A-mazing rat-robot man aims to mimic humans

Continuing our series on young researchers, Ayala Ochert meets Neil Burgess, the neuroscientist with his own line in laboratory rats

It is a job that children surely dream of having when they grow up. Coming to work in the morning and playing with your robot cyberpet, and then later turning to your virtual reality computer game, like DOOM or Duke Nukem.

Neil Burgess is the lucky job-owner, so it is hardly surprising that he looks forward to coming to work each day. What is more surprising is that work is based in the department of anatomy at University College London, where he is a research fellow. But, he insists, there is a serious side to his work, namely, understanding how the human brain works.

The small robot that sits on a table top in the corner of his office is not there just to keep him amused. He has designed it to simulate the behaviour of a rat as it tries to find its way around, looking for food. Though it does not look much like a rat - more a mass of wires mounted on wheels - its behaviour is a lot like a rat getting its bearings in a new environment.

Burgess, 31, achieves this life-like effect by combining known features of rat brains with others that he thinks they would need to get about. While "place cells", which respond to a rat's position in space, and "head-direction cells", which respond to the direction it is facing, have both been identified by neuroscientists, "goal cells", which mark the position of a reward, are his own invention, although he would not be surprised if they really do exist.

But the idea that every cell in the brain holds a little bit of information has sometimes been derided. "The so-called 'grandmother-cell hypothesis' says you have got one grandmother so you have got one cell coding for your grandmother," explains Burgess. "People argue that's not very likely because cells die all the time, so if you lost that cell you would have no concept of your grandmother anymore." So in the 1980s, when neural networks were being developed, it was assumed that representations were distributed across the brain among cells that only held information collectively. "But," says Burgess, "this assumption has been proved wrong again and again in different parts of the brain."

Already the robot has inspired new work on how rats find their way around,and Burgess hopes that his experiments with computer games will do the same for human navigation. A game like Duke Nukem simulates a 3-D world in which the player must defend himself against attack from every direction. "Behind the game is a really good graphics program, so you can take the monsters out and use it to simulate a small town," he explains. Volunteers in his experiments have the blood flow in their brains monitored while they "wander around" the virtual town. The experience can feel quite real. "Compared with most stimulation that you get in these functional imaging experiments, it's quite absorbing. Occasionally you even dream about the places afterwards," adds Burgess.

The images produced indicate that human and rat brains share more in common than was previously assumed. While rats use their hippocampus, an area of the brain, primarily for navigation, we use our hippocampus primarily for remembering past events. So few researchers expected that the human hippocampus would also be employed in navigation. But, while volunteers "played" the computer game, blood flow was apparently diverted to the hippocampus on the right side of their brains. Burgess reckons that a sense of space, combined in the hippocampus with a profoundly human sense of time, could together generate a memory of events. He is optimistic that he will one day be able to mimic human memory on a computer, just as he simulated rat navigation.

It is probably no coincidence that the parameters of space and time are central to Burgess's way of thinking about the brain - he initially trained as a theoretical physicist before he became interested in studying the brain. "I slowly came to the realisation that I was more interested in animate than inanimate objects," he recalls. "But I still use maths and computing in the process of developing these models, so the techniques may not be that different."

Burgess finds himself pleasantly settled in academic life. Had the offers not been so tempting, he might have left for a more lucrative career in the City, joining friends who are providing the Stock Exchange with a neural network solution to insider trading. Although he has missed out on that particular opportunity, he may yet have success in business with the company that he has set up with colleagues to distribute hardware and software to other labs. And a robot that navigates like a rat has many potential applications - perhaps even as the cyberpet of the future.