Hive minds: Honeybee intelligence creates a buzz
http://www.newscientist.com/article/mg2 ... ?full=true
28 November 2012 by David Robson
Magazine issue 2892.
Bees can do remarkable things with a brain the size of a pinhead, raising some intriguing questions about the nature of intelligence
I CERTAINLY wouldn't want to meet this bunch in a dark alley. Some are sitting and glowering at me from the shadows, and others are brawling in an unruly scrum, their wings and limbs flailing against the sides of their Perspex prison. Every last one of them is armed, and I can't help wondering if they are planning some kind of coup. Fortunately, I am assured that they can be easily placated with a quick fix of the sweet stuff. "Mostly, our bees collaborate quite happily," says Lars Chittka, whose lab I am visiting at Queen Mary, University of London.
That's just as well, because these miniature brawlers show an extraordinary intelligence when they are given the chance to shine. Chittka and others have found that bees can count, read symbols and solve problems that would perplex some of the smartest mammals. Some have an eye for art appreciation, having been trained to pick either Monet or Picasso's paintings from a choice of the two artists' work. They may even have a form of self-awareness, and all of this with a brain the size of a pinhead. Studying how they are capable of such great ingenuity promises to reveal much about the evolution of intelligence. It might even provide a new perspective on the workings of our own brains.
Bees have long enjoyed our admiration. Ever since the ancient Egyptians began to cultivate their taste for honey, the hive has been revered for its apparent altruism and tireless work ethic. Whether bees themselves are intelligent has been a matter of dispute, however, with many considering each individual to be relatively stupid - a mindless cog in the greater honey-making machine. As the Latin proverb had it: "una apis, nulla apis" - "one bee is no bee".
Hints of apian intellect began to emerge with the research of Austrian zoologist Karl von Frisch. Working in the years around the second world war, he observed that foraging bees often perform a strange jive across the honeycomb - the famous "waggle dance", the steps of which signal the direction and distance to nearby flowers.
We have now uncovered a rich repertoire of behaviours under the hive lid. Studies of the choreography of the waggle dance, for instance, have revealed that a worker will interrupt another's jive with a butt to the head if it has found danger - a spider, say - at the location (Current Biology, vol 20, p 310). Bees also display an extraordinary range of housekeeping chores, including spring cleaning, mutual grooming and a form of surveillance in which "bouncers" guard entrances against intruders. The hive has even evolved its own air conditioning; when temperatures soar, the workers sprinkle water over the honeycomb and beat their wings to produce a cooling draft.
In total, Chittka estimates that we have now recorded around 60 separate behaviours for worker honeybees, including six different kinds of dance (Current Biology, vol 19, p R995). These achievements seem to overshadow the abilities of many mammals. Rabbits are thought to show about 30 distinct behaviours, and the beaver has about 50 in its busy life felling trees, building damns and storing food. Even the bottlenose dolphin's 120 or so routines are only about twice the number a worker honeybee manages.
Despite this bulging portfolio of behaviours, many zoologists have remained sceptical about apian intelligence, believing they were seeing hard-wired instinct rather than flexible thought. "The brain of a bee is the size of a grass seed and is not made for thinking," said von Frisch in 1962. However, that view is now changing, as Chittka and others discover a surprising mental agility behind the bee's bumbling exterior.
Chittka's first revelation came while he was investigating the way honeybees navigate to a flower patch. Varying the number of 3.5-metre-tall tents between a hive and a feeder - "It looked more like an art installation than an experiment" - he found that foragers seemed to count landmarks rather than using the overall distance when working out where to land. Subsequent research has confirmed this numeracy, showing that bees can match the quantity shown in simple pictures of shapes to find a reward. In one trial they were shown three leaves and then had to choose between two and three lemons, for instance - a test they passed with ease (see diagram). The ability to match signs using different symbols is crucial, showing that the bees did not just rely on a memory of a specific image but understood the underlying number. But this ability is limited: bees can only count to four.
Might bees be able to grapple with other abstract rules? This is a question that Martin Giurfa at the University of Toulouse, France, has explored over the past 10 years by testing bees' powers of categorisation. "Many people told us we were crazy - but we liked the challenge," he says. Giurfa started out by training his bees on the concept of symmetry and they quickly learned to sup on a sweet reward under symmetrical signs while avoiding asymmetrical pictures (Nature, vol 382, p 458). They have since learned spatial relationships such as above/below and left/right and have also mastered the concept of same/different. What is more, the bees easily transfer their learning to new situations - if they are trained to search for smells that are the same, they are subsequently able to pick visual signs that match, for instance (Nature, vol 410, p 930).
Giurfa's latest results are more impressive still. With colleague Aurore Avargués-Weber, he found that bees can combine concepts they have learned. When trained to search through pairs of shapes, for instance, they based their choice on the colour (whether the two shapes were the same shade, or different) and spatial arrangement (whether they were stacked vertically, rather than aligned side by side). "It presupposes a greater level of abstraction," says Giurfa. His bees mastered the task after just 30 trials, compared with the thousands of attempts required by some primates (PNAS, vol 109, p 7481). That's not all. Putting these skills to the test in a labyrinth, honeybees can learn to use abstract signs to find the way to a reward. Importantly, they can then grasp that the same signs mean different things in different mazes - suggesting an understanding of context (Journal of Comparative Physiology A, vol 181, p 343).
Many cognitive scientists believe that such deliberation reaches its apex in a trait known as metacognition. This ability to introspect and judge the quality of your own thoughts - whether you are certain about something or simply going on a hunch - is often considered to be the keystone of a conscious mind. Identifying metacognition in animals that lack language is a tough challenge but, through a series of canny tests, it has been demonstrated in just a small group of primates and dolphins. Now there is preliminary evidence that the honeybee may be a member of this select club.
