Goldfish might look as if they are just swimming around with their mouths open, but what goes on inside those mouths is a lot like a high-efficiency industrial filter, according to mathematician Angela Cheer at the University of California, Davis, and biologist Laurie Sanderson at the College of William and Mary.

Cheer and Sanderson combined computer models with careful observations of live fish to work out how fish filter feed -- a problem that had puzzled biologists for decades. In a study published in the journal Nature, they show that fish use crossflow filtration, a method widely used in industry, to concentrate food particles.

Many species of fish, including herring, mackerel, tilapia and goldfish, feed by swimming around with their mouths open or by pumping water into their mouths. Water is pushed out through gill slits in the side of the head. Food particles in the water are trapped and swallowed. Fish don't swallow much water with their food, so somehow food and water are separated.

Look in the mouth of one of these fish, and you'll see a series of arches lined with combs called gill rakers. They look like a sieve, and for a long time, biologists assumed that the rakers collected food particles as water flowed through them.

This sieving model raised a number of questions, such as why the filters don't clog with food, how food particles get from the rakers to the esophagus, and how fish can filter particles as small as a bacterium, said Sanderson.

Sanderson's team used fiber-optic endoscopes, like those cardiologists use to probe blocked blood vessels, to watch what actually happens when fish are feeding. They found that food particles were not hitting or sticking to the rakers at all. They were being carried straight past and collected at the roof or the back of the mouth.

"The theory is that they're sieving, but there's no food on the rakers," said Cheer.

Based on Sanderson's observations, Cheer built computer models to stimulate water flow through the mouths of these fishes under different conditions.

They found that instead of flowing smoothly through the mouth and out of the gill slits, the gill arches caused a swirling vortex to form at the roof of the mouth in some species. Food particles are trapped in the vortex, stick to mucus on the roof of the mouth and are then swallowed.

According to Cheer's model, water should still be passing through the rakers to get out of the gill slits. Sanderson's observations showed food particles flowing past as if the rakers were solid.

Cheer and Sanderson realized that the rakers act as a crossflow filter. Crossflow filtration is widely used in industrial applications such as winemaking, brewing and water purification. It's a valuable method for filtering large volumes of liquid without clogging filters. Instead of pouring a mixture of liquid and solid particles through a filter, the mixture is run past the filter to separate the liquid.

"What fish and industrial filters have in common is a need to concentrate particles without rapidly clogging the filters," said Sanderson. Industrial crossflow filters do eventually clog, but those in fish don't seem to, she said.

The research is published in the July 26 issue of Nature and the June 21 issue of The Journal of Theoretical Biology.

Editor's note: Images of the flow models are available. Contact Andy Fell for details.