Homepage

Back

Summer 2003

Archives

With a computer and his expertise in animal behavior, especially spatial cognition and spatial memory, Villanova Psychology Professor Michael Brown, is teaching rats to reinvent themselves. Or you might say, he is giving them a legitimate profession, in a manner of speaking. Dr. Brown, working with University of Baltimore computer scientist Dr. James Otto (a former faculty member of Villanova’s College of Commerce and Finance decision and information technologies department), have taught rats to rear up on their hind legs when they smell certain odors in a lab.

Their special project, completed in 2001, was sponsored by the U.S. Defense Department’s Defense Advanced Research Project Agency, or DARPA. The goal was to see if rats actually could be taught to detect odors in the way the people have traditionally used working dogs for detecting contraband explosives, prohibited foodstuffs, or other substances that are of interest to authorities. The research was rated classified information while it was ongoing, and the professors were able to discuss it only after completion.

Brown, Otto and colleague William Long III, outlined their behavioral experiments testing the new concept of using rats to detect contraband odors, in a paper titled “Training rats to search and alert on contraband odors,” published in 2002 in the journal, Applied Animal Behavior Science.

On paper, the goal sounds fairly simple: see if rats can duplicate certain tasks traditionally performed by dogs. The behavorial experiments Brown and Otto conducted tested a new concept using rats to detect contraband odors, such as explosives, drugs or prohibited foodstuffs. The advantage of using rats is that because of their size, temperament and cost, they may be useful in situations where using dogs would be inappropriate.

“Dogs have been used to perform search and detection tests,” said Dr. Brown. “Rats, being smaller, could be used in a similar way, and their advantage is that you could use rats where you could not use a dog, for example--because of the fact it is obvious that the dog is there.” Rats, in some situations, would be totally or almost undetectable. Also, rats or other animals are able to detect substances through their sensitive olfactory systems more readily than any machine that’s been invented so far, according to Brown. Although instruments can be used to detect some contraband and explosives, and these systems have been tested in airport-type baggage screening systems, engineers have not yet built machines to detect all of them.

Brown’s research has shown that rats are capable of recognizing and alerting when they detect scents in the laboratory setting. But when all is said and done, it’s not as easy as it sounds. The testing involved countless hours of pre-training and conditioning before the rats were able to be put through their paces and ready to make them capable to be trained as “sniffer undercover agents.” So how do you train rats to distinguish odors? Brown discussed the details of his work and what it takes to transform ordinary rats into potential rodent sniffer agents.

How do you teach a rat to sniff odors? By trial and error and by introducing several different types of odors. And with a lot of patience on the part of the researcher. First, the rats need to undergo sophisticated but very basic, repetitive training before they can be taught to do the tasks foreseen by the researchers.
The odors used in the lab setting are simulated commercially available substances, such as odors that mimic cocaine, almond extract, explosives such as TNT, and motor oil, among others. These are the same simulated odors that are used to train bomb- and drug-sniffing dogs. Rats have certain physiological limitations, however. “There are certain things they can and cannot smell,” says Brown. That’s part of the challenge he faced. “We don’t know what’s included in the odors they can’t smell.”

At the start of the training, Brown used a Skinner box, or an instrumented conditioning chamber, in teaching the rats to discriminate odors. In the tests, Brown wanted to train the rats to rear up on their hind legs when they smelled a cocaine, almond or other mimic odor in the lab. “The rats especially like the almond scent,” noted Brown. “Almond extract is a favorite odor used in lab tests because rats smell it very well.”

Brown chose the rearing response as the alert signal when the rat found the target odor because rearing is a natural behavior for rats who are exploring an environment. Nevertheless, the professor needed to apply what psychologists call a “shaping” procedure to encourage a prolonged rear and make it work. “We couldn’t just use the naturally occurring response in its original form because if rats rear up anyway, how would we know they found the target?” By rewarding the rats as they reared up, he reinforced that behavior, teaching the rats to respond for seconds longer than usual. Eventually, the rats learned to rear up for prolonged intervals.

Brown also took a unique approach in developing how the animals would respond to odors. “What’s different about how we worked is that instead of bringing an air stream or the substance odor to the animal, we tested the idea that you could allow the animal to go out, search and seek a target substance, just as the bomb- and drug-sniffing dogs do,” he said. This method is more complicated because it limits the amount of control that the researcher can have over the animal. It’s up to the animal to learn the searching, seeking and detecting aspects on its own. This method also requires repeated trials.

For the test, the rat was placed inside the Skinner Box and rewarded when it detected an odor. “Every minute, a signal via a tone or beep tells the rat that a trial will start,” explained Brown. “Everything the rat sees and hears are under the researcher’s control. A stream of air is pumped into the box through a tube inserted into a hole on each side of the chamber. The target odor is pumped into one of the tubes. The rat, on hearing the signal tone or beep, will notice a light that goes off, signaling which side contains the odor. The rat eventually learns to stick his nose in the hole, which is called a “nose-poke apparatus.” When the rat inserts his nose into the apparatus, the air stream passes his nose. The air stream either does or does not contain the target odor.”

Once the rat inserts his nose into the “nose-poke apparatus,” a light comes on, two levers are activated which extend out into the box. The rat sees and learns to recognize the light. If the rat pushes either lever, and a target odor is present in the air stream, it gets rewarded with food which is automatically dispensed. A motion sensor attached to the rat via a harness detects the rat’s response and alerts the computer to the rat’s response. Likewise, if no odor is present and the rat learns to press one of the two levers, and he can get a reward by pressing the correct lever.

