Bumblebee Social Learning: How Insects Solved a Puzzle and Taught Other Bees Without Any Instruction

A bumblebee given a two-step puzzle box, one that required rotating a red tab before a second disc could be pushed aside to reach a sugar reward, had no prior experience with either step. Given enough time alone, most bees failed. Bees that watched a trained demonstrator complete the sequence cracked it themselves. Then other bees watched them, and the solution spread.
The 2023 study, published in PLOS Biology and led by Alice Bridges and Lars Chittka at Queen Mary University of London, presented bumblebees (Bombus terrestris) with this sequential puzzle box. The box had two possible solutions, each requiring distinct actions in order. Researcher-trained demonstrator bees were introduced into colonies that had never seen the box. Naive bees observed. Then they tried.
Single-step puzzle learning in insects had been documented before. Two-step sequential learning, transmitted socially through a colony, had not.No bee was rewarded for demonstrating. No bee was punished for failing. The transmission happened through observation alone. The researchers ran conditions where some colonies had access to a demonstrator and others did not. In demonstrator-absent colonies, the two-step solution almost never appeared spontaneously.
This rules out the simpler explanation: that bees are just individually smart enough to figure it out given time. Some are. But the rate of spread through a colony with a demonstrator was dramatically higher, and the bees that learned from watching tended to use the exact sequence the demonstrator used, even when an alternative route was available. The behaviour was copied, not independently reinvented.

Bumblebees have roughly one million neurons. The human brain has around 86 billion. A brain one-millionth the size encoding a novel two-step procedure from a single observation session, and reproducing it reliably, forces a reconsideration of what cognition actually requires.The researchers tracked which bees had watched which demonstrators, and which naive bees subsequently solved the puzzle. The solution moved in a pattern consistent with social chains: bee A learned from the demonstrator, bee B learned from bee A, bee C from bee B. The researchers described this as cultural transmission, a behaviour spreading through a population not through genetics or individual trial and error, but through social observation.
Culture, in biology, is defined as information that passes between individuals through social learning rather than through genes. By that definition, the bumblebee colonies in this experiment had culture. The word sounds large for an insect. The data supports it.Bumblebees already had a documented list of cognitive abilities most people associate with larger animals. They learn to associate colours with rewards. They use tools: a 2017 study by Olli Loukola and colleagues, published in Science, showed bumblebees learning to move a ball to a target for a reward after watching a demonstrator. They show pessimistic cognitive bias when stressed. They roll wooden balls with no apparent reward, which researchers classify as play.

The 2023 puzzle study adds sequential learning and cultural transmission to that list. What the research collectively suggests is that insect behaviour is constrained by brain size far less than assumed. The bee's small brain is densely connected and specialised. It does not need billions of neurons to encode a procedure. It needs the right architecture.
For foraging insects, social learning is a practical advantage. A bee that can copy a successful foraging technique from a nestmate is more efficient than one that must discover every flower, every entry point, every sequence, by itself. The puzzle box is artificial. The cognitive machinery it reveals is the same machinery bees use in the field.

The puzzle box was never really about the puzzle. It was a controlled way to isolate one question: can a bee learn a procedure by watching, and can that procedure travel through a colony without anyone teaching it? The answer to both is yes. What that tells us is that the gap between insect cognition and the cognition of animals we consider complex is a matter of degree, not of kind, and the degree keeps shrinking every time someone designs a careful enough experiment.