ver since Pavlov trained dogs to salivate for meat powder at the sound of a bell, psychologists have used the principles of classical conditioning to study how animals and humans learn.

But only recently have they been able to peer into the brain and watch that learning take place.

Now a team of English researchers, using a sophisticated brain scanning technique called functional M.R.I., has provided a vivid demonstration of the neural processes at work in a simple Pavlovian conditioning experiment.

Like Pavlov's dogs, the subjects in the study were conditioned to associate a neutral stimulus — in this case, abstract images presented on a computer screen — with food. One image was paired with the smell of peanut butter, wafted to the subjects' noses through a tube. Another image was paired with the smell of vanilla.

The 13 subjects, said Dr. Jay A. Gottfried, a postdoctoral fellow at the Wellcome department of neuroimaging science at University College London and the lead author of the study, "were all young, healthy, right-handed volunteers who came in hungry and professed to liking both peanut butter and vanilla, which isn't so easy in England."

The subjects, who thought they were participating in an experiment about learning computer tasks, were quickly trained to associate the images with the food smells.

The subjects reacted faster to the images paired with the food odors than to other images that had no pleasant associations. At the same time, their brains surged into action, with areas known to be involved in motivation and emotional processing — the amygdala, deep in the temporal lobe, the orbitofrontal cortex, and other structures lighting up on the brain scan.

Then the researchers took their study, published last week in the journal Science, a step further. When the subjects were fed either a peanut butter sandwich or a bowl of vanilla ice cream, Dr. Gottfried and his colleagues found, the images associated with that food no longer drew as strong a response, and the subjects' emotional brain circuits quieted down. But the image associated with whichever food the subjects did not receive continued to elicit faster reaction times and a flurry of chemical activity in the amygdala and other brain areas.

Psychologists refer to this as "selective satiation." Dr. Gottfried calls it "the restaurant phenomenon."

"You go out to Lutèce and have an eight-course meal and just when you think you can't stuff another crumb into your mouth, they bring the dessert cart by and, miraculously, you have a spot for that chocolate cake," he said.

Whatever its label, added Dr. Gottfried, who did the study with two other scientists, Dr. John O'Doherty and Dr. Raymond J. Dolan, the effect reflects the fact that learning is a tool designed by evolution for survival and as such, is infinitely flexible.

"If you think about a rabbit jumping around a carrot patch," he said, "it may learn that a pile of rocks comes to predict that carrot patch. But once the patch is depleted, it behooves the rabbit no longer to follow the promptings of that pile of rocks."

The study's findings, he added, may eventually help scientists understand more about why people with eating disorders fail to become satiated on foods even after they have eaten them.

Patients with injuries to the brain's frontal and temporal lobes, Dr. Gottfried noted, areas that encompass the circuit involved in hunger and satiation, often have eating problems and may eat indiscriminately or fail to stop eating when they are full.

But Dr. P. Read Montague, a professor of neuroscience at the Baylor College of Medicine and an expert on mental function and the brain, said the study was most noteworthy for offering a glimpse of what takes place inside the brain when a rat learns to run a maze or a human to forage for lunch.

"This is what psychologists really couldn't do" before, he said. "They could sit and observe the behavior, but they couldn't eavesdrop on the black box as they were doing it."