Neuroscientists have recorded the first video footage of a tiny killer fly catching its prey in mid-air.
They were investigating how the 4mm-long insect decides when to pounce.
Apparently the flies are not much good at judging the size of a target, so they use a strategy based partly on how fast a potential meal is buzzing past.
That means the researchers could trick the flies into going for targets that were far too big, but further away and faster-moving than expected.
The experiments, published in the journal Brain, Behaviour and Evolution, used a species called Coenosia attenuata, which is so good at killing other airborne critters - including fruit flies - that organic farmers use it as a biological control mechanism.
C. attenuata is an unfussy eating machine, explained Dr Paloma Gonzalez-Bellido - and that was partly why she started to study the little predators.
"It didn't look like they have a template for what they're looking for," she told BBC News.
"They go after things that are very slow, after things that are fast, after things that are white, things that are black..."
But a fly can't just pounce on everything. So what cue is the creature's brain looking for?
Many predators are able to gauge the size of their prey, using information like the comparison between two eyes to judge how far away it is and then calculate its size accordingly, before they decide whether to strike.
This includes dragonflies, which have much bigger eyes and brain than C. attenuata.
Dr Gonzalez-Bellido and her colleagues at the University of Cambridge set out to see whether these diminutive diners could do the same sums.
"We don't really know how well such small animals see; we don't really know what the constraints are on the system," she said.
In video experiments, her team presented the flies with moving beads of different sizes and speeds, as well as real fruit-fly meals. Every pounce was monitored using two cameras, so that the movement could be tracked precisely.
The cameras witnessed some extremely poor decisions, in which the 4mm insects set off after beads 12mm across. And the team saw similar mistakes when they observed the insects outside the lab.
"In the wild, we see them take off after bees - and then turn around, halfway through the flight," Dr Gonzalez-Bellido said.
Since the flies seemed unable to use a target's actual size in their decision, the researchers set about testing the role of various other factors.
This included calculating how big - and how fast - each target would appear to the fly, taking into account the exact distance between them.
Using these apparent or "subtended" values, instead of actual ones, the researchers discovered a particular ratio between size and speed that usually triggers a pounce.
"[A target] has a subtended size. But something that subtends a size could be small and close, or large and far away. So in theory, they could go after an aeroplane," Dr Gonzalez-Bellido explained.
"One way of solving this problem is to match up the size with the speed. So a plane may have the right subtended size, but it won't come across the retina at the right speed."
This relatively simple formula for picking a target, the researchers say, has probably evolved because of the tiny size of the fly's brain and eyes - but also because the decision has to be made extremely fast.
The killer flies in the study, on average, went for targets about 8cm away and they covered that distance in less than 0.4 seconds.
"If they take too long, it'll be gone," Dr Gonzalez-Bellido said. "And one way to take less time is not to bother working out how far away it is."
Furthermore, the experiments showed that the flies can detect extremely small targets, which will only trigger activation in a single lens of the fly's compound eye.
"That means that they've really tuned the system," said co-author Dr Trevor Wardill. "They really are doing quite well with a pretty tiny eye."
Follow Jonathan on Twitter