Monday, February 29, 2016

Drumming Ducks and Dancing Chickens

This little video extract is from Patient like the Chipmunks by Bob Bailey & Marian Breland-Bailey, and it details the "IQ Zoo" created by Keller and Marian Breland, which expanded on the work of B.F. Skinner. Skinner, of course, was just expanding on the work  of Edward Thorndike.

I wrote about the Bob Bailey and the Brelands some years back in a blog post that eventually became a magazine piece.

This article was written for the June, 2010 issue of Dogs Today magazine. Illustration by Kevin Brockbank.

Who invented animal training and what can we learn from them?

The question may seem silly. Animal training is older than the hills. For certain, it is as old as the dog.

That said, most of what we call animal training today fits into the folder of "operant conditioning," a term first coined by American psychologist B.F. Skinner in the early 1930s.

Chase it around the room, and operant conditioning is simply learning from consequences.

Skinner codified the basic principals of operant conditioning, and he invented a laboratory-based mechanical-training machine which rewarded animals with food when they pulled levers and pecked at spots.

Skinner's real claim to fame in the world of dogs, however, is that he hired Keller and Marian Breland -- two young assistants who more or less invented the modern art of animal training.

Teaching Pigeons to Bowl

The Breland's made their first big discovery while working with Skinner in 1943.

As odd as it sounds, the goal on this occasion was to train pigeons to "bowl" a ping-pong ball down a short alley to knock down a few pins. Keller Breland decided to use a hand-held switch to trigger a food reward, rather than a purely mechanical device. A small problem was that the pigeons had no interest in pecking at the ping-pong ball! Though Keller waited for hours for the right behavior to express itself, it never happened. In frustration, Keller decided to give the pigeons a reward for doing anything approximating movement in the right direction. To his amazement, the pigeons caught on pretty quickly, and the completed trick was learned in short order.

Thus was born "shaping," or progressive rewards based on approximating a task.

The second big development occurred in 1945. By this time, B. F. Skinner had left the University of Minnesota, and the Brelands had decided to strike out on their own as professional animal trainers.

While shaping tricks, the Brelands noticed that the animals themselves seemed to be paying attention to the noises made by the hand-held food-reward switches.

Keller and Marian Breland soon discovered that an acoustic secondary enforcer, such as a click or whistle, could communicate to an animal the precise action being rewarded, and it could do so from a distance.

The Brelands called this a "bridging stimulus," (now generally called a bridge) and it dramatically sped up animal training by increasing the amount of information going to an animal.

Thus was born clicker training.

The Rise of Commercial Animal Acts

By now the Brelands had created their own animal training company -- Animal Behavior Enterprises. Their first contract, with General Mills, was so successful that other contracts with movies, circuses, museums, fairs, and zoos soon followed.

Over the next several decades, the Brelands trained more than 15,000 animals representing more than 140 species. At one point, the Brelands had more than 1,000 animals under training at a single time -- a jaw-dropping level of production.

The Brelands did not just train animals; they also trained other animal trainers who went on to work at such venues as Busch Gardens, Disney World, and Sea World. The Brelands themselves signed contracts with such major amusement parks as Marineland of Florida, Marineland of the Pacific, Parrot Jungle, and Six Flags.

Rewards-based clicker training worked so well that in 1951, the Brelands authored an article in American Psychologist, in which they said they thought rewards-based clicker training might work on any animal to train just about anything.

And then something happened. They noticed that clicker training was, in certain circumstances, beginning to fail in ways they could no longer overlook.

When Clicker Training Failed

In a 1961 paper entitled, The Misbehavior of Organisms, Keller and Marian Breland described their first experience with the failure of reward-based operant conditioning.

It seems that when working with pigs, chickens and raccoons, the animals would often learn a trick and then begin to drift away from the learned behavior and towards more instinctive, unreinforced, foraging actions.

What was going on?

Put simply, instinct was raising its inconvenient head.

Though Skinner and his disciples had always maintained that performance was driven by external rewards or punishments, here was clear evidence that there was an internal code that could not always be ignored.

The Brelands wrote:

These egregious failures came as a rather considerable shock to us, for there was nothing in our background in behaviorism to prepare us for such gross inabilities to predict and control the behavior of animals with which we had been working for years.... [T]he diagnosis of theory failure does not depend on subtle statistical interpretations or on semantic legerdemain - the animal simply does not do what he has been conditioned to do.

The Brelands did not overstate the problem, nor did they quantify it. They simply stated a fact: instinct existed, and sometimes it bubbled up and over-rode trained behaviors.

Clearly, every species had different instincts, and just as clearly, a great deal of animal training could be done without ever triggering overpowering instinct. Still, the Brelands noted,

After 14 years of continuous conditioning and observation of thousands of animals, it is our reluctant conclusion that the behavior of any species cannot be adequately understood, predicted, or controlled without knowledge of its instinctive patterns, evolutionary history, and ecological niche.

The Problem with Dogs

What does this have to do with dogs?

Quite a lot. You see a small, but vocal and militant group of clicker trainers believe everything a dog does is learned by external rewards, and internal drives are "old school" fiction.

While the Brelands argued that a species could not be adequately controlled without “knowledge of its instinctive patterns, evolutionary history, and ecological niche," the most extreme militants in the world of clicker training now seek to minimize and disavow the very nature and history of dogs.

Dog packs? There are no such things, we are told.

Dominance? It does not exist in feral dogs or in wolves, and never mind the experts who disagree.

Prey drive? Not too much said about that!

Of course, instinctive behaviors and drives do not disappear simply because they are inconvenient.

As Keller and Marian Breland put it,

[A]lthough it was easy to banish the Instinctivists from the science during the Behavioristic Revolution, it was not possible to banish instinct so easily.

Of course, one must be careful to qualify the role of instinct.

Yes, dogs have instincts, but the history of dog breeding has largely been about reducing instinctive drives. As a consequence, most breeds have instinctive drives that are sufficiently attenuated that they are not much of an impediment to basic rewards-based training.

That said, not all dog breeds are alike. Not every dog is a blank slate, as the owner of any herding dog or game-bred terrier will tell you. Prey drive does not disappear because you want it to. Many problematic behaviors in dogs -- especially behaviors in hard-wired working dogs that are being raised as pets -- are self-reinforcing behaviors that express themselves without any external reinforcement at all!

Clicker training, the Brelands remind us, cannot solve everything.

Is rewards-based training the most important tool in any trainer’s box of tricks and methods?

Absolutely. There is not much debate there.

But the Brelands remind us that dogs do not come to the trainer as a tabula rasa, nor should we think of all dog breeds as being more or less the same, or that all responses are equally conditionable to all stimuli.

Dogs and other animals, it turns out, are a bit more complicated that white rats, and the real world is not a laboratory.

In the wild and on the farm, animals have managed to learn, all by themselves, since the Dawn of Time and long before clickers came on the scene. How did they do that? Does the real world have as much to teach us as the lab? Keller and Marian Breland thought it did.

No comments: