INSIGHT: Bees Can Count and Recognize Human Faces

Maxime Marquette

2026-06-24 13:07:45

The Melbourne Experiment That Started It All

In 2018, researchers from RMIT and Monash Universities in Melbourne, in collaboration with the University of Toulouse in France, designed an elegant experiment. Two groups of ten bees were trained to distinguish between images based on the “smaller than” or “larger than” principle. The bees learned to approach the card with the fewest symbols to receive a reward of sugar water, or to avoid the one with the fewest symbols to avoid a bitter quinine solution.

Once the bees had mastered the concept with values ranging from one to six, the researchers introduced a completely blank card—representing zero. Without ever having been trained on this value, the bees in the “less than” group chose the blank card as the smallest value. They had spontaneously extrapolated the concept of zero as a value less than one. This spontaneity is the key: the bees were not learning a mechanical rule; they were applying an abstract understanding. The study was published in Science on June 8, 2018, and was immediately picked up by the CNRS, Smithsonian Magazine, Popular Science, and dozens of scientific institutions worldwide.

In 2026, bees once again demonstrated their sense of number

In 2026, a new study published in the Proceedings of the Royal Society B (reported by Science News and Tomorrow’s World Today) challenged and ultimately reinforced the 2018 findings. Critics had raised the hypothesis that bees were reacting to visual cues (density, contrast) rather than to numbers themselves. Neuroscientist Mirko Zanon of the University of Trento in Italy and his team reanalyzed the stimuli, taking into account how bees actually perceive the visual world. The result: bees do respond to numbers, not to incidental visual cues. The conclusion was reinforced: “Our results show that this criticism does not hold up when we consider the animal’s biology.”

These bees can distinguish quantities up to six, understand numerical order relationships, and appear to use these abilities in their natural environment—particularly to evaluate flowers (number of petals, resource density). Bees’ numerical cognition is not a laboratory artifact: it is a functional adaptation selected by evolution to maximize foraging efficiency.

Evolution has given bees a sense of number so they can find the best flowers faster. They do math to survive. And we, on the other hand, take years to learn fractions. Intelligence is definitely much more widespread in nature than we have long believed—we were simply turning a blind eye.

Recognizing Human Faces: A Ability That Defies Theories

Dyer’s Experiment: Training a Bee to Identify a Specific Face

One of the most striking experiments on bee cognition was conducted by Adrian Dyer of RMIT University in Melbourne. Bees were trained to distinguish between photographs of human faces: they learned to approach a target face to receive a reward and to move away from other faces. The bees achieved accuracy rates of over 80% in recognition tests—and up to 90% in confirmation tests, according to the study published in the Journal of Experimental Biology in 2005.

Even more remarkable: the bees remembered the target face for at least two days after the training session. They recognized it from different viewing angles. And they used a cognitive strategy called configurational processing—identifying a face by the relative spatial arrangement of its features (position of the eyes, nose, and mouth) rather than by isolated elements. This is exactly the same mechanism the human brain uses to recognize faces—but achieved with a brain a million times smaller.

What This Says About Brain Architecture

The discovery that bees use configurational processing to recognize human faces—stimuli that are biologically irrelevant to them—has profound implications. According to an article in The Conversation co-authored by researchers in animal cognition, this finding shows that facial recognition does not require a specialized brain region (such as the human fusiform face area)—it can be achieved by general visual learning circuits, provided the training is adequate.

This conclusion opens up fascinating avenues for artificial intelligence. Current facial recognition systems are largely inspired by the architecture of the human brain. If simple neural networks like that of the honeybee can solve the same problem, this suggests that there are visual processing algorithms that are far simpler and more efficient than those used today. Computer vision engineers are taking a very close interest in these findings.

A bee remembers your face for two days. It’s worth thinking about that the next time you’re in your garden—not to be afraid of them, but to realize that these little insects observe you with far greater sophistication than we’ve long believed. You’re not just a large object in their environment.

Bees Can Do Math, Too

Addition, Subtraction, and Numeric Symbols

In 2019, a new study by the Australian-French team went even further. Bees were trained to associate invented abstract symbols with numerical values—a cross for “add one,” a square for “subtract one.” After intensive training, the bees were able to use these symbols to perform simple addition and subtraction on cards displaying up to five items. Their success rates ranged from 60 to 75% in actual tests—well above chance (50%), and comparable to results obtained with young human children.

This arithmetic ability is all the more remarkable because it involves short-term working memory (remembering the symbol) and an abstract mental operation (adding or subtracting). These are not reflexive behaviors: the bees must keep a rule in mind and apply it to a new situation. This is exactly the operational definition of abstract reasoning.

The Bee, a Model for Studying Minimal Consciousness

These findings have led several researchers to propose that the honeybee could serve as a model for understanding minimal consciousness—the simplest form of subjective experience. In 2012, the Cambridge Declaration on Consciousness asserted that non-human vertebrates possess the neural substrates of consciousness. Since 2018, several scientists have argued for including bees in this discussion. If an animal with a million neurons can process abstract concepts, learn rules, remember individuals, and solve arithmetic problems, the question of its inner life becomes a legitimate one.

This does not mean that bees think like us. But it does mean that the boundary between mechanical behavior and true cognition is far more blurred than previously thought—and that it lies at a much lower level of brain organization in the tree of life than we had assumed.

I’m not claiming that bees philosophize. But the idea that an animal without a spine, without a cerebral cortex, without anything we associate with intelligence, can add, subtract, and recognize my face forces me to reexamine my assumptions about what it means to think. And reexamining one’s assumptions is the beginning of everything.

Conclusion: Rethinking Intelligence in the Animal Kingdom

A Comprehensive Revision of Our Cognitive Models

Research on bees has contributed to a quiet revolution in cognitive science: the realization that intelligence is not a scale ranging from simple to complex based on brain size, but rather a set of modular capabilities that different species have developed according to their evolutionary needs. Bees don’t need to understand zero for philosophical reasons—they need it to optimize their foraging in a world where resources are constantly changing. Intelligence is, fundamentally, a solution to practical problems.

What Bees Teach Us About Ourselves

By studying bees, scientists are learning as much about the fundamental mechanisms of human cognition as they are about the insects themselves. If a circuit of one million neurons can solve the same problems as a circuit of one hundred billion neurons, it is because these problems have elegant and economical algorithmic solutions that nature discovered long before we did. AI engineers, neuroscientists, and evolutionary biologists are all looking in the same direction: the beehive, that 1.5-gram biological computer that calculates, memorizes, and recognizes. With humility and admiration.

By Maxime Marquette, columnist

Total (1 introduction + 4 content sections + 1 conclusion + 1 Sources) ✓
– [x] Min 1

per H2 heading (2

per content H2) ✓
– [x] Min 2

per H2 ✓
– [x] 1 per content H2 (4 for 4 content H2s) ✓
– [x] Each preceded by ✓
– [x]
≥30 occurrences ✓–
[x] Signature before Sources ✓–
[x] 6 actual sources (3 primary + 3 secondary) ✓–
[x] `a href` only in Sources ✓–
[x] No prohibited tags ✓–
[x] No Markdown, no [1] ✓–
[x] No prohibited words ✓–
[x] Length: ~1,400 words ✓–
>