they measured to calculate how many unique mixtures a person could theoretically distinguish across the full combinatorial space of that 128-molecule palette. The result: at least one trillion, presented as a conservative lower bound.
A contemporary news report emphasized how dramatically this estimate expanded the presumed range of human olfaction. The finding reframed smell as a far more powerful sense than vision or hearing in terms of raw discriminatory capacity, since the eye distinguishes a few million colors and the ear about 340,000 tones. If correct, the work implies that everyday experience samples only a tiny fraction of the potential odor space our noses could, in principle, resolve.
How 26 noses and 128 molecules produced a trillion-scent estimate
The experimental design hinged on a simple but clever structure. Keller and Vosshall selected 128 odorant molecules that spanned a wide range of chemical classes, from floral and fruity to sulfurous and woody. They then created mixtures containing either 10, 20, or 30 of those molecules in equal proportions. In each trial, two of the three vials were identical, and the third differed by a controlled number of swapped components. By varying the percentage of overlap between the “different” mixture and its matched pair, the researchers could pinpoint the threshold at which subjects lost the ability to discriminate.
Subjects performed hundreds of trials in a forced-choice format. The data showed that discrimination dropped off as mixtures became more similar, but even at high overlap levels, performance stayed above chance for many comparisons. From those measured thresholds, the team extrapolated across the full space of possible 128-component combinations to arrive at the trillion figure. They described the estimate as conservative because it used a simple model of mixture similarity and because the real world contains far more than 128 odorant molecules.
That extrapolation step is where the controversy begins. The combinatorial space of possible mixtures is astronomically large, and only a minuscule fraction can ever be tested directly. To bridge that gap, the researchers assumed that the discrimination performance observed in their sample of mixtures would generalize across the entire space. They also adopted a particular statistical criterion for deciding when two mixtures should count as perceptually distinct. Small tweaks to those assumptions can produce very different totals, a sensitivity that critics later seized upon.
Beyond the headline number, the work fed into a broader effort to understand how roughly 400 types of human odorant receptors can encode a vast variety of percepts. Reviews of olfactory coding, such as one in a neuroscience journal, have highlighted how each odorant can activate multiple receptors and each receptor can respond to multiple odorants, creating a high-dimensional pattern space. The trillion-scent claim, if borne out, would underscore how efficiently this combinatorial code can carve up the space of possible molecular mixtures.
The Gerkin and Castro critique and unresolved questions
The trillion-scent estimate did not go unchallenged. In 2015, researchers Richard Gerkin and Jason Castro published a detailed methodological critique in eLife, arguing that the trillion-odors estimate is not statistically stable. Their analysis showed that small changes in the discrimination criterion, the number of subjects tested, or the statistical model used for extrapolation could shift the final number by many orders of magnitude in either direction. With only 26 participants and a specific set of mixture parameters, the original estimate was highly sensitive to design choices that could have gone differently.
Gerkin and Castro did not claim the true number is smaller than a trillion. They argued that the data simply cannot support any single number with confidence. In their view, the behavioral results demonstrate that humans can distinguish more odors than the old 10,000-odor figure allowed, but they do not justify pinning that capacity to a specific order of magnitude. The title of their paper made the point clear: the number of olfactory stimuli that humans can discriminate is still unknown.
The critique also stressed that the extrapolation did not incorporate individual variability. Some participants in the Rockefeller study performed much better than others, and the analysis effectively averaged across them. If real-world populations span a wide range of olfactory acuity due to genetics, age, or experience, then a single capacity estimate may obscure meaningful differences. A more robust approach, the critics suggested, would model individual performance and then ask how population-level variation shapes the overall space of discriminable odors.
Several gaps in the evidence remain unaddressed. No independent lab has replicated the experiment using the same 128-odorant library with a larger or more demographically varied group of subjects. The raw trial-by-trial data and exact mixture compositions from the original 26 participants have not been widely available for reanalysis, limiting outside efforts to test alternative models. And no study has directly tied behavioral discrimination thresholds to physiological recordings of receptor activation patterns in humans, which would help connect perceptual limits to underlying neural mechanisms.
The debate has spilled into broader discussions about how to quantify sensory capacities at all. One commentary, accessible via a publisher portal, raised the concern that headline numbers can overshadow the subtler insights that come from mapping how perception changes with stimulus similarity. In this view, arguing over whether the true figure is billions, trillions, or more may matter less than understanding the structure of odor space: which mixtures cluster together, which dimensions are most salient, and how experience reshapes those boundaries.
What the trillion-odor debate means for future research
Despite the controversy, most researchers agree on two points. First, the old 10,000-odor claim is untenable in light of modern psychophysics. Humans can likely distinguish far more than that, especially when complex mixtures are considered. Second, current data are insufficient to anchor a precise upper bound. The trillion-odor estimate is best seen as a provocative hypothesis that spurred new work rather than a settled fact.
Future experiments could narrow the uncertainty by testing larger, more diverse cohorts with standardized mixture libraries and open data sharing. Combining behavioral tasks with brain imaging or receptor-level assays might reveal how different mixtures that smell distinct map onto patterns of neural activity. Computational models of olfactory coding, informed by such data, could then generate predictions about discrimination limits that are testable without sampling the entire combinatorial space.
For now, the human sense of smell occupies an ambiguous position: clearly more powerful and flexible than once assumed, yet still hard to quantify in a single number. Whether the final tally lands in the billions, trillions, or beyond, the ongoing debate has already reshaped how scientists think about odor perception, pushing the field toward more rigorous, transparent, and mechanistically grounded approaches to one of our most ancient senses.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
