In the recent study “Can domestic dogs (Canis familiaris) use referential emotional expressions to locate hidden food?”, researchers investigated whether dogs have the ability to locate hidden food based on the emotions shown by people’s facial expressions. It’s a fascinating question, and I was eager to learn what the experiment revealed.
Unfortunately, I had concerns related to the methods, and the result was that I found that the experiment lacked the strength it could otherwise have had. When I read scientific papers, I pay careful attention to the methods because unless I fully grasp the details of the experimental design, it is impossible to draw conclusions about the meaning of the study. Often, flaws in the methodology result in studies whose conclusions should not be accepted without further evidence from better experiments.
In the study I just read, people picked up two boxes, one at a time, and made facial expressions of happiness or disgust or they kept their expression neutral. Then, the dog was given a choice about which box to go to. (The dogs had been trained to understand that these boxes could contain food.) The idea was to see if the dogs would choose the box that was associated with happiness rather than the box associated with disgust or with a neutral expression. Overall, there was an effect of the human expression on the choices made, though many individual dogs made choices throughout the trials that were no better than chance.
The location of the testing was not always the same. Some of the dogs were tested indoors, and some were tested outdoors. Four breeds of dogs (Labs, Goldens, Border Collies, and German Shepherds) were pet dogs who lived with families in homes, and were tested indoors, but the Huskies lived in a facility for racing dogs and were tested outdoors. The investigators did in fact find a difference between dogs tested in these two conditions, but there is no way to tell if this is because of the breed, because they were tested outdoors, or because of their different social situation, which the experimenters acknowledge. With more than one variable present, determining which variable matters is a challenge.
GET THE BARK NEWSLETTER IN YOUR INBOX!
Sign up and get the answers to your questions.
Since the whole point of the experiment is to test the effect of human facial expressions, then the facial expression should be the only variable in the experiment. Unfortunately, in this study, other variables besides the ones already mentioned were not held constant. The “happy” box contained sausage, the “neutral” box contained wood shavings, and the “disgust” box had garlic in it. The neutral expression had no accompanying vocalization, but both the happy and the disgusted expression did. If the dogs did respond differently to any of these conditions, it’s hard to know whether or not the contents of the box, the presence or absence of a vocalization, or the people’s expressions (or some combination) was the cause.
In a better-designed experiment, all the boxes would contain exactly the same item, and the other differences not being tested would be eliminated as well. The researchers did a series of tests to determine if the dogs were choosing boxes based on smell alone, but only in the outdoor condition with the Huskies. (They found that these dogs were not using smell.)
To be fair, the researchers spend quite a bit of space in their discussion explaining how this experiment could have been done better in a cleaner way, scientifically speaking. I agree with their analysis that improvements could be made to the design that would add strength to their conclusions. I would have preferred for them to conduct their experiments with the improved designs before publishing, though I understand that it is immensely challenging to explore cleanly the role of human emotion in influencing dog behavior.
There are so many variables with pet dogs because of the different amounts and types of experience they have in their lives, and not only are they impossible to control for. Even if you could control many factors by raising dogs in labs and controlling their environment, then you introduce the problem of dogs who are not in the “pet dog” environment. I’m not saying that an experiment into these issues can ever be perfect, but I do think that this study could be improved upon.