When feeling peckish, we might take a leisurely stroll into the kitchen. A fruit fly on the other hand must dodge countless obstacles before alighting on a fruity morsel. But when zipping about in the air, nearby objects are little more to a fly than a series of looming shapes. To determine how a fly decides when to steer away and when to land, Michael Dickinson and his student Lance Tammero designed the Fly-O-Rama', a virtual reality flight simulator, to trick the flies into landing or taking avoiding action. To find which visual cues cause a fly to land or veer away from a collision, they presented different visual patterns and measured what the flies did with their wings and legs (p. 2785).

Having worked for the last five years on flight aerodynamics, Dickinson wanted to get to grips with the higher order behaviour of flying insects: to know what causes a fly to starting zigzagging about and to know what visual cues are crucial in decision making.

Feedback from the fly's wings drives the position of the square target in the Fly-O-Rama, so Dickinson and Tammero could expand the square target in different positions that had been chosen by the fly itself. Therefore,sometimes the target was straight ahead when expansion occurred and at other times, it was only visible out of the corner of the fly's eye. The speed at which the square got bigger was also changed, by the scientists, so that sometimes it appeared to approach slowly and at other times more rapidly.

This cunning experiment reveals that when objects were directly ahead,flies prepare to bring down their leggy undercarriage for landing, whereas when the square is seen from the side, they try to navigate away. To Dickinson's surprise, however, the two behaviours are not mutually exclusive. Sometimes the flies bring their legs down, as if about to land, while executing a typical avoidance response. This makes sense' says Dickinson,since it's not costly to be cautious.'

Interestingly, the flies reacted quicker when avoiding a collision than when preparing to land. This clue gave Tammero and Dickinson the idea that the two behaviours must have separate pathways in the fruit fly's brain: a slow,fixed pathway for landing and a quicker, flexible pathway for crash-avoidance. Based on this realisation, Tammero and Dickinson have proposed a model for the fly's decision-making process. This new model takes both control pathways into account, as well as adding the effects of cells in the fly's brain that are tuned to specific patterns of visual motion, such as those caused by looming objects.

Dickinson admits that gluing a fly to a stick could introduce some minor behavioural artefacts, but he believes that understanding the sensory information that controls a fly's eye view of the world could help us designmicromechanical flying insects'. He explains that these microscopic robots could gather meteorological observations, or even be used for surveillance. Maybe it won't be long before were all seeing the world from a fly's perspective!