Spelling Bees? Maybe Not, but They Can Do Math

Analysis by Dr. Karen Shaw Becker

Story at-a-glance -

  • Researchers from Australia and France found that honeybees can learn how to add and subtract and indicate the smaller of two objects
  • Bees may only understand numbers at a basic level, but enough to learn rules and concepts for problem solving, and they can be trained to count and use number discrimination
  • The fact that bees can perceive math concepts helps humans understand the relationship between brain size and brain power, and implies the possibility of future development of artificial intelligence (AI) for rapid learning
  • Honeybees return repeatedly to places they’ve learned are good sources for food, which is how the researchers lured them to visit a specially designed testing apparatus — a Y-shaped maze
  • Scientists have debated whether animals have sufficient brain power to learn complex number skills, as it takes a higher level of mental control to parse numbers, especially the concept of zero

If you thought the term “pea brain” meant the capacity for cognitive thinking might not be up to par because a pea is so small, consider this: If someone was called a “bee brain” — obviously smaller than a pea — it would actually be a compliment, because a new study suggests that you don't need a large brain to solve tasks. 

Here’s why: Scientists recently revealed that bees may only understand numbers at a basic level, but enough for problem solving. In fact, they say honeybees are a model for insect cognition, perception and vision.1 They can learn complex rules and concepts2 and be trained to count and use number discrimination. 

The study, published in Science Advances,3 reflects a collaboration between colleagues from universities in Australia and France. Co-author Adrian Dyer, Ph.D. from Australia’s RMIT University says honeybees can learn how to add and subtract and indicate the smaller of two objects. Dyer explains:

“You need to be able to hold the rules around adding and subtracting in your long-term memory, while mentally manipulating a set of given numbers in your short-term memory. On top of this, our bees also used their short-term memories to solve arithmetic problems, as they learned to recognize plus or minus as abstract concepts rather than being given visual aids.

Our findings suggest that advanced numerical cognition may be found much more widely in nature among [nonhuman] animals than previously suspected. If (math) doesn’t require a massive brain, there might also be new ways for us to incorporate interactions of both long-term rules and working memory into designs to improve rapid AI [artificial intelligence] learning of new problems.”4

According to the featured video, math has been vital to human society for millennia, but the fact that bees have their own math language and can perceive math concepts helps humans understand the relationship between brain size and brain power. Dyer says it implies the possibility for future development of AI, especially in the realm of rapid learning.

Math 101 for Bees: How the Honeybees Were Trained

Needless to say, arithmetic requires a sophisticated level of cognitive awareness; the fact that honeybees “get” the concept of zero is indeed sophisticated. Further, Dyer contends that the number zero could be considered the backbone of modern math and technological advancements. “It’s a difficult concept to grasp and a mathematical skill that doesn’t come easily,” he says. “It takes children a few years to learn.”5

Addition and subtraction are complex because they require two levels of processing. It “involves the complex mental management of numbers, long-term rules and short term working memory,” Dyer adds. Honeybees are known to return to places they’ve learned are good sources for food. Consequently, the bees return repeatedly to collect nutrition and continue learning, just as the researchers had arranged it.

The bee experiments, conducted by Scarlett Howard, a Ph.D. researcher in RMIT’s Bio Inspired Digital Sensing-Lab (BIDS-Lab), started with marking selected honeybees for identification, then training them (essentially luring them) to visit a specially designed testing apparatus — a Y-shaped maze.

What helped scientists in their experimentation, though, was the successful use of punishment and reward. When they did it right, they received sugar water as a reward. When they didn’t, they got a bitter-tasting quinine solution instead. An article from RMIT described the drill:

“When a bee flew into the entrance of the maze they would see a set of elements, between [one] to [five] shapes. The shapes were either blue, which meant the bees had to add, or yellow, which meant the bee had to subtract.

After viewing the initial number, the bee would fly through a hole into a decision chamber where it could choose to fly to the left or right side of the maze. One side had an incorrect solution to the problem, and the other side had the correct solution of either plus or minus one. The correct answer was changed randomly throughout the experiment to avoid bees learning to visit just one side of the maze.”

Bees: Applying Long-Term Rules — A Different Type of Number Processing

When the experiment first started, the bees chose randomly until they figured out how to solve the problem. After more than 100 trials, each lasting four to seven hours, eventually, the bees demonstrated the new skill set of learning the colors associated with numbers. Blue meant plus one (or +1), while yellow meant minus one (or -1). The bees’ skill was even more specific, however:

“When Howard periodically tested the bees with an image that contained no elements versus an image that had one or more, the bees understood that the set of zero was the lower number, despite never having been exposed to an ‘empty set.’”6

“This is a tricky neuroscience problem,” Dyer explains. “It is relatively easy for neurons to respond to stimuli such as light or the presence of an object, but how do we, or even an insect, understand what nothing is? Their brain can manage a long-term rule and applying that to a mathematical problem to come up with a correct answer. That is a different type of number processing to spontaneous quantity judgments.”

Children learn early on that plus or minus signs are symbols that mean quantities are either added or taken away. Howard explained that such an understanding of math symbols could be considered a complex language, at least when it’s being done by bees.

Other Animals Exhibit Math Skills, Too

Scientists have been debating whether animals either already know or have sufficient brain power to learn complex number skills, as it takes a higher level of mental control to parse numbers, long-term rules and intelligently utilize short-term working memory. According to Dyer:

“We’ve long believed only humans had the intelligence to get the concept, but recent research has shown monkeys and birds have the brains for it, as well. What we haven’t known — until now — is whether insects can also understand zero.”7

African parrots are known for their abstract, inferential reasoning. Crows have the ability to solve complex problems, and pigeons demonstrate mathematical abilities. Some experts believe pigs are even smarter than cats and dogs. Chimpanzees have shown they can learn sign language, and bottlenose dolphins are one of the few species with the ability to both develop their own vocal signature and recognize the same in other dolphins.

The researchers observed that while bee brains contain fewer than 1 million neurons, humans possess 86,000 million. One problem with developing artificial intelligence is that it requires enabling robots to function in very complex environments, but Dyer noted that the study’s findings opened the door to ways different brains could represent zero.

One of the most surprising concepts gleaned from the experiments, at least for some of the scientists, was that a large brain is not necessary in order to solve tasks. “If bees can perceive zero with a brain of less than a million neurons,” Dyer concluded, “it suggests there are simple, efficient ways to teach AI new tricks.”