A recent study in the journal Animal Cognition involving drug and explosive detection dogs has been widely reported in the online media (e.g. Medical News Daily, Sacramento Today, etc). The study, conducted at the University of California at Davis, was cleverly designed to explore the unconscious influence of detection dog handlers on their canine partners. It provides an excellent example of how greatly the beliefs and expectations of humans can unintentionally affect the behavior of dogs, even when both the human and the dog are highly trained professionals. The study is interesting in its own right, but it is also relevant to veterinary medical research since the problem of the unintended effects of human beliefs and expectations creates significant challenges for assessing the efficacy of veterinary medical therapies.
In this study, 18 detection teams were tested in each of 4 different rooms, none of which contained any explosives or drugs as a target. Therefore, any positive alert response from the dogs was an incorrect (or false positive) response. Handlers were told that a red piece of paper would be present in some rooms and would indicate the location of a substance the dog should detect. In reality, this piece of paper was a decoy intended only to create and expectation in the mind of the handlers that their dogs should exhibit a positive response. The 4 rooms contained the following:
1. No target, no paper decoy for the handler, no scent decoy for the dog
2. No target, a paper decoy for the handler, no scent decoy for the dog
3. No target, no paper decoy for the handler, a scent decoy for the dog (sausages and tennis balls)
4. No target, a paper decoy for the handler at the same location as a scent decoy for the dog (sausages and tennis balls)
Each of the teams was tested in each room, and there were a total of 225 incorrect responses (in which the dog sat/lay down and/or vocalized as it had been trained to do to indicate it had detected a target substance). There were false responses in all rooms, but interestingly there were more mistakes in the rooms with only a paper decoy for the handler than in those with a scent decoy for the dog. This would seem to suggest that the unconscious behavioral cues given by the handler affected the dogs’ performance even more than the presence of food or toys!
It is easy to see the implications of this finding for veterinary medicine. Many of the symptoms we look at in evaluating the effect of treating a patient, and many of the variables we measure in research studies of medical treatments, rely on owners or veterinarians observing the behavior of our animal patients. Most attempts to assess pain, itching, activity level, nausea, appetite, behavior and many other key indicators of health are ultimately dependant on subjective interpretation by owners and veterinarians.
We already know that our own expectations and biases can influence what we see and how we interpret it. That is, after all, the major basis for the placebo effect, and why it disappears when we don’t know if the patient is getting a real treatment or a placebo.
But what this study suggests is that the patients’ behavior is likely also affected by our expectations and biases. This is yet another element to the placebo-by-proxy effect, in which ineffective therapies are believed to be working because of non-specific treatment effects (aka placebo) on the owners or the doctors evaluating the treated patients. One of the best examples of this is the use of glucosamine for arthritis in pets, which the balance of the evidence pretty clearly shows doesn’t work but which many veterinarians and owners cling to tenaciously as a useful therapy regardless.
A study like this one shows quite clearly how our expectations and beliefs strongly effect the behavior of our dogs. And this was a study of experienced professional handlers who undoubtedly had been trained in how to avoid misleading their canine partners. Those of us without such training or unaware of the risks of such unconscious influences are likely at even greater risk for unintentionally changing our pets’ behavior in ways that conform to our beliefs. Could our dogs play more, scratch less, eat better, seem happier, or otherwise appear to benefit from treatments we give them partly because we want them to and we expect they will? This study suggests the answer is “Yes!” and reinforces the importance of properly controlled research in evaluating medical therapies.
That is a fascinating study, and I think your analysis of the implications is spot on. Thanks for posting about it!
Excellent. My own observations suggest that this effect accounts for at least some of the benefits of equine massage treatments reported by horse owners (and the perceived benefits can happen even when the horse has not, in fact, received the treatment), so it’s good to see some actual supporting evidence.
This study as you describe it is flawed. By stating quite factually to the handler that a red paper indicates the location of the planted substance as opposed to telling the handler it “may” be a positive location; the handler will likely represent this location many times to the dog and command an alert response. If told the red paper “may” indicate an odor; you are testing the handler trust in his dog’s ability and able to see if the handler is trying “talk his dog into an alert” which may be false. Tagging hide locations is common in the early stages of detection dog training when imprinting new odors, reinforcing the correct alert response and offering the reward to the dog. Especially with explosive detection dogs, there are 30 or so odors and compounds the dog should detect and many agencies do not have access to them all or train on them all. Again, was this study presented as a training exercise or a test? If presented as a training exercise, the handlers were mislead (lied to) and reinforced the alert response where they were told a positive alert should have occurred. Because in the training environment there is an element of trust in the person running the training; that the event is for the betterment of the team. In training, if the dog false to alert at source, you enforce the alert response to imprint a new odor or reinforce one the dog has not been exposed to for a long period of time. Those who put on the study just screwed up training on 18 canine teams by forcing through deception, the handler to imprint false odors on the dogs. This will have to be trained out of them since now some will likely alert to sausages and tennis balls. If it was presented as a test, the instructions were flawed and anyone who ran their dog on it was not professional or smart enough to see flaws and the damage that occur to a well trained dog. Yes, I work an explosive detection dog for a large agency.
I appreciate your insight into the details of detection dog training. It is not an area in which I have any special expertise.
I can’t answer in detail your questions about the protocol of the study, as I only have access to the abstract which I linked to in the article. It does appear that handlers were intentionally deceived into believing the red tag marked a positive location, since the whole point of the study was to see if the handler’s belief inflluenced the responses of the dog. It was presented as a test, not a training exercise, so as I understand it the handlers should not have intentionally reinforced a response to the tagged site. As for whether this significantly impaired the dogs involved in the study, I have no idea and I’ll defer to your experience on the subject. If you have the appropriate connections to contact the agencies involved in the test and could let us know whether or not this seemed to be a problem for them or their handlers, I would be interested to know the answer.
My only purpose in reporting the study was that, as a veterinarian, I frequently see situations in which owner or doctor expectations influence how they perceive the symptoms or response to treatment of my patients. This study suggests that more than that, our expectations might very well influence the behavior of the dogs, which has important implications for how we interpret their behavior when trying to figure out whether or not our treatments are helping them. It would be a shame if this experiment interfered with the dogs’ working abilities, but I don’t think even if that is the case it changes the conclusion of the study, which is that we can unconsciously influence our dogs’ behavior and not be aware that we are doing so, even if we are highly trained in canine behavior.
**Actually, I found a full text copy of the study, which might answer your questions. You can read it here:
http://www.springerlink.com/content/j477277481125291/fulltext.html