There is a dog I see at agility trials frequently who does something adorable. As she exits the weave poles, she always does what looks like a little leap of joy by hopping up a little as she continues to the next obstacle. If you didn’t know this dog’s history, you might think that it is a wonderful indicator of her joy at playing agility with her handler. The truth behind this behaviour is that the dog was rewarded during training by having her handler throw a treat bag ahead of her as she finished the weaves correctly. During that training, the dog would try to leap up to try to catch the threat bag in the air rather than wait for it to land. Over time, she came to believe that the leaping up was part of what was being rewarded. That was the pattern she recognized so she repeated it to keep the rewards coming.
Patterns are important. Patterns provide structure and order to our dogs’ lives. Everything from being let out at certain times of the day to meal preparation rituals become important indicators to make life predictable and easy to navigate. Some of these patterns evolve organically, we do not set them up intentionally. Like choosing the same shoes to go walking at the park. Our dogs will soon come to recognize the pattern that the putting on of those special shoes means a trip to the park. It might not be surprising to learn that this tendency to recognize patterns, or “Patternicity” as Michael Shermer calls it, might actually be a biological survival mechanism.
Michael Shermer defines Patternicity as “the tendency to find meaningful patterns in both meaningful and meaningless noise.” Shermer suggests that animals constantly look for patterns and that this pattern seeking is very much a part of both Pavlov’s Classical Conditioning and Skinner’s Operant Conditioning. In Classical or Respondent Conditioning the dog will recognize the pattern that a particular sound or signal always precedes a particular thing (e.g., hearing “Sit!” signals an opportunity to earn a reward for sitting or that being in a certain place means that scary noises will happen). In Operant Conditioning, a dog learns that if he performs a particular behaviour, a predictable result will happen (e.g., leaning against the owner’s leg will get them some petting or that barking out the window will earn them a scolding from the owner).
This reflex to find patterns is so strong that animals will find patterns even where none exist! Skinner showed this in experiments with pigeons where the birds were randomly rewarded for pecking one of two keys. Researchers were instructed not to be consistent in when they rewarded the bird over several trials. Interestingly, the pigeons developed all kinds of behaviour rituals that they believed would make the reward more likely. This is called “superstitious learning” and is actually one of two errors that can happen in Shermer’s “Patternicity.”
Shermer describes a Type I error in Patternicity as a False Positive – believing a pattern is real when no such pattern actually exists. This is what the pigeons in Skinner’s experiment showed by performing their behaviour rituals, their superstitious learning. A Type II error in Patternicity is a False Negative – a failure to recognize a pattern when one actually exists. And this is where it gets interesting. There is actually a solid biological reason why animals more likely to make Type I mistakes than Type II mistakes. Let’s call it survival arithmetic. If an animal hears a noise in the bushes nearby, it might believe that the noise is a predator about to attack and they will flee. If they are wrong, a Type I false positive, the worst that happens is they got some unnecessary exercise and lose a little time. But if they commit a Type II error and disbelieve that there is a predator when there really is one, well, they are dead! From a mathematical perspective, Type I errors are less costly than Type II errors. Put more simply – “Better safe than sorry!”
Here’s an interesting statistic: the greatest batters in professional baseball only hit the ball 35-43% of the time. And here’s something else that is interesting. Baseball players are among the most superstitious in all of sport. Science has traced this tendency of baseball players to develop superstitious rituals to improve their hitting percentage to a lack of control and consistency while batting. In other words, the less control the player feels they have, the more likely they are to develop superstitious rituals to improve their odds. The less predictable the situation, the more Type I errors in Patternicity occur.
Another interesting correlation has been found with the amounts of dopamine in the brain. Dopamine is a neurotransmitter involved in the brain’s reward system and we associate it with feeling “good.” Studies have found that subjects that have high levels of dopamine will make more Type I Patternicity errors and that subjects with too little dopamine will make more Type II errors. What that means for our dogs is that if they are too excited or happy, we may have trouble teaching them because they are seeing patterns that aren’t there, they aren’t getting the lesson that we are trying to teach. In cases where dogs are too anxious or stressed and have too little dopamine, they will make more Type II errors and not recognize the lesson, the pattern, we are trying to teach them!
We are not immune
It would be unwise to think that, as humans, we are immune to Patternicity. If anything, we are more likely to fall victim the Type I and Type II errors that Shermer describes just because we are more complex in our thinking. There are just more patterns for us to see. But that also means that there are more mistakes for us to make. And it goes beyond obvious errors like thinking our dog has learned what we are teaching because we think we see a pattern where there isn’t one.
Our human ability to perceive patterns can be profoundly influenced by the suggestion that patterns exist. If we shown a pattern in something, we are more likely to see it again in similar circumstances even if that pattern doesn’t exist! We will often see the patterns that we most want to see or are most familiar with even if they aren’t really there. These are Type I errors and the behavioural arithmetic says there is less risk for believing than not believing if we are wrong. And this can lead to another perception problem that Shermer calls “Agenticity.”
Agenticity is defined as infusing recognized patterns with an intentional agent – someone or something is doing it on purpose. In the case of working with our dogs, we see this when we are told that the dog is “blowing off” the handler or that the dog is just upset over something that happened recently and is acting out. The interesting thing about Agenticity as it applies to the human-dog relationship is that we apply our own human emotions and motivations to a dog who may not have those capabilities!
Patterns, agents, and being better trainers
Patternicity can be a problem in working with dogs but, ironically, it is also a part of the mechanism that makes learning possible when we train with our dogs. We are deliberately trying to create patterns for them to recognize so that we can teach them behaviours and then the cues we want to use to get them to do those behaviours in future. Having an understanding of how Patternicity works can help us be better trainers. It shows us clearly that not providing enough structure to the patterns we use to train our dogs can lead to superstitious learning and then we have to “fix” the problems that our sloppy training has caused. Not having a good understanding of how Patternicity can effect us as trainers can lead to believing that our dogs “know” things or have learned things that aren’t really that well trained yet. Then Agenticity allows us to blame our sloppy training on the dog by believing that the dog is intentionally offering us an incorrect pattern for some imagined emotional reason.
We want to believe. That much is clear. Our dogs want to believe too. We all seek order and predictability and, if we don’t see it clearly, we can fool ourselves or allow ourselves to be fooled by others into believing in things that may not be there. One of the things I do with my training is to keep written records or video of training sessions. This allows me to review the patterns I think I observe more than once to be sure that I am not falling victim to the Type I or Type II errors that Michael Shermer describes. It also allows me to show my result to my fellow trainers and get their input to help me keep a good perspective on what is really happening.
I think honesty is important in our relationships with dogs. And that means not fooling ourselves about what we are seeing and not allowing our dogs to fool themselves with superstitious learning because we haven’t been clear enough with them. The more we know about ourselves and about our dogs, the better we will be as trainers and caretakers.
Until next time, have fun with your dogs!
The NEW Canine Nation ebook is now available –
“Relationships: Life with Dogs”
Photo credits –