When Your Dog Reads Your Subtle Cues

Every animal cognition researcher lives in Clever Hans's shadow. This horse was clever, but not in the way that people thought he was. Hans lived in Berlin at the turn of the twentieth century. His owner, a Mr. von Osten, had trained him to do arithmetic (or so his owner believed). Given a multiplication or division problem, Hans tapped out the answer with his hoof. He knew factors, decimal places; he could count the people present in a room, or the number of bespectacled people present in a room. Given a new problem, Hans was unerringly correct. Hans went, in the parlance of our time, global: crowds lined up to see the famous counting horse. Suddenly the possibly hidden and latent intelligence of animals was on everyone's mind. Skeptics lined up, too, but not until one Oskar Pfungst, a psychologist, came along, did they have their proof. Pfungst demonstrated that Hans was using cues unintentionally provided by his owner or anyone asking Hans questions. Without being aware of it, they were telling him when he'd reached the final answer with their bodies: relaxing their shoulders, tilting their head, widening their nostrils, and so on. These subtle acts were Hans's clue to stop tapping.

I think Hans was pretty clever, myself. He didn't need to learn arithmetic: he could just get the "answers" from the humans around him. But his is a cautionary tale to behavioral scientists: beware of inadvertent cueing. This is especially relevant in studies looking at animals' performance in experimental tests, because often the animal acts in interaction with a human experimenter. Is the chimp indicating which of two piles of M&Ms is larger because he knows which is larger, or because the human accidentally glances at the larger pile?

Today, researchers do a good job of avoiding the pitfalls of inadvertent cueing. But when thinking about how dogs behave, "cueing" takes on a new dimension. For dogs are especially sensitive to a myriad of cues that indicate what we humans, the ones who bred them and brought them into our homes, are going to do. Dogs' sensitivity to cues is in some sense, as with Hans, what makes them so "smart". They predict us; they follow our smallest gestures; they seem to know us.

The cues we give are often subtle. We might not even know that we are giving a different cue, when we rise from our chairs to go to the fridge, as opposed to rising to take the dog for a walk. Your dog does.

I was thinking about the subtlety of cues when reading a paper which might speak about cues we give to dogs. In this paper, the authors (Lit et al, 2011) tested 18 drug- and explosive-detection dogs and their handlers. The teams were told to search for the scent of drugs or explosives in an experimental room. Only, there were no none there. However, the handlers had been told that some of the locations were baited. And in the end, the teams "found" the presence of the scent again and again. Of course, they were wrong every time: there were no drugs nor explosives present.

These were experienced teams, who have clearly proven their abilities to find hidden contraband. So what was happening? It appears that the handlers, thinking there were scents present where there weren't, cued the dogs. At least, that is what has been reported. I read about the paper in Scientific American, whose headline read, in part, "When dog handlers believed there were drugs or bomb materials, their dogs called more false alarms".

Is that true? It's

dogs look where we look

a subtle point, but the headline isn't quite right. The team called more false alarms. That is, the handlers reported that the dog found the drug. Two things could have happened, and the study's data does not distinguish between them. Either the handlers tipped off the dogs to what they thought were correct locations -- the Clever Hans "effect" -- or the handlers simply misreported what the dogs were doing. They might have "thought" or "sensed" that they saw the dog alert to a scent. This is classic confirmation bias: you see what you expect to see. An experienced handler expects to see his dog find the scent.

The study is presented as a cautionary tale against cueing a dog, but I think it is also a cautionary tale about research: the findings are often more subtle than the later headlines will tell you. Look more closely. (Your dog certainly is.)

Cited: Lit, Lisa, Schweitzer, Julie B., Oberbauer, Anita M. 2011. "Handler beliefs affect scent detection dog outcomes." Animal Cognition (on-line).