“If the rat is able to learn discrimination, then over many trials the probability of pushing the correct lever increases,” said Brown. “And rats can discriminate many stimuli well enough that they eventually are almost always right.” A computer program monitors the rat’s responses and sends the results to the researchers. Brown’s co-investigator, Otto, developed some of the technology used in detecting the rat’s responses in the conditioning chamber and in the sensor that remotely issues a signal from the rat’s harness. (The lab rats were fitted with a specially made harness that looks like a small life jacket. Attached to it, is a lead containing a sensor that remotely supplies the rat’s behavior, including the alert behavior. Brown notes that the rats used in the lab trials, are treated especially well. There are well fed to maintain a controlled weight and receive regular veterinary check-ups. The lab rats also savored the sucrose pellets that they were rewarded with in the odor tests. As a result special care given them, the lab rats live to a ripe age of two years old, which is one year longer than rats in nature live.)

Once the rats learned to detect and alert for odors in the Skinner Box, they also were tested in an arena (roughly one square meter in size), which was an open setting. The arena contained 25 locations where possible odors were buried within cups containing food pellet rewards. The rat’s task here was to travel around the arena and search out the various locations where the odors were hidden. In the arena, the rat also rears up for two to three seconds when it detects odor and gets the food. A motion sensor connected to rat’s harness signaled via computer to the researcher when the rats found the target odors.

A computer and software program controlled the stimuli and measured the rat’s responses, in the Skinner Box, as well as turning on and off the odor that was channeled into the nose-poke apparatus. Brown’s colleague, Jim Otto, developed some of the technology for the project.

Brown used a species called Wistar albino rats in his research. He typically prefers to use males in lab tests because there is less variance in behavior than in females. In the training sessions, the professor used about 80 rats in three different cycles. It was a learning process for both man and beast. Each time he and his graduate students trained a group of rats, the researchers also learned how to train the rats more effectively. “In the first cycle of rats that we trained, my students worked at least six or seven months before the rats performed the task, and they didn’t perform all that well,” said Brown. “By the third batch of rats we trained, the rats performed really well, but this took three months."

Normally, psychologists conduct these kinds of experiments to discover the underlying psychological and biological processes in animals. They are not too concerned with how long the training takes. For this process, however, time was one of the key issues. In order to train rats which could be used in an applied setting, “it would be necessary to get the training protocol to the point where it is feasible and timely,” said Brown.

Spy Rats in a real-world-application?

In a real-world situation, the projected practical use for training rats to detect real odors might be useful for military applications, in air-port screening, or in disaster situations. An article about Brown’s research published in the journal Nature in June 2002 states that rats are better suited than dogs to perform certain tasks such as minefield clearance. It cites that projects to train African pouched rats to detect mines are already underway in Angola and Tanzania. But the if the federal government could train spy rats as sniffer agents similar to how Brown trained rats, and if they could be deployed for real-life missions, it would require a lengthy training processes and sophisticated technology.

On the other hand, the advantage to using rats for search and detection is that they reproduce in large numbers fairly quickly to make it feasible. Also, some people do not object to using rats rather than dogs in dangerous situations such as disaster recovery.

Overall, the results of the research project were favorable. Brown proved that the concept for training rats to sniff odors and alert by a rearing response worked. Ultimately, the professor found that it was feasible to train rats to search and respond to odors of interest in a fairly short amount of time. His research demonstrated that existing technology can be used to allow this type of research to be monitored and used. To take the research to the next step, perhaps to apply it in real-life situations, different and, possibly, new technology would be needed. “I don’t know what that technology is, and I’m not sure that Jim Otto does either,” Brown said. Based on that assumption, “it’s not completely clear that the research can be taken far enough to be applicable,” he muses. “But the limitations, the things that need to be worked on, are at the technology end, not so much at the animal behavioral end. I think we have demonstrated that the animal behavior part of the system is going to work, or could work.”
State-of-the-art technology for sniffer rat agents?

Like James Bond, the potential sniffer rodent agents would need state-of-the-art technology to enable them to perform the type of research Brown has demonstrated rats can do in a lab setting.

In a real-world setting, the demands are more rigorous. Sniffer rat agents would need to work unencumbered and a technological system would need to transmit their results. While the professor believes rats could be trained to go where directed, they would need to be controlled by some type of highly sophisticated sensor or sensors. The sensor would need to allow the person in charge to remotely locate and keep track of the rat and also to detect its response. Unlike the technology used in this lab research, it would need to be wireless. That’s a challenging problem to resolve. Brown observes: “Detecting the response is no big deal; we’ve done that on a small spatial scale and we know we could do it at a large spatial scale, with GPS technology. But in between, there’s an unknown. Jim [Otto] wasn’t able to identify technology that would work at an intermediate scale, that is, a spatial scale that would be most useful in real life. So, it’s not clear whether such technology exists. In theory, certainly it could be developed, but it might cost a fortune.” At this point, the research is cutting-edge, but the application surely needs a cutting-edge remote guidance system. Developing appropriate technology would be the next step in this project.” And that’s up to Otto, or the engineers, says Brown.

Brown also estimates that even if proper technology existed that it would be quite a stretch, perhaps at least 10 years or more, before sniffer rats would be used in the real world. No need to move over yet James Bond.