Clint Perry at Macquarie University in Sydney, Australia, first trained his bees to discriminate between different images using tests of varying difficulty. In some later trials, he then gave them the option of an escape port if they didn't want to risk getting it wrong. Not only were the bees more likely to avoid the more difficult trials than the easier ones, they also took longer to decide on the harder tests. As a result of the escape port, they were also more likely to answer correctly on the trials they did decide to take, suggesting that the bees had accurately judged which trials they could and couldn't pass. Perry presented his preliminary results this year at the Tenth International Congress of Neuroethology at the University of Maryland in College Park, and although he hasn't yet published the full details of his experiments, Chittka thinks his conclusions are plausible. "This performance would certainly be taken as evidence for metacognition if the study was done with vertebrate subjects," he says.
As the bee's CV continues to grow, researchers have begun to ask how and why they evolved such a rich cognitive palette. Some clues might come from their nearest family. Although honeybees are among the most studied insects, there is plenty of evidence that many of their cousins - including bumblebees and ants - also show advanced learning. All these insects have particularly large and intricate "mushroom bodies" - the dense orbs of neural networks involved in learning and memory in the insect brain. Because bees and ants are mostly social creatures, this capacity was thought to have evolved to deal with the demands of living in a big group. A recent comparison, however, suggests that the expansion kicked off 90 million years ago in a solitary wasp that ultimately gave rise to all these social insects. If so, apian intelligence may have originated for hunting and overwhelming prey, before later being co-opted for a more cooperative and peaceful lifestyle.
Despite the expansion of its mushroom body, a bee's neural machinery is still minuscule compared with other intelligent creatures such as primates and cetaceans. Human skulls house about 85 billion brain cells, whereas a bee has fewer than a million in a brain measuring less than a millimetre cubed. How bees achieve so much with so little is a mystery, although their size might have some advantages. It takes less time to pass signals between neurons if they are closer together, which should mean that insect brains can process information more efficiently. These messages may also be less susceptible to electrical noise - something akin to the static sound on a bad phone line. Noise is a particular problem when communicating over long distances, so bigger brains use digital "on/off" firing that can persist through the crackle. Within the tiny insect brains, however, a graduated, "analogue" signal can hold up, with nuances in the amplitude conveying the information. "That can transmit huge amounts of information over a short amount of time," says Jeremy Niven, a neuroscientist at the University of Sussex in Brighton, UK.
Cognitive shortcuts
Even so, it is likely that bees take cognitive shortcuts. "Insects force us to think about whether behaviours are as sophisticated as we think they are, or whether they are based on heuristics using simple assumptions," says Niven. Indeed, computer programs simulating the activity of neural networks suggest that apparently complex abilities such as counting and categorisation can be conjured from just a few hundred brain cells. Such results should be considered with some scepticism, because these models scrub away the messiness of the real world that undoubtedly needs more machinery to process. Nevertheless, alongside the achievements of bees, they do seem to confirm that relatively few neurons can go a long way when used efficiently. This could tell us as much about our own brains as those of the bee. "It could be that even we humans use rather simple techniques," says Mandyam Srinivasan at the Queensland Brain Institute in Brisbane, Australia.
We shouldn't get carried away, though. Bees may have wide-ranging abilities, but Chittka suspects their talents do not run very deep. Consider perception. Bee vision is generally pretty poor - they tend to rely on the outlines of objects, while missing the fine details. That is partly a question of eyesight, but it may also be down to the amount of grey matter devoted to vision. Our visual world is much richer, and we are very good at processing many parts of a scene at once - which is why a familiar face in a crowd will jump out at us. Honeybees apparently lack this capacity for "parallel processing" - if they are looking for a certain colour in a sea of objects, they check each one in turn, as if they can't take in the whole scene with a single glance.
There is also a huge gulf in memory capacity. The limit of human recall has yet to be found; think of all the words you know, the people you recognise. Even smaller-brained animals such as pigeons can learn to recognise thousands of images. Although bees are quick to pick up new rules, they are soon overwhelmed by large quantities of new information. Honeybees can be trained to associate certain smells with the location of different feeding sites, for instance, but they become less accurate once the number of sites exceeds two. Their inability to form connections between different events will be one of their biggest limitations, says Niven. "It gives them less opportunity to make predictions about what's going to happen in the world around them."
Still, as a simple model of intelligence, bees have huge potential. "They give us a much better handle on the neural underpinnings," says Niven. As we learn more about how bees think, the hope is that eventually we will pin down the anatomy of intelligence - working out how different networks of neurons give rise to different skills.
If nothing else, the bright sparks I met in the lab might draw attention to the myriad other intelligent life forms hiding in our gardens and cupboards and crawling under our floorboards. Not just the honeybee's closest relatives, but also its mortal enemy, the spider, and even such maligned creatures as the cockroach. If bees have taught us anything, it is that we should be prepared to be surprised by what a tiny brain can do. "After so many years, I've lost my prejudice," says Giurfa. "I've learned to respect these animals."
Agent bee
Given bees' extraordinary sense of smell and quick wits, some researchers are wondering whether they could be used to sniff out trouble in war zones. Bees trained to associate a sugar solution with the smell of a commonly used explosive would hover around landmines, for example. A laser radar system could then be tuned to detect the light scattered from their beating wings, allowing the operators to view their movements over a large area, at a safe distance.
Bees' detective abilities might also be used in hospitals. Some illnesses, including certain kinds of cancer and tuberculosis, leave patients with telltale odours that a bee could be trained to associate with food. Then, if the distinctive smell is present on a sample of breath or urine, the bee would extend its tongue, a small movement that could be picked up by a camera to give a diagnosis.
