Honeybees can learn to do basic addition and subtraction
Despite possessing a brain with fewer than one million neurons, honeybees can learn basic arithmetic. In lab experiments, researchers trained bees using color-coded mazes: blue shapes signaled that bees needed to add one item, while yellow shapes meant subtract one item. By choosing the correct visual option, bees consistently solved basic math problems to receive a sweet reward, proving that complex numerical reasoning does not require a large vertebrate brain.
Rethinking the Limits of Insect Intelligence
For decades, the standard assumption in cognitive science was that abstract mathematical operations required large, layered brains. Processing abstract relationships, maintaining values in short-term working memory, and applying dynamic rules like addition and subtraction were considered exclusive to vertebrates with complex cerebral cortices, such as primates, marine mammals, and certain birds like parrots. Honeybees, possessing fewer than one million neurons compared to the human brain's roughly eighty-six billion, were thought to rely almost entirely on hardwired instincts, simple associative cues, and basic sensory-motor patterns.
However, researchers studying insect behavior have increasingly found that the compact neural architecture of bees can perform remarkable cognitive feats. Far from being simple biological automata, honeybees display flexible learning, rule abstraction, and complex decision-making. Their survival depends on navigating miles across diverse landscapes, evaluating nectar quality, and remembering the locations and visual characteristics of changing floral resources. This behavioral plasticity prompted researchers to test whether a bee's compact nervous system could manage symbolic numerical reasoning.
Inside the Arithmetic Y-Maze Experiment
To investigate whether honeybees could perform arithmetic, scientists constructed a controlled experimental apparatus known as a Y-maze. Individual free-flying bees were trained to enter the maze through an opening where they first encountered a sample visual stimulus. This stimulus displayed a set of geometric shapes, such as squares, triangles, or circles, rendered in a specific color. The color of the shapes served as a symbolic instruction: blue signaled that the bee needed to perform addition by one (+1), while yellow signaled that the bee needed to perform subtraction by one (-1).
After observing the initial sample display, the bee flew through an opening into a decision chamber featuring two separate arms. Each arm presented a new visual card with a different quantity of shapes. If the initial display showed two blue shapes, the correct choice in the decision chamber was a card showing three shapes. If the initial display showed three yellow shapes, the correct choice was a card showing two shapes. The alternative arm displayed an incorrect number, such as an identical count to the sample or a shift in the wrong mathematical direction.
Rule Learning, Working Memory, and Feedback
To motivate the bees and evaluate their ability to learn, researchers used a differential conditioning protocol. Choosing the mathematically correct display rewarded the bee with a drop of sweet sucrose solution, whereas choosing the incorrect display yielded a bitter-tasting quinine solution or plain water. Over repeated training trials, the positions of the correct answers were randomized between the left and right arms to ensure the bees were not simply developing a directional bias.
Crucially, the experimenters controlled for non-numerical visual cues. To prevent the bees from making choices based on total surface area, brightness, color saturation, or shape configuration, the testing displays varied the sizes and arrangements of the geometric shapes. A correct answer with fewer shapes might have a larger overall surface area than an incorrect option with more shapes. The bees consistently chose the mathematically correct number of items regardless of spatial layout, proving they were responding to the numerical quantity rather than superficial visual patterns.
The test demonstrated that the bees successfully maintained the initial quantity in working memory, processed the color as an abstract mathematical operator, applied the operation to modify the remembered value by one, and matched that new value to the visual options in the decision chamber. Across multiple trials with novel shapes they had not encountered during training, the bees solved both addition and subtraction problems at levels significantly above chance.
A Broader Foundation of Numerical Cognition
The ability of honeybees to add and subtract did not emerge in a vacuum; it builds upon a sophisticated suite of basic numerical skills identified across different behavioral studies. Prior research showed that foraging bees can count visual landmarks along a flight path to gauge distance, helping them navigate to reliable food sources and return accurately to the hive. When landmarks are shifted or rearranged in experimental setups, bees adjust their flight expectations accordingly.
Furthermore, honeybees have demonstrated an understanding of relative quantity, successfully learning concepts such as 'greater than' and 'less than.' In separate experiments, bees were trained to consistently choose the smaller quantity among visual displays containing varying numbers of elements. When presented with an entirely empty display containing zero items alongside displays containing one or more shapes, the bees treated the empty set as a numerical value sitting at the lowest end of the continuum, effectively understanding the abstract mathematical concept of zero.
Neural Mechanisms and Mushroom Bodies
The ability of honeybees to carry out such operations highlights the functional efficiency of insect brain anatomy. High-level associative learning in bees takes place predominantly in paired structures known as mushroom bodies, located within the insect brain. These dense neuropils receive pre-processed sensory information from the visual and olfactory lobes, integrating signals through complex networks of neurons known as Kenyon cells.
Because insect neurons can form intricate local feedback loops and maintain multiple functional connections per cell, a circuit containing relatively few neurons can accomplish computational tasks that were once thought to require extensive mammalian neocortex layers. The bee's nervous system optimizes its wiring by using specialized modular circuits that process symbolic rules and sensory memory in parallel, maximizing computational output while minimizing metabolic and anatomical costs.
Laboratory Capabilities Versus Natural Ecology
While honeybees reliably demonstrate numerical reasoning in controlled laboratory mazes, there is no direct evidence that wild bees solve arithmetic equations during their everyday foraging routines. In nature, bees rely on a combination of visual scanning, floral odors, spatial mapping, and communication methods like the waggle dance, which encodes distance and solar direction to recruit nestmates to rich floral patches.
Instead, the capacity for basic math reflects a broader underlying cognitive plasticity. Natural selection has endowed honeybees with general-purpose associative learning systems that allow them to adapt to unpredictable floral patterns, seasonal changes, and navigating complex outdoor environments. When placed in human-designed laboratory tasks that present arithmetic rules, the bees can co-opt these generalized learning circuits to solve problems that they never encounter in the wild. This reveals that complex numerical reasoning is an emergent property of compact, highly connected neural networks.
Key takeaways
•Honeybees can learn to use color cues as operational symbols, adding or subtracting one item based on blue or yellow stimuli.
•The bees solve arithmetic problems by combining working memory, rule application, and quantity discrimination, while ignoring confounding factors like shape area.
•This mathematical capacity builds on existing numerical competencies in bees, including landmark counting, relative quantity judgment, and understanding zero as an empty set.
•Complex numerical reasoning does not require a large vertebrate brain, showing that compact insect neural structures like mushroom bodies can support sophisticated cognitive processing.