Inside of a Dog_ What Dogs See, Smell, and Know - Alexandra Horowitz [71]
To verify this anecdotal impression, some have designed so-called intelligence tests for dogs. We're all familiar with intelligence tests for humans: pen-and-paper creations that require you to solve SAT-like problems of word choice, spatial relationships, and reasoning. There are questions that test your memory, your vocabulary, your declining math skills, and your simple pattern-finding ability and attention to detail. Even putting aside whether the result is a fair assessment of intelligence, the design does not translate obviously to testing dogs. So revisions are made. Instead of tests of advanced vocabulary, there are tests of simple command recognition. Instead of repeating a list of digits read aloud, a dog may be asked to remember where a treat was hidden. Willingness to learn a new trick may replace the ability to figure complex sums. Questions loosely mimic experimental psychology paradigms: of object permanence (if a cup is placed over a treat, is it still there?), learning (does your dog realize what foolish trick you desire him to do?), and problem solving (how can he get his mouth on that food you've got?).
Formal studies of groups of dogs on these kinds of abilities—mostly cognition about physical objects and the environment—yield what at first seem to be unsurprising results. By bringing dogs to a field baited with treats and timing dogs' speed in finding them, researchers have confirmed that dogs use landmarks to navigate and find shortcuts. This behavior is consistent with what their wolflike ancestors would probably have done in finding food and finding their way. Dogs are, of course, pretty good at all tasks that involve getting themselves to food. Given a choice of two piles of food, dogs have no trouble choosing the larger one—especially as the contrast between them grows. Turn a cup over a bit of food and dogs go right for it, knocking the cup and revealing the treat. Dog subjects have even learned how to use a simple tool—pulling a string—to get an attached biscuit that was otherwise out of reach.
But dogs don't pass all the tests. They typically make lots of mistakes when presented with piles of three versus four biscuits, or of five and seven: they choose the smaller amounts just as often as the larger. And they develop preferences for piles on the left or the right, which lead them to make even more blatant errors. Similarly, their skill at finding hidden food gets worse as the hiding gets more complicated. And their tool use also starts to look less impressive as the trials get trickier. When there are two strings, and only the more distant one is attached to an alluring biscuit, dogs nonetheless go for the nearer string, the one attached to nothing. They don't seem to understand the string as a tool: as a means to an end. Indeed, they may have succeeded in the original case simply by pawing and mouthing at the problem until accidentally solving it.
A dog owner tallying her dog's score in these dog intelligence tests might find that he's scoring closer to Dim but happy than Top of the obedience class. Is that it, then? Is he not smart after all?
A closer look at the intelligence tests and the psychological experiments reveals a flaw: they are unintentionally rigged against dogs. The flaw is in the experimental method, not in the experimented dog. It has to do with the very presence of people—experimenters or owners. Let's look more closely at a typical experimental setup. It might begin as follows: A dog is sitting at attention and being restrained by a leash. An experimenter comes before him and shows him a great new toy. This dog loves new toys.* The toy and a bucket are clearly shown to the dog, the toy is put into the bucket, and then the experimenter disappears with the booty behind one of two screens in the room. She returns with the bucket—emptied of its treat. This turns out not to be a cruel hoax, but a standard