Dog talk: Extracting unambiguous answers to questions
There's a story in the dog research community about a dog being tested for its limits of detection. Seemingly confirming the myth of supersensitive odor-detecting abilities of canines, this dog appeared to be able to detect the target even at infinite dilution of samples. It turns out this dog was responding to a change in air pressure whenever it was presented with a sample-and not to the target.

Stories such as this reflect the difficulty of quantifying canine olfaction. Dogs respond to stimuli other than what researchers think they are presenting. Dogs are unable to speak and answer questions directly, and some methods that have been used to detect a dog's response to stimuli are subjective, relying on the experimenter's interpretation of sometimes subtle behavior.

Researchers at the Institute for Biological Detection Systems, Auburn University, in Alabama, have developed psychophysical methods that allow objective and unambiguous communication with dogs during testing. They eliminate competing stimuli when testing the ability of a dog to detect an odor and they teach the dog to use levers to signal clearly their response to a stimulus.

"We do it by testing the dog in a semi-sound-proof chamber," says James M. Johnston, the institute's director of behavioral research. The chamber takes away visual and auditory distractions, leaving the dog to respond only to odor.

The wooden chamber is big enough to comfortably accommodate fairly large dogs. The walls are insulated to muffle sounds, and the inside is painted gray. When a dog steps in, all it sees is an aluminum panel with a hole in the middle and levers on either side. Attached to the bottom of the panel is a pan where food appears whenever the dog performs a task correctly.

At the back of the panel is a glass chamber through which air is constantly flowing. During testing, vapor from odorous substances is inserted into the air stream, which the dog samples through the hole in the panel. Generating and delivering the samples requires especially designed vapor generators and close collaboration with the institute's chemistry laboratory, supervised by Cindy C. Edge.

The dogs are trained to stick their noses in the hole for at least one second, to ensure uniform access to the odor. They are also taught to communicate what they are smelling by pressing either lever: the right if it's the target odor, and the left if it's clean air. If they respond correctly, the dogs are rewarded with food on an intermittent schedule.

The limit of detection-or odor threshold-is found by providing the dog with a range of concentrations of the target and determining the point at which the dog's accuracy is no better than what it can achieve by randomly choosing either lever.

Identifying the detection odor signature is more complicated. The dogs work with three levers: the right for the target odor and the left for clean air. And they press a middle lever when they're smelling neither clean air nor the target odor-that is, anything else.

Using this setup, Johnston has acquired some basic facts of canine olfaction. For example, odor thresholds do not vary greatly from dog to dog, at least for those dogs they have tested, which came from dog pounds. Also, to dogs, 99% nitroglycerine doesn't smell like nitroglycerine-based smokeless powder.

Johnston says there are a lot of issues that need to be examined, such as the effects of age, sex, neutering, diseases, and veterinary drugs on odor thresholds or the relationship between detection odor signatures and the biochemistry of the dog's nose. "The dog is an impressive detection instrument," he says. "We just need to generate the science to understand it."

RETURN TO ARTICLE