A test that chimps made famous, aced by an insect

A century ago, psychologist Wolfgang Köhler ran an experiment that became a foundational moment in animal cognition research. He put chimpanzees in a room with a banana hanging out of reach and a scattering of boxes on the floor, and watched as the animals figured out, on their own, that stacking the boxes would get them to the fruit. No trial and error, no training — just a sudden, working solution to a problem the chimp had never encountered before. That kind of behavior, called insight, was long treated as evidence of the kind of intelligence only large-brained animals possess.

A study published June 4, 2026 in the journal Science just complicated that assumption considerably. Researchers at the University of Oulu in Finland, working with colleagues at the University of Helsinki and the University of Turku, built an insect version of Köhler's classic setup — and bumble bees solved it. Untrained, with a brain smaller than a sesame seed, the bees figured out how to move a ball into position and climb onto it to reach a reward they'd never been shown how to access.

The setup: two facts, no instructions

The experiment, led by researcher Akshaye A. Bhambore and described in the paper "Spontaneous problem-solving in bumble bees," gave the test subjects — Bombus terrestris workers — exactly two pieces of prior knowledge and nothing more. First, the bees learned that a blue artificial flower reliably contained a sugar reward. Second, separately, they learned that a nearby ball was something they could physically move. Neither lesson involved combining the two ideas, and neither hinted at what the actual test would require.

Then researchers moved the goalposts, literally. The rewarding blue flower was relocated to the ceiling of a transparent arena, completely out of reach for a bee walking or flying along the ground. To get to it, a bee would need to independently connect two facts it had never been taught to connect: that the ball, which it already knew it could move, could be pushed underneath the suspended flower and used as a platform to climb up and reach the reward. Nobody demonstrated that sequence. Nobody rewarded any intermediate step. The bees that solved it did so entirely on their own initiative.

Ruling out luck, one control experiment at a time

The most convincing part of this study isn't the headline result — it's how aggressively the researchers tried to disprove it before accepting it. Bhambore and the team ran a series of progressively stricter control experiments specifically designed to rule out simpler, less impressive explanations: that bees were solving the puzzle through blind trial and error, that they were simply attracted to visual movement and stumbled onto the ball by accident, or that they were relying on direct visual guidance toward the flower the entire time.

The decisive test hid the flower from view entirely during the trial. If bees were solving the puzzle just because they could see the reward and were homing in on it visually, removing that visual cue should have stopped them cold. It didn't. Bees still successfully positioned the ball and climbed onto it even when the target reward was completely out of sight, according to reporting on the study by EurekAlert. That result is what pushed the researchers toward describing the behavior as genuinely goal-directed rather than a byproduct of simpler sensory guidance — the bees appeared to be working from an internal plan, not just reacting to what they could see in the moment.

What "goal-directed" actually looked like in practice

Bhambore, describing the bees' behavior in comments picked up by Interesting Engineering, noted that successful individuals showed more directed movement patterns than unsuccessful ones — meaning the bees that eventually solved the puzzle didn't wander randomly toward the solution. Their movements looked purposeful in a way that's harder to explain through accidental discovery. That distinction matters enormously for how researchers classify the behavior. Random success dressed up as intelligence has fooled scientists before; directed movement toward an untrained solution is a considerably higher bar to clear.

The researchers were careful, per coverage of the study by Backyard Brains, not to overstate what this means about bee minds relative to human minds. The finding isn't that bees think the way people do, or that they're running some miniature version of human cognition. It's that a specific kind of problem-solving — combining two previously separate pieces of learned information into a novel behavioral solution, without training and without visual guidance — doesn't require the neural architecture of a large-brained vertebrate to occur. What matters, as one researcher put it, is how the system is organized, not how much hardware it has.

Why this reopens a much bigger question about where intelligence starts

For most of the history of animal cognition research, spontaneous insight-based problem-solving sat on a fairly exclusive list: humans, great apes, some corvids, a handful of other large-brained species. Insects, despite decades of research documenting their sophisticated navigation, communication, and social learning abilities, were generally assumed to solve problems through instinct or association rather than anything resembling forward planning.

This study is the first confirmed instance of that kind of insight-like, goal-directed problem-solving in any insect species, according to coverage of the research. That's a narrow, specific claim, but its implications ripple outward. If a brain this small can pull off behavior this cognitively demanding, the line separating "instinct-driven" from "genuinely intelligent" animal behavior looks considerably blurrier than researchers assumed even a few years ago — and it raises an uncomfortable question about how many other small-brained species have been quietly capable of this all along, simply because nobody built the right experiment to catch them doing it.

*This article was researched using publicly available reporting from Science, ScienceDaily, Phys.org, EurekAlert, Interesting Engineering, SciTechDaily, and Backyard Brains coverage of the peer-reviewed study published in Science. It is intended for informational purposes.*