Quail are smarter than computers, according to a study that compared the abilities of each to recognize the sounds of human speech.

Researchers thought it would be simple for a computer to recognize spoken words. For example, "b" and "d" sound different. But each of these sounds varies enough individually when spoken and are similar enough to one another that researchers have failed, despite numerous attempts over the years, to make computers recognize spoken language. Even the most sophisticated work to date has shown only limited success.

The "d" sound does not have any single acoustic feature that is exclusive to it; no single aspect of the "d" sound allows it to be distinguished from a "b." Instead, it appears, there are several features, some of which may sometimes be absent, that together make the "d" sound.

As a result, linguistic researchers suggested that the human ability to distinguish the sounds must be the work of a separate mental ability by which people can analyze and categorize different sounds.

Now, the quail study shows, you don't have to be human to pick out the sounds and categorize them. You don't even have to be smart.

In the Sept. 4 issue of Science, researchers at the University of Texas at Austin and at Arizona State University reported that Japanese quail can be trained to listen to human speech and spot the "b"s, "d"s, and "g"s.

The birds were trained to peck at a lever whenever they heard a sound in the correct category. The lever activated a food dispenser.

Given little words beginning with a "b" or a "d" or a "g" and containing different vowel sounds, the quail apparently made mental categories so that when new words with variants of the consonant sounds were presented, they could spot the ones with the sound they had been trained to recognize.

Keith Kuender, a psychologist at Texas who led the research group, said that the point is that the power to make and use mental categories, even with unfamiliar experiences such as human speech, is a power that extends far down the evolutionary scale.