Honey bees are living proof that bigger doesn't necessarily mean better.
Despite their modest brains, made up of about 1 million neurons, we've known for years that these insects are capable of learning and performing complex tasks.
This includes being able to count to four, understand what zero means and even solve simple addition and subtraction problems.
Now new research, published last week in the Journal of Experimental Biology, suggests we've underrated their mathematical ability.
In the experiment, honey bees were trained to enter a Y-shaped maze and given the task of distinguishing between a card with four shapes on it, or another card with either eight, seven, six or five shapes on it.
When they correctly selected the card with four shapes, they were rewarded with a sip of sweet sugary water.
But half the bees, when they incorrectly picked the other card, were given a sip of bitter quinine-flavoured water instead.
It turns out their success at the task came down to how they'd been trained.
Bees that were trained with the reward when they got the answer right and the penalty when they got the answer wrong performed much better than bees that were trained with the reward alone.
They could distinguish between cards with four and eight shapes on them, and even the trickier choice of distinguishing between four and five.
This puts honey bees on par with humans and mosquitofish.
"It shows us definitively that changing the training regime with the bees ... really impacts their performance on tasks," said lead author of the study, Scarlett Howard, who carried out the research as part of her PhD at RMIT University.
And it raises the possibility that bees (and potentially other animals) can be trained to count to higher numbers than was previously thought.
That's according to sensory and behaviour ecologist Lars Chittka of Queen Mary University of London, who was not involved in the study.
He suggested we may underestimate the intelligence of many animals, so long as training procedures are inadequate.
"This applies to lots of animal intelligence tests, not just counting tasks," he said.
Motivated to perform
Dr Howard was surprised to find that bees which only received rewards weren't able to distinguish between the card with four items and the card with eight items.
This is despite the cards showing a relatively simple ratio of one to two, which many other animals are able to distinguish between.
"I think I thought because honey bees are such great learners, this would be an easy one for them to do, but they weren't able to do it," she said.
The other surprise she encountered was that the bees trained with rewards and aversion were able to distinguish between the card with four items and the card with five items, an extremely difficult comparison for them to make.
But why? Could it all be down to how motivated they were?
Other research looking at bumblebees has shown that having a penalty for when a bee makes the wrong choice slows them down, and they become much more accurate.
In the case of the honey bees, avoiding a sip of the bitter water could increase their motivation to choose more accurately.
"It's almost like they're thinking of the consequences of their actions," Dr Howard said.
Professor Chittka compares only having rewards to giving school kids a multiple choice exam where you get points for ticking the correct answers, but no penalties for ticking false statements.
"In that case, a clever kid would not even read the statements, but tick everything as correct – that would be the simplest and fastest strategy," he said.
Small and mighty brains
Bees' small brains make them interesting to study, Professor Chittka said.
"How much cognitive capacity and flexibility can one squeeze into the microcomputer that is the bee brain?" he asked.
"This is key to finding out the minimal neural circuits required for such tasks."
This new understanding could allow us to use bees as inspiration for our own inventions.
Professor Chittka is part of a team working on a project called Brains on Board.
They're aiming to design autonomous flying robots that could mimic the navigational and learning abilities of honey bees.
