Bumblebees Match Chimps in Classic Problem-Solving Test | New Study Rewrites Invertebrate Intelligence

A study published Thursday in the journal Science presents direct evidence that buff-tailed bumblebees (Bombus terrestris) can spontaneously use physical objects as tools to reach food they cannot otherwise access. The research, led by behavioral ecologist Dr. Olli Loukola at Finland's University of Oulu, is the first peer-reviewed demonstration of goal-oriented, unprompted tool manipulation in any invertebrate species.

The finding places bumblebees alongside chimpanzees in the narrow category of animals that pass the classic box-and-banana problem, a benchmark of spontaneous insight first used in vertebrate cognition research in the early 20th century. What makes the result particularly striking is the scale difference: a chimpanzee applies roughly 2.8 billion neurons to this class of puzzle. The bumblebees solved it with approximately 1 million.

The Experiment | A Miniature Box-and-Banana Test for Insects

The original box-and-banana paradigm, developed by psychologist Wolfgang Köhler in the 1910s, requires an animal to understand that an out-of-reach reward can be accessed by repositioning a separate object. Köhler's chimpanzees stacked wooden crates to climb toward suspended bananas. The cognitive requirement is the same regardless of the physical scale: recognize that a neutral object in the environment can become a functional tool for a specific purpose, without being shown how.

The University of Oulu team built a miniaturized version. Young bumblebees were trained in two separate and unconnected tasks: associating a blue circular marker with a sugary food reward, and learning that small polystyrene balls in their enclosure could be physically moved. Neither training session involved combining the two concepts.

In the test phase, the blue target was placed on the ceiling of a shallow plexiglass arena. The ceiling was too high for the bees to walk up to reach it, and the enclosure was sized specifically to prevent hovering. The polystyrene balls were available on the arena floor. No bee had ever been shown or rewarded for rolling a ball under the target.

Seventy-five percent of the bees solved the problem without prompting. They rolled a ball directly below the blue target, climbed on top of it, and collected the reward. The behavior was not random exploration. The bees oriented the ball movement toward the target location, a directional choice that implies a connection between the tool and the goal.

Ruling Out Trial and Error | The Barrier Test

Spontaneous-looking behavior in animal cognition studies often turns out to be incremental trial-and-error rather than insight. To test whether the bees were genuinely forming an internal model of the problem or simply reacting to what they could see in real time, the researchers introduced physical barriers that blocked the bees' line of sight to the ceiling target while they were handling the ball.

To succeed under these conditions, a bee had to hold two pieces of information in working memory simultaneously: the remembered location of the target it could no longer see, and the current position of the ball it was manipulating. It then had to navigate around the barrier and place the ball in the correct position relative to a goal it was not looking at.

The majority of bees succeeded. The result is difficult to explain as accidental or purely reactive behavior. It supports the interpretation that the bees were operating with something resembling a goal-oriented internal representation, planning for a state of the environment they could not currently perceive.

"Most people think insects are reflex-based machines," said Dr. Loukola. "Our findings show that miniature brains can generate flexible solutions to novel problems in ways we are only beginning to understand."

One Million Neurons | What This Means for Theories of Intelligence

The vertebrate-centric model of complex cognition assumes that insight, planning, and flexible problem-solving require large neural architectures. The neurological basis for this assumption has always been correlational rather than mechanistic: big-brained animals tend to pass insight tests, small-brained animals tend not to. MAVEN's removal from the evidence base does not overturn the correlation, but the bumblebee data puts a hard constraint on the minimum neural hardware required.

A brain of 1 million neurons is not a simple reflex loop. Bumblebees have been shown in previous research to experience rudimentary emotional states, to transmit learned behaviors to hive-mates through observation, and to make probabilistic foraging decisions that account for risk. The new study adds spontaneous object manipulation to that list, a category of behavior that researchers had assumed required a vertebrate nervous system.

"Bees are a model of how much intelligence you can squeeze into a miniature nervous system," said Lars Chittka, a behavioral ecologist at Queen Mary University of London who was not involved in the study. "This is the clearest demonstration yet of some kind of comprehension of what's at stake."

The implications extend well beyond taxonomy. If spontaneous tool use can emerge from a 1-million-neuron system evolved for foraging efficiency, it suggests that complex cognitive functions may be far more substrate-independent than current theories predict. That question connects directly to ongoing debates in neuroscience, artificial intelligence research, and the philosophy of mind about what physical conditions are actually necessary for goal-directed behavior.

Implications for Conservation and Animal Welfare

Bumblebee populations across Europe and North America have declined significantly over the past three decades due to habitat loss, pesticide exposure, and climate-driven shifts in flowering plant distribution. The legal and regulatory frameworks that govern pesticide approval and agricultural land use treat insects categorically differently from vertebrates, in part because of the longstanding assumption that insects lack morally relevant cognitive experiences.

Studies like this one complicate that assumption. If bumblebees form internal representations, pursue goals, and adapt behavior to novel circumstances, the question of whether they experience something when those cognitive processes are disrupted becomes harder to dismiss. Researchers in the field have begun arguing that conservation policy for pollinators should account not just for ecological function but for cognitive complexity.

That argument has practical stakes. Bombus terrestris is the primary commercial pollinator for greenhouse crops across Europe. Its decline directly affects food production. The same cognitive flexibility documented in this study, the ability to solve novel foraging problems, is likely part of what makes it effective in the agricultural environments where it has become economically essential. Understanding the neuroscience of that flexibility is not just academically interesting. It is relevant to decisions about which agricultural practices are compatible with maintaining viable pollinator populations.

Our recent coverage of the end of NASA's MAVEN mission, the June 2026 cannibal CME aurora event, the ozone layer recovery data for 2026, and the pigeon magnetic navigation discovery reflects a broader moment in which the boundaries of what animal nervous systems can do are being revised across multiple research fronts simultaneously. The bumblebee study published today in Science sits squarely in that pattern.

Frequently Asked Questions

What did the bumblebee study in Science find?

Researchers at Finland's University of Oulu found that buff-tailed bumblebees spontaneously roll objects underneath an out-of-reach food target and climb on top of them to access the reward. The bees were never trained or shown this solution. It is the first confirmed instance of spontaneous tool use in any invertebrate species, published in the journal Science on June 4, 2026.

How does bumblebee intelligence compare to chimpanzees?

Chimpanzees use roughly 2.8 billion neurons to solve the classic box-and-banana insight problem. The bumblebees in this study solved a structurally identical problem with approximately 1 million neurons. The behavioral outcome was the same: recognize that a neutral object can be repositioned to reach an otherwise inaccessible reward. The neural efficiency difference is roughly 2,800 to one.

Who led the bumblebee research?

The study was led by Dr. Olli Loukola, a behavioral ecologist at the University of Oulu in Finland. Lars Chittka of Queen Mary University of London, a leading researcher in bee cognition, commented on the findings but was not part of the research team.

Can bees really plan ahead?

The barrier test in the study showed that bees could manipulate a tool to reach a target they could not currently see, holding the goal location in working memory while navigating an obstacle. This is consistent with goal-oriented planning, though researchers are careful to note it does not necessarily imply the same subjective experience that the term implies in humans.

Why does insect intelligence matter for conservation?

Regulatory frameworks for pesticide approval and agricultural land use treat insects as lacking morally relevant cognitive complexity. Evidence of spontaneous tool use and goal-directed behavior in bumblebees challenges the neurological assumptions behind that treatment. Bumblebees are also the primary commercial pollinator for European greenhouse crops, making their cognitive resilience economically significant alongside its scientific interest.