German psychologist Hermann Ebbinghaus (1885) came up with the concept of a “learning curve” while studying the effects of repeated performances of a task and its effect on memory. Generally, the more an activity is practiced, the more likely it will be remembered and repeated correctly. Thus, you could plot the increased performance on a Cartesian graph, with one plane, call it the X axis representing the number of learning repetitions, and the other plane, call it the Y axis, representing the correctly performed repetitions, and a line that slopes upwards between the two axes representing the improved performance of that skill over time. Similarly, Ebbinghaus postulated a “forgetting curve”, which showed the opposite effect, which shows the down sloping curve, representing the deterioration of the skill when practice stops over time.
How has the learning curve been used for training animals? In B. F. Skinner’s famous book, The Behavior of Organisms (1938), he uses learning curves to display the results of his experiments on rats. His learning curves generally smoothly slope upwards as the rat is reinforced for pressing a bar to get a kernel of food.
OK, fine. So, what is the big deal?
Well, the learning curves he used were aggregate curves for a population of rats. However, when you do more digging on this effect, you see that the curves for individual rats were anything but smooth. Instead, they were jagged.
This phenomenon didn’t go unnoticed, either by Skinner or scientists. It wasn’t that Skinner was cheating. But he did feel the need to create smooth curves to demonstrate his theory of operant conditioning, without explaining why he had to aggregate the data. (Pavlov did the same thing; to his credit, Pavlov was ultimately using his investigations to predict higher nervous system functions.)
Now comes the interesting part… at least to me. If the theory is accurate, then the learning curve for the individual should be as smooth as the learning curve for the group. But it isn’t. The problem goes back to Skinner’s premise that individuals can be controlled using operant conditioning. If that was true, then every reinforcing repetition should have resulted in a perfectly smooth upward curve of performance. That meant something else was going on.
Before I moved to Phoenix, AZ, I lived in the Seattle, WA area and, as a hobby, I trained at a Brazilian Jiu-Jitsu martial arts school. About the time I was getting ready to move to Phoenix, I had been training there 3 times a week, 3 hours per session, for about 2 ½ years and had attained a Blue Belt with one white stripe. But I was getting frustrated. I asked the Professor (the head teacher) a question: “Why is it that sometimes I come in here and do well, but at other times I come in and I think I’m no better than a beginning white belt?” His answer was, “In jiu-jitsu, sometimes you are the windshield, sometimes you are the bug. It takes many years to become a black belt, you are doing fine. If you are concerned about your progress, look at your progress in 4-month intervals.” My progress wasn’t a smooth learning curve. It was a jagged curve.
I find this same phenomenon when studying dog behavior. When I read Skinner’s book, it didn’t go smoothly. Even B. F. Skinner couldn’t teach his readers in such a way that every word, every paragraph, and every chapter, would smoothly lead to a complete comprehension of his theory. Regardless of the author, for me, there are always detours along the way. Sometimes I must re-read sentences or chapters. Sometimes I must consult with others for their feedback. Sometimes I must try out the concepts and see how they work in the real world. Sometimes I must put a book aside, and read a supplementary study, to understand how a theory was developed, or how others have attempted to test those theories. And many times, I have to put the book aside and let my brain sleep on what I have learned before trying to go at it again.
How does this apply to dog training? Well, whenever I build any dog with lessons, there is never a smooth learning curve. Some days the dog is better, some days there are setbacks, or sometimes we need to take a detour. If we looked at the learning curve for that dog from 10 miles up, the curve would look like it was smoothly sloping upwards, up close the situation is much different. It would be an individualized jagged line sloping upwards over time.
It is important for me to let owners know how this works. First, they need a good understanding that without doing their homework, there will be little progress. Second, they need a clear understanding of a correct repetition. And lastly, that their dogs are not machines and they will output an individualized jagged line of progress.
Dogs have a brain that mediates all this input, and it has effects on the output. The dogs brain has a say in all of our training since dogs are not just machines. The promises of “stimulus control” by theorists like Skinner are never going to be realized. Skinner had the utopian idea that with sufficient reinforcement of correct responses, that all human (and animal) behavior could be controlled. He pursued a path so that all life could be controlled like a machine. He even advocated for a new form of totalitarian government to do just that! He was blinded by his biases and his approach. Followers have taken up this mantle and promised the same, on the micro and macro levels. I saw this early on when the clicker training fad was coming on. Prior to that, the force trainers made the same claims. But in the real world, with real dogs, it didn’t work that way. Dogs had their own minds. No system should promise perfect progress or results.
As a novice, I figured I was doing something wrong. I believed all these promises, thinking if I knew what these gurus knew, I could pump out perfect dogs. But over the years, I saw that even they weren’t producing perfect dogs, even with their own pets. I don’t know if they were just blind or selling us all a lie. Looking back, I think it was a combination of both. Trainers still market themselves as gurus. It’s either ignorance or dishonesty. And believe me, even the gurus have experienced Impostor Syndrome (even if they won’t admit it). The “purely positive” cult is wrong. And a lot of other approaches are wrong. A dog has a brain and you can’t defeat that with some warped and blind theoretical system.
No dog is going to make perfect progress every day or turn out to be the perfect dog. Some days will be better, some days will be worse, and some days will require detours to get back on track. And that is also true if you stop training. Sometimes you’ll be the windshield, sometimes the bug, and your dog won’t be perfect. But if your approach is valid because of good instruction, and you give your dog proper time and effort, the learning curve will slope upwards… limited only by your correct efforts combined with your dog’s biological talent. If you quit on your dog whenever you see a plateau, or must do a detour, or must invest more time and effort, then you misunderstand what it takes to maximize your dog. It is never going to look like a smooth curve. There is more to learning than Classical and Instrumental Conditioning. Your dog is not a machine. Your dog has a brain and will have a say regarding the rate of progress. Modern theorists have rejected the original promises of the mechanical input/ output theories, and you should, too. Life is jagged. Get used to it.