But where does this dog intelligence hierarchy come from? A quick search reveals the source to be the book The Intelligence of Dogs by psychologist and trainer Stanley Coren. The resulting rankings are widely available, for example here where dog owners will be either delighted or confounded by their pet's placing. Either way, they will be interested in knowing how dog intelligence was determined.
According to this website the three criteria are
According to this review of the book these tests mostly come down to the dog's ability to learn and obey commands. It must be admitted that this measure has validity in the sense that it can be measured objectively. This will also come as a relief to perplexed owners of the low ranked terriers, since they know that their bright companions understand commands perfectly, but are just too stubborn to obey them mindlessly. In fact, the book does raise this point: Even though terriers are extremely bright, they are bred to be independent and thus "do not care about human responses to their behaviors" while dogs bred to work with man like the herding dogs, do extremely well in obedience because of the genetic wiring that compels them to look to their human owners for direction. This excerpt from the book is taken from an excellent summary and review written by KrisHur on the newsgroup rec.pets.dogs.behavior.
The rankings do give a first impression of accuracy, since the top dog is the Border Collie, which everyone agrees is the smartest breed, no matter how you figure it. However, it is my belief that most of the list is completely meaningless because of the author's misguided attempt to rank a large number of breeds. The large number of breeds tested implies that the number of subjects in each breed would be rather limited, invalidating any possibility of statistical significance of the results. With 79 breeds represented, a sample of 100 dogs of each breed would require many thousands of careful tests, which seems like an improbably large number. Testing problems would be compounded by the difficulty in locating a hundred dogs of the same breed with owners willing to let them undergo a lengthy testing protocol.
As definitive proof of the shoddiness of the rankings, I note that the Norwich Terrier is listed in 38th position, while the Norfolk Terrier is given in 56th position. As far as I can tell, the only difference between these two breeds is the shape of their ears. Moreover, the American Kennel Club describes the Norfolk Terrier as being one of the most intelligent terriers, whereas no such mention is made for the Norwich Terrier.
This obvious error is exactly where the rankings go wrong. With insufficient data, it is impossible to make find statistically significant differences between closely related breeds. On the contrary, my anectodal experience suggests that a carefully done statistical analysis would find no significant difference between breeds in the same major groups, that is, hound, terrier, herding, etc. In other words, I conjecture that the variation among individuals in a single breed could be on the same order as differences among breeds in the same major group. However, it is clear that there are significant behavioural differences between the major groups, which leads one to believe that Coren's results could have some validity if limited to these more general classes of dog.
Well, I am not at all surprised to find such a glaring error in this report of scientific findings. It is fairly well known that research and pets do not make a good mix in academia, and that doing research on pets can mean the termination of your career. The upside of such research is that you can become a popular author and even appear on television, as Stanley has, but even this has been known to be damage one's academic reputation. For this reason, Coren's study has remained unchallenged, and without serious criticism.
In conclusion, I believe that the dog intelligence ranking as it now stands has little validity. On the other hand, I also believe that there could be some merit to these results, if the author revised his list by limiting the categories to the major groups: hound, terrier, herding, etc., while making sure that the sample sizes were large enough to ensure some statistical significance.
Back to ilanpi