Pessimistic Pooches? Depressed Doggies? Not So Fast!

Does Fido see the cup as half full? Is your dog pessimistic? Last time we saw headlines like these they were about a certain barnyard animal. Remember "Pampered pigs 'feel optimistic'"? I didn't like it then, and I don't like it now.

Roughly half of the population of dogs in the UK are likely to - at some point in their lives - exhibit "undesirable separation-related behavior (SRB)." These are things that sometimes happen when left alone, like barking, chewing up or otherwise destroying objects, and urinating (or worse!) inside the house. While some owners view these behaviors as fine and normative, and while some will seek out professional guidance, many are motivated to give up the dog. In this paper, veterinary scientist Michael Mendl and colleagues were interested in what makes people surrender their pet dogs for adoption (or, as they call it across the pond, "re-homing.") They wondered: do the dogs with the most SRB have some kind of negative underlying affect? Are they grumpy? Can dogs be, essentially, depressed?

The research team rounded up twenty-four dogs (12 male, 12 female) from two dog re-homing centres (they are in the UK, after all, where they spell things like that). After the dogs had been in the shelters for between seven and twelve days, the dogs were given an "SRB test," to see how it responded to being left alone. The dog was taken into a room where one of the researchers played with it for twenty minutes. The next day, the dog was taken back to the same room with the same researcher, and following ten minutes of interaction, the dog was left alone for five minutes. While the dog was alone in the room, the researchers recorded video footage in order to calculate an "SRB score" for each dog. The SRB score was determined by the total amount of seconds during the five minutes that the dog spent barking, whining, howling, scratching at the door, urinating, crapping, or destroying objects. The range of time spent exhibiting SRB ranged from zero to 169 seconds, suggesting significant variation. Initial analyses showed that neither gender, sterilization status (i.e. intact/neutered), breed, nor re-homing centre could account for the variation in SRB.

Then, one or two days after the SRB test, the dog was given the "cognitive bias" (or CB) test, which was meant to assess his or her affective state. Dogs were trained to inspect a bowl for food. When the bowl was on one side of the room, it contained food (we'll call this the "positive side"); when the bowl was on the opposite side of the room, it was empty (we'll call this the "negative side"). It was determined that the dog had learned the association when for the preceding three positive trials, the longest amount of time to approach the positive location was quicker than for all three of the preceding negative trials. The dogs were given up to thirty seconds to approach the bowl - if they did not, the trial was aborted.

Once the dog had "mastered" the game, the test phase began. The bowl would be placed, empty, at one of three locations located spatially between the two trained locations. They were labeled (M)iddle, (N)ear (N)egative, and (N)ear (P)ositive.

The theoretical basis for this test is that an individual's underlying mood will affect his or her decision-making. If the dog expects food to be found in the bowl, he or she will approach it quicker, suggesting an optimistic mood. If the dog expects the bowl to be empty, he or she will approach it slower, suggesting pessimism. (The relevant statistical controls were used, to account for things like baseline running speed and dog size.) So the first question to be addressed was if there were differences among the three test locations in terms of approach time. And there were.

What this means is that the dogs generalized from the training session to the testing session: food was more likely to be found when the bowl was nearer to the "positive" location than to the "negative" location.

But the main question is this: is there a statistical relationship between the results of the SRB test and the CB test? Are the dogs that have more separation-related behaviors also more likely to display the "pessimistic" response in the CB test? Well, yes. The dogs with higher SRB scores took significantly longer to approach when the bowl was in the middle location, compared with the lower SRB dogs (note, however, that this is not a group comparison, but a correlation). They had a "tendency towards" longer approaches (read: "not statistically significant," or, if you prefer, "almost but not quite." Unfortunately, "close" only matters in the game of horseshoes. Or in hand grenades) at the NN location. And there was no difference for the NP location. So basically, the only relationship between the SRB test and the CB test was seen in the most ambiguous, center location. Critically, there was no difference in approach time for either the positive or negative locations, meaning that the high-SRB group didn't simply run slower to every location.

By now, I'm sure you've figured out where this is going. The researchers' interpretation of this data is that the correlation between SRB score and approach time for the center location reflects the underlying mood of the dog.

Dogs at re-homing centres showing higher levels of SRB in a test that predicts subsequent SRB in the home situation also showed pessimistic-like behaviour in a cognitive bias test of affective state, which was unlikely to be explained by differences in running speed/motivation, learning ability, or other dog characteristics. Studies on a variety of non-human species indicate that, as in humans, pessimism is related to negative affect or mood. Dogs showing SRB may thus also be in an underlying negative affective state, although the conscious experience of such a state cannot be known for sure.

In other words, the dogs that took longer to approach the bowl when it was in the center, theoretically, were grumpier. There might have been food in it, but there was an equal chance to not find food - it was truly ambiguous. And those dogs decided it wasn't worth checking. Except...they did check. They just took their sweet time doing it. Except, they didn't REALLY take their sweet time doing it. Because if they didn't get to the bowl within thirty seconds, the trial was aborted. So they just took a slightly sweeter time doing it.

So I don't buy it. It's not the data or results I find problematic, but the interpretation. For one thing, does the behavior of a dog after a total of thirty minutes of interaction with a human, separated across two days, really relate to separation anxiety? Is thirty minutes over the course of twenty-four hours enough to develop an attachment relationship? Unlikely. What if its just generalized anxiety? Dog shelters are pretty stressful places. The kennels are 1.5x1.5 meters, connected with a 1.5x2.5 meter outside "pen." Often, two individuals share the same kennel (though in this study, apparently, all dogs were individually housed). Shelters are also relatively unhealthy places for dogs. Viruses easily pass back and forth between individuals. They even have a name for the respiratory symptoms that dogs often leave the shelter with: kennel cough. Sometimes the dogs get walked if there are volunteers, or if the staff has time, but sometimes they are not. This does not seem the ideal situation for an experiment that purports to be assessing underlying affect.

Also, the measurement of separation anxiety was limited to a five-minute period. The longest a dog spent in separation-related behaviors was about two-and-a-half minutes. Is five minutes really long enough to assess separation anxiety? I'm skeptical that in five minutes or less a dog could adequately self-regulate. Surely, some of the dogs would bark for hours if given the chance, but some of them might settle down after a while. If they had used a ten minute period instead of five, would there be any significant results? I'm not clear. They give no justification for the use of a five minute assessment period.

And another thing, with respect to the CB task. There is no indication about what the slower dogs were doing while they weren't approaching the bowls. Were they truly walking more slowly? Did they run all over the room before approaching the bowl? Did they try to interact with the experimenters instead of checking for food? Were they scratching themselves? Maybe the correlation doesn't reflect the dogs' underlying affect, but some sort of differential impulsivity or distractibility or interest in humans. Maybe the less anxious dogs simply had higher energy. Or maybe they were hungrier, and thus more motivated to check for food in an ambiguous situation.

The researchers seem to be assigning certain temperamental traits to these dogs, despite the very strong possibility that if there are indeed differences in affect, they could be due to their current states. In other words, do these tests address the dogs' personality, or simply their current situation?

Mendl, M., Brooks, J., Basse, C., Burman, O., Paul, E., Blackwell, E., & Casey, R. (2010). Dogs showing separation-related behaviour exhibit a 'pessimistic' cognitive bias Current Biology, 20 (19) : 10.1016/j.cub.2010.08.030

And one more thing: first, I downloaded just the paper. It is about a page and a half long, which is normal for a "Correspondence" article in Current Biology. But I couldn't adequately blog the paper - which means I couldn't adequately understand the paper - without also downloading the supplemental materials. The supplemental materials were nine pages long. This is a problem.