BERKELEY -- Flies are among the most visual of insects as they barnstorm through the air and engage one another in aerial dogfights.

So what better way to study their quick maneuvering than to stick them in a "virtual reality" chamber and record how they bank and roll in response to changing images.

Using just such a virtual reality chamber, Michael Dickinson and his UC Berkeley colleagues have solved a long-standing puzzle about how flies fly. For being such a visual creature, no one could find a direct connection between the fly's visual system and the muscles that control their single set of wings.

In this week's issue of Science (April 10, vol. 280), Dickinson, an assistant professor of integrative biology at UC Berkeley, reports that information from the eyes feeds instead to vestigial organs called halteres - the evolutionary remnants of a second set of wings - which act as the fly's gyroscope. The halteres then relay signals to the wing muscles to alter their stroke or angle of attack.

This seemingly illogical system, relaying visual information through a muscle-bound organ, nevertheless is extremely fast. House flies can change course in response to visual images within an amazingly short 30 milliseconds.

"We knew behaviorally that a lot of flight is under the voluntary control of vision, but we've had difficulty identifying functional connections - wing steering muscles don't respond to changing visual images," says Dickinson, a neuroethologist. "But when we looked at the steering muscles of the halteres and presented the animal with a visual pattern, we got robust activation of the muscles."

"One possibility is that visual control in the fly works by fooling its gyroscope," he adds.

The discovery is important not only for what it tells us about how flies fly and how they evolved, but also for the novel tip it gives designers on how to stabilize small, insect-like robots during flight.

Unlike most flying insects, flies have a single pair of wings. The hindwings have diminished in size to millimeter long, lollipop-like organs called halteres that beat like a normal wing during flight but play an entirely different role. They essentially act as gyroscopes, telling the fly how its body is rotating and sending signals to the wing muscles to correct its orientation. They are analogous to the human inner ear, which is critical to maintaining equilibrium.

The halteres, beating out of sync with the forewings, are the key to the fly's aerodynamic prowess.

"Flies are the most accomplished fliers on the planet in terms of aerodynamics," Dickinson says. "They can do things no other animal can, like land on ceilings or inclined surfaces. And they are especially deft at takeoffs and landings -- their skill far exceeds that of any other insect or bird."

Dickinson has been studying how the sensory cells at the base of the haltere detect changes in the haltere's position resulting from forces exerted during flight. The major factor is the Coriolis force, which pushes things sideways as they move on a rotating body. This force, which causes winds on the spinning Earth to curl into eddies and cyclones, pushes the beating haltere to the side when the fly's body rotates. The sensory cells, called campaniform sensilla, then send signals to the steering neurons of the wing to alter the reflex beating to stabilize flight or change direction.

Remove a fly's halteres and it becomes unstable and quickly crashes to the ground, he says.

The key to Dickinson's new finding was discovering a 1948 paper in which P. F. Bonhag of Cornell University reported his dissection of a set of tiny muscles attached to the halteres in the horse fly. Long since forgotten, these muscles appear to be vestiges of muscles used to steer the hindwing before it became specialized into the sensory structures we recognize as halteres.

Though these steering muscles - 11 of them in the house fly, analogous to the 17 steering muscles attached to the fly's forewing - evidently are no longer important in generating aerodynamic forces, Dickinson had a hunch they might be the missing connection between the visual system and the flight muscles.

Looking instead at blowflies, Calliphora vicina, he and postdoctoral researchers Wai Pang Chan and Frederick Prete stuck glass recording electrodes into several of the 11 minuscule steering muscles of the haltere and measured their activity when the fly was presented with various moving images in the virtual reality chamber.

"Lo and behold, the steering muscles were strongly activated," Dickinson says. Different muscles contracted depending upon whether the pattern of dark lines moved up, down, across or diagonally.

He suspects that these contracting muscles tweak the halteres, which in turn relay the effect to the wing muscles to control flight.

Dickinson plans further studies to determine exactly how the steering muscles affect the halteres.

"Flies use visual information combined with mechanosensory information to fool the halteres, probably changing the way the halteres beat or the sensitivity of its receptors," Dickinson says. "That information is then sent forward to the wing muscles."

Contrary to expectations, channeling visual information through the halteres is probably a more stable way to achieve visual control of flight, he says. Rather than turning off or overriding the gyroscopes - the halteres - it is more effective to fool them.

By connecting to the haltere steering muscles, the fly also is taking advantage of an already existing fast, reflexive control system. The halteres, just one nerve cell away from the motor neurons of the wing, are designed to react quickly - reflexively - to yaw, pitch and roll in the fly. This allows, for example, a male fly to rapidly change course when pursuing a female.

Moreover, it seems certain now that the halteres derived from an earlier set of hindwings, and that flies adapted the hindwings' steering muscles to a different purpose. In other insects, such as locusts, the hindwings beat out of phase with the forewings - just like the halteres - but probably are not able to affect equilibrium, he notes.

"In many insects the forewing follows what the hindwing is doing, and this is still going on in flies," Dickinson says. "The same basic circuitry is there in the fly, the hindwing entrains the forewing, they've just reused the muscles and sensors on the hindwing in a very clever way."

The virtual reality chambers Dickinson uses in his laboratory are cylinders lined with about 2,000 green diodes that present black stripes moving at various angles, and at a speed between 3,000 and 4,000 frames per second. Flies are tethered to a post in the center of the cylinder.

The high-speed images are necessary because fly's eyes can see movement 10 times faster than the human eye. In other words, while humans see a constant image when it flickers on and off more than 30 times per second, flies do not see a continuous fused image until the flicker rate reaches 300 times per second.

Their compound eyes, on the other hand, contain between 550 and 600 individual ommatidia (in the fruit fly) that see very little detail.

"Flies have poor spatial resolution but spectacular temporal resolution," Dickinson says. "Their eyes are built for speed."

The research was supported by the National Science Foundation and the David and Lucile Packard Foundation.