collision detection: Bees can recognize human faces. <i>Run for your lives!</i>

December 13, 2005

Bees can recognize human faces. Run for your lives!

An old folk myth claims that if you kill a bee, its hivemates will remember your face and come and get you. That's a canard, of course, but a team of British and German scientists recently discovered something even spookier: Bees can recognize individual human faces. In an experiment, they put a bunch of pictures of faces in front of some bees, and put sugary liquid on a select few of the photos to attract the insects. When the bees were later shown the photos without any sugary liquid, they made a beeline (sorry, I couldn't resist) for the ones that had previously been sugar-ified. From World Science:

The bees learned to distinguish the correct face from the wrong one with better than 80 percent accuracy, even when the faces were similar, and regardless of where the photos were placed, the researchers found. Also, just like humans, the bees performed worse when the faces were flipped upside-down. [snip]
Moreover, "Two bees tested two days after the initial training retained the information in long-term memory," they wrote. One scored about 94 percent on the first day and 79 percent two days later; the second bee's score dropped from about 87 to 76 percent during the same time frame.

Among the many mindblowing implications of the study is that it bursts apart the idea that facial recognition is rilly rilly difficult. Up until now, many neuroscientists have assumed that because facial recognition is evolutionarily crucial, and because we appear to evolved a section of the brain -- the fusiform gyrus -- that deals primarily with facial recognition, that this task is hellishly hard. But if bees can do it, and educated fleas can do it (man, I just cannot stop myself today, sorry), maybe face-recognition isn't as cognitively difficult as we assume.

(Thanks to Boing Boing for this one!)

Posted by Clive Thompson at December 13, 2005 12:01 PM

Trackback Pings

TrackBack URL for this entry: http://www.collisiondetection.net/mt3/mt-tb.cgi/1396

Comments

I wonder what happens if they took a different photo of the same person. Maybe the bees use very simple regonition and can only remember the exact same photo. This is hardly facial regonition, more like shape regonition. It needs to be dynamic to be facial, so more tests needed.

Posted by: nova9 at December 13, 2005 1:14 PM

That's definitely true that a dynamic, moving face would be an even more precise test!

But I don't think this is mere shape recognition. Human faces all have very similar outside peripheral shapes -- yet the bees were able to distinguish between two faces that seem to be, even the human eye, quite similar.

Posted by: Clive at December 13, 2005 2:44 PM

This research is interesting, but I share nova9's view that more research is needed. However, you only need a small cluster of neurons to produce a neural net associative memory capable of distinguishing closely related entities (such as human faces, or different instances of the same tree, or different flower beds). We shouldn't be suprised if insects possess such basic neural talents, I suppose.

Posted by: Chris Bateman at December 14, 2005 3:03 AM

Also, could it be that the bees are recognizing features of the photograph paper itself, and not the actual image representing a human on the paper? This experiment would be more telling if there were actual humans standing behind a glass screen.

Posted by: Dave Sandoval at December 14, 2005 10:31 AM

Chris, yes, totally -- very small neural nets can perform some extremely complicated recognition tests. I was once hanging out with a couple of AI guys who'd created a net to drive a car around a simple track. They experimented by knocking out various parts of the net until they discovered that it only required six functioning "neurons" to capably drive the car.

Dave, excellent point about the paper. Bees can see UV-A and UV-B light, so is it possible the photographs had properties in those spectra that weren't obvious to the scientists?

Posted by: Clive at December 14, 2005 11:50 AM

I find this very distressing. Time was, I thought it meant something to be a human. That we were real special. First it was reported that we have a lot in common with garden slugs what with 99% of genes being the same. Now this. It makes me want to cry, and I would do so if I hadn't read that scientists have recently discovered that certain types of bacteria are acutally able to cry too.

Posted by: daniel luke at December 14, 2005 11:23 PM

You know, Daniel, I think absolutely the oposite. It´s nice to have so many animal relatives. Less lonely!

Anyway, here's an experiment: Ten people move around (dynamically so) in a yard full of bees. Five of the people are soaked in strawberry jam. Then they all have showers and come back.

I think choosing to do painful things in order to achieve extremely abstract and often not even well thought goals, is a particuarly human trait. Does that make you feel better? :)

Posted by: eke at December 15, 2005 5:33 AM

Completely off topic here, but you have to get yourself one of these bad boys, Clive!

By the way, cool post. Freaky.

Posted by: Dusty Bear at December 16, 2005 5:19 PM

Clive - what do we conclude from the fact that, on the other hand, *we* can't recognize bees' faces?

Posted by: tmandel at December 18, 2005 4:37 PM