Dogs have been used to detect everything from drugs to cell phones. An enormous amount of trust is placed in their ability to accurately track down scents. Unfortunately, according to a new study by Lisa Lit, Julie Schweitzer and Anita Oberbauer of UC Davis, that trust may be misplaced.
To be clear, it isn’t the dogs’ ability to detect scents that is called into question, it’s how accurately they can relay that information. Lit and her colleagues found that the dogs routinely give their handlers the answers that the dogs thought their humans want.
The researchers set up a testing site with four rooms. The handlers were told that there could be up to three target scents (the scents to which the dogs had been trained to alert) in each room. In two rooms, there would be markers indicating where the scents were located. However, in reality, there were no target scents in any of the rooms. Two of the rooms did contain markers, but they were randomly placed. In addition, one of the rooms with markers and one without also contained distraction items: sausages and tennis balls.
Eighteen teams of dogs and handlers passed through each room twice. Despite the fact that none of the rooms contained any target scents, the dogs alerted a total of 225 times. Most of the alerts occurred at locations indicated by the markers. In other words, the dogs reacted more to their handlers’ expectations than to dog distractions like sausages.
This result should be troubling for any agency that relies on dog detection. There is no doubt that dogs are capable of accurately detecting all the things for which they are currently used. The question is whether their human handlers can be properly trained to avoid giving subconscious prompts that guide their dogs in the wrong direction.
In the meantime, perhaps we should look to other animals that care a lot less about what we think than dogs do. For example, Bart Weetjens has trained rats to detect landmines and to diagnose tuberculosis.