A species of bumblebee with a brain smaller than a sesame seed has solved a physical problem that researchers never demonstrated for it, repositioning a ball to use as a platform and reach a sugar reward hidden overhead. The peer-reviewed finding, published in Science by a team of six researchers, places Bombus terrestris alongside primates and corvids in the short list of animals documented performing spontaneous object manipulation to access food. The result directly challenges the long-held assumption that flexible problem-solving of this kind requires a large brain or extensive prior training.
Why spontaneous insect problem-solving upends old assumptions
The experimental setup draws on a test first used by psychologist Wolfgang Köhler more than a century ago: place a reward out of reach and provide an object that can be moved into position to access it. In Köhler’s original version, chimpanzees stacked boxes beneath a hanging banana. The bumblebee version, described in the Science report, replaced the banana with a sugar-dispensing artificial flower mounted on the ceiling of a small arena and gave the bees a movable ball instead of a box. Bombus terrestris workers that had never been shown the solution repositioned the ball beneath the flower and climbed on top of it to feed.
That outcome is striking because tool-use research has historically treated brain size as a rough predictor of cognitive flexibility. Great apes, New Caledonian crows, and a handful of other vertebrates have passed versions of the box-and-banana test. Insects were not expected to join that group. The bumblebee result suggests that the neural architecture needed for this kind of spatial reasoning can be packed into far fewer neurons than scientists assumed, and it raises hard questions about how many other invertebrate species might share the capacity if tested under the right conditions.
One plausible explanation is that bees are not deploying a general-purpose tool-use module at all. Instead, they may rely on a rapid spatial-reasoning shortcut tuned to vertical displacement, the kind of calculation a foraging insect already performs when assessing flower height and landing angles. If that hypothesis holds, follow-up experiments varying ball mass and ceiling height should produce predictable shifts in success rates. That test has not yet been reported, but the conceptual link to everyday foraging decisions offers a starting point for future work.
Controlled experiments ruled out simple trial and error
The research team, whose author list on PubMed includes Bhambore, Akmese, Hakkinen, Jussila, Kantola, and Loukola, built several control conditions to rule out the possibility that bees stumbled onto the answer by accident. In one condition, barriers hid the flower from the bees’ line of sight, forcing them to recall its location from a prior exploration phase rather than simply walking toward a visible target. In another, bees were given an exploration period before facing the choice, separating the discovery of the flower’s position from the act of solving the problem. A third set of controls compared bees that had partial or no exposure to the individual task components, such as the ball or the flower, against bees that had encountered both.
The most telling result came from the behavior of successful bees. Rather than cycling through wrong options, many went directly to the correct location and began moving the ball without hesitation. That pattern is difficult to explain through random exploration. It points instead to some form of internal representation, a mental map of where the reward sits and what physical steps are needed to reach it, assembled during the exploration phase and executed in a single directed sequence.
The study’s framing as an analogue of Köhler’s classic test is deliberate. By replicating the logical structure of the box-and-banana problem while stripping away the vertebrate context, the authors created a direct comparison point. The bees did not need to observe a demonstrator. They did not need repeated failed attempts. They arrived at a functional solution on their own, under controlled laboratory conditions, with their behavior recorded and compared against multiple baselines.
This design matters because insects are often assumed to rely heavily on incremental trial and error or simple associative learning tied to immediate sensory cues. In the bumblebee arena, however, the key cue-the flower-could be concealed during the crucial phase, leaving only the remembered location and the presence of a movable object. When bees still managed to navigate to the correct spot, reposition the ball, and climb onto it, they appeared to be integrating stored spatial information with an understanding that elevating themselves would bring them into contact with the reward.
Alternative explanations cannot be dismissed outright. It remains possible that some bees relied on a history of interactions with elevated flowers in other contexts, or that subtle arena features guided their movements more than observers realized. Yet the convergence of performance across different control conditions, and the rapidity with which some individuals executed the correct sequence, argue against a purely incremental search strategy. The behavior looks organized rather than haphazard.
Open questions about insect cognition after the Science study
Several gaps remain. The publicly available summaries of the study do not include raw trial-by-trial data, exact sample sizes per condition, or the statistical outputs that would let outside researchers evaluate effect sizes. Those details sit behind the journal’s paywall. Without them, independent assessment of how robust the success rates were across conditions is limited to the qualitative descriptions released alongside the paper.
Direct quotes from the authors have not appeared in the institutional press materials distributed so far. That absence makes it harder to pin down how the team interprets its own findings, particularly on the question of mechanism. Did the bees solve the problem through spatial reasoning, associative learning, or something else entirely? The published article presumably addresses this, but the secondary descriptions do not settle it, leaving room for competing interpretations among outside commentators.
One line of inquiry concerns individual variation. If only a small fraction of bees solved the task, that would suggest a specialized subset of particularly exploratory or cognitively flexible workers. If many solved it, the capacity might be widespread within colonies. Without detailed numbers, it is unclear whether Bombus terrestris should be thought of as a species that generally handles object-manipulation problems, or as a society in which rare innovators occasionally emerge.
Another open question involves learning over time. The current descriptions focus on bees that solved the problem without demonstration, but do not clarify how performance changed with repeated exposure. If a naïve bee watched a successful peer move the ball and climb onto it, would it copy the sequence on its next attempt? Demonstrator-based learning has been shown in other bee contexts, such as route following and color preferences, and extending that work to physical problem-solving would help distinguish between individual insight and social transmission.
A broader question hangs over the field: if bumblebees can pass a test designed for apes, what does that say about the dozens of other social insect species that have never been tested with similar paradigms? Honeybees, stingless bees, and several ant genera show complex foraging behaviors that could, in principle, involve comparable spatial reasoning. The Bombus terrestris result is a single data point, but it is a data point that rewrites expectations about what a small brain can accomplish without instruction. Researchers studying insect navigation, division of labor, and communication will now have to consider whether comparable object-based problem-solving might be latent in their own study species.
The implications extend beyond entomology. If sophisticated problem-solving can be implemented in such compact neural hardware, it strengthens the case that cognition depends as much on circuit organization and ecological pressures as on sheer neuron count. That perspective could influence how neuroscientists think about efficiency in artificial systems and how ethicists weigh the cognitive lives of small animals. Future work that exposes more of the underlying data, systematically varies task parameters, and broadens the range of tested species will determine whether the bumblebee’s ball-and-flower performance is an outlier or an early glimpse of a much wider invertebrate intelligence.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
