In her quest to define odor, inside and out, Christine Chee has recently collaborated with Caltech scientists building an electronic nose. The so-called "e-nose" is designed to detect and identify odors by means of polymer sensors that "sniff" the chemical offered to them. In recent experiments the researchers tested the electronic nose's ability to detect differences in odor quality. Does it work as well as a biological nose? Does it do the same thing? Do things that smell similar to us smell similar to the e-nose and vice versa: Do things that smell different to us smell different to the computer sniffer?

An electronic-nose-versus-rat experiment underscored the difference.

Chee based the experiment on what she already knew --that odor perception could be divided into three categories: Intensity -- how strong is it; Quality -- how much like a rose smell, say, is this smell; Hedonics -- how pleasant or unpleasant is it. What role does the emotional system play in olfactory processing? And, does the animal discriminate between odors based on issues of "dislike" and "like?"

Chee found out something interesting --Hedonics is important for the rat, but not the machine. Indeed, odors that are chemically similar may both smell like oranges. Rats could distinguish between "good" oranges and "bad" oranges, but the e-nose could not.

"We know the two odors appear similarly in the electronic nose's `perceptual space,' so to speak," she says. "When we look for the data for that chemical, the odors group strongly together. Nonetheless, the rats could tell the difference right away."

One can't expect an electronic nose to have memories of oranges --good or bad -- or any other positive or negative feeling about a smell. Yet. It will take some time before an incipient silicon Proust savors the odor of a something it "loved" in "childhood," churning out the cybernetic equivalent of inspired prose.

Chee wants to rule out the Proustian experience --the subjective evaluation of a smell created by memory-- in her experiments as well. To discover what is purely olfactory in odor discrimination, she needs to tightly control the experiments so that complex processes such as remembering all the "good" or "bad" oranges and getting emotional about them are not part of the discrimination process. "Now we know it's an issue we must control for," she says. "When you have an electronic nose you can look at odor processing without having to deal with the constraints of a particular neural system. It's a nice abstract way of investigating olfactory perception."

-- Kathleen Stein