Clever Hans Revisited
You probably know the story of Clever Hans, the horse owned by a math teacher named von Osten who decided to teach his horse to do math in the same way that he did his pupils. After extensive training, Clever Hans appeared able to solve relatively advanced mathematical problems, including multiplication and long division. Clever Hans showcased his abilities around Europe, although von Osten never charged for an exhibition. His owner and trainer sincerely believed that his horse understood what he was being asked, and wanted the world to see it for themselves.
Scientists were so interested that a panel was formed, led by psychologist Carl Stumpf, which verified that no tricks were visibly involved, but passed the issue onto psychologist Oskar Pfungst. After an extensive series of tests, Pfungst found that Clever Hans was unconsciously being cued by his trainer. Hans could only answer questions if in visual range of a human who knew the answer. In other words, Pfungst found that Clever Hans was clever indeed, but in a different way than thought by his owner. The horse used subtle cues from humans (head tilt, eyebrow raise) to know when the correct number was coming, and thus when to stop pawing. (Hans communicated by pawing the ground; his answer to 2 + 2 was to paw the ground 4 times.)
I’m reprising this story, familiar to most of you, because of a great talk given by Dan Estep and Suzanne Hetts of Animal Behavior Associates at the Interdisciplinary Forum on Applied Animal Behavior in Arizona last weekend. The owner of a dog named Sheba had asked for a “scientific investigation” into his dog’s intelligence, and a local TV station asked Dan and Suzanne to look into it.
Sheba’s owner, Bob, was convinced that his dog was brilliant. Not smart, but brilliant. She knew just about everything about anything, and what’s more, she’d acquired this information all on her own; the owner swore he’d never taught her a thing except how to communicate. Sheba could answer yes/no questions with one paw or two, and multiple choice questions by pawing 1 to 4 times. Is it cold in the Arctic? Yup, answered Sheba, slapping her paw once onto her owner’s palm. Did the Green Bay Packer’s win the Super Bowl this year? Of course not, everyone knows that: Here’s two paw slaps for a resounding no.
Dan played the video tape of Sheba’s performance and we all watched, fascinated, while Sheba accurately answered question after question when asked by her owner. And then, predictably, it all fell apart when Sheba was asked to answer questions when her owner didn’t know the answer or she couldn’t see him.
But, of course, Sheba WAS brilliant, just brilliant at readings subtle cues that her owner wasn’t aware of. Imagine us at the meeting – a gloriously eclectic group made up of Certified Applied Animal Behaviorists, Board Certified Veterinary Behaviorists and some trainers brilliant in their own right – all straining forward, watching the video over and over, trying to figure out what cues Sheba was using to figure out the correct answer. We never did, and neither did Dan or Suzanne (although the best guess is a combination of tactile cues from his hand and visual cues from his face) because Bob refused them any chance to work with her some more. What a shame that instead of acknowledging Sheba’s intelligence, albeit not in the way he imagined, he sent them packing, insisting that they were wrong, that Sheba really did know that San Diego was south of San Fransisco. He also eventually admitted to them that Sheba was actually Albert Einstein reincarnated. Oh my.
There’s more to this than an amusing story. Dan used this case to remind us that the real Clever Hans taught us a lot more than “be aware that you might be cuing an animal in subtle ways.” Pfungst discovered not just that Clever Hans was astoundingly good at reading visual cues, but that it was almost impossible not to produce them. Once he figured out the cues that Hans was reading, he found that when he or others consciously tried their hardest to avoid creating them, they were unable to do so. In other words, even if you tried your hardest to stay absolutely still in every way, it was impossible NOT to cue Clever Hans if you knew the answer and he could see you. Wow. Think about that in relation to you and your own dog. (By the way, Pfungst’s book, Clever Hans, is available for free. I highly recommend it, it’s fascinating.)
As Sheba reminds us, it’s not just horses that read us like a book. Note a study by Lisa Lit that found that dog/handler scent detection teams reported finding scents 260 times (18 teams, each run 6 times) in areas in which there were NO scents planted. But the handlers had been told that scents had been planted, and that one room even had a red symbol marking its location. Mark Hines, who works with scent detection dogs all over the world (he works for Kong, encouraging the use of positive reinforcement in the training of military, detection and protection dogs; I call him The King of Kong) was in attendance at the meetings and said this was a common problem in detection dogs. Really good handlers and trainers are well aware of the problem, but it’s more extensive than you might think.
Those of you who studied psychology might be remembering the study by Rosenthal & Fode in 1963 in which experimenters were told that some rats were “Maze Bright” and others “Maze Dull.” That wasn’t true, the rats were actually all the same. And of course, when the rats ran the maze, the ones believed to be “bright” really did run the maze faster. (It turns out the experimenters handled and interacted with the “bright” ones more often.)
Now… think about your own dog, his or her behavior and your own expectations. How many times do you think you are unconsciously cuing your dog? Eeeeps, the mind boggles. How many tiny pupil dilations, head bobs or changes in scent do we make every day that communicate with our dog? How many times do our beliefs about our dogs and our expectations of their behavior influence it?
I’d love to hear examples from you of when you think this might be relevant in your life. I’ll tell you when I think I most have to be aware of it: knowing Willie’s history with unfamiliar dogs, I have to be extra careful not to set him up to be tense during greetings. Perhaps this is why I’ve seen clients (and myself) have the best luck with classically counter conditioning both the handler and the dog to have a different response. Replacing an action on our part (rather than trying to just stand still like Pfungst did) with another behavior like signaling “Watch” or doing BAT or “Look at Me” changes our behavior as much as our dog’s.
MEANWHILE, back on the farm: I wrote most of this on Sunday afternoon on a plane from Denver to Madison on my way home from IFAAB, but now it’s Monday morning and it’s lovely to be home. Especially since I leave again on Friday for the Tucson Book Festival. But as much as I love home, IFAAB was wonderful this year: interesting, interactive, inspiring and supportive, sort of an academic slumber party. I’ll write more about some of the other interesting talks I heard there as the weeks go on.
Here’s a little contrast for you: First, the colors in Arizona on Sunday:
And here’s the colors here in Wisconsin: