Behold the Big Baud Fly

The common fly is a miracle of miniaturization, a tiny flying machine capable of split-second zigzags--and of outmaneuvering most humans. Now a group of scientists has peered at the pilot's brain and found that it too has some fancy moves. They have found a way to measure how efficiently a neuron encodes information and shown that a single neuron in the fly's eye, despite the electrically noisy environment, can handle a remarkable 100 binary bits per second (bps). The new technique, appearing in Physical Review Letters this month, could be a powerful tool for understanding how brains process information.

Most neurons work like tiny computer modems--when they fire, that's like sending out a binary 1; when they don't, it's a 0. By watching the firing patterns of neurons in flies and other animals, scientists have been able to make rough estimates of the bandwidth--the maximum data rate--that a single neuron can carry. Those studies showed that the rate was at least 60 bps. Hoping to develop an absolute measure, biophysicist Steven Strong and colleagues at the NEC Research Institute in Princeton, New Jersey, hit on the idea of measuring the entropy or "disorder" of the electrical pulses passed on by the neuron. Entropy and information content are roughly equivalent, Strong explains, because "the more disordered something is, the more information you would need to describe it." Turning that equation around, if the fly is watching something complex, then the nerve pulses should appear disordered.

So Strong, who is now at the Institute for Advanced Study in Princeton, and colleagues buried a common fly up to its neck in wax and inserted an electrode near its H1 neuron, which fires when an object passes horizontally before the fly's eyes. Then they recorded the neuron's firing while the fly watched an oscilloscope that showed a random pattern of bars moving horizontally. A predictable pattern would not tax the visual neurons, but random motions would test the maximum capacity of the neuron to carry information. To estimate how much of the entropy in nerve-pulse timing represented information and how much was just random variation, they showed the fly the same sequence several times and looked to see how the response changed. The result, says Strong, was that about half of this bandwidth contained information.

The new method is "an important step forward," says Markus Meister, a biologist at Harvard University who has already applied it in similar studies of rabbit and salamander brains. There, too, he says, the technique reveals that the neurons pack in more information than researchers had expected.