Definition of a Superstitious Behavior: Accidentally or unintentionally reinforced behavior where a behavior is reinforced but the reinforcement occurred by random chance instead of in accordance with a specific contingency. —From the Association of Animal Behavior Professionals glossary
When I started this blog I wrote that I was going to share my mistakes with the goal of helping others learn. Here are some nice big embarrassing mistakes, but at least they date mostly from my earlier training days.
When one of my agility buddies encouraged me in 2008 to start using a clicker, I didn’t know that I should practice timing. I didn’t know that there were mechanical and observational skills involved. A clicker seemed like a fun thing and I had heard that dogs got motivated and enjoyed it. So I also didn’t know that it would be wise to start with a behavior that involved only gross motor movements.
I should have picked something that would be easy for me to perceive and mark. What I did was to get a really loud box clicker and decided right away that my first project would be to click Summer for eye contact.
The very first behavior I actually got with a clicker was a head nod. I realized this after a couple of weeks of practice. But even though Summer did learn (despite me) that I was trying to teach eye contact, the head nod remained. Strangely, it drifted to the period after the click and before treat delivery. The sequence went: eye contact, click, head nod, treat. The nod, immediately preceding the food, accordingly got a ton of reinforcement.
Four years later, I still get little nods from Summer. Interestingly, she doesn’t offer it in shaping sessions. When it comes back, it is in its old place between the marker and the treat when I have just clicked her for something else.
It has faded some over the years but I found a couple of examples. Want to see?
This is an example of a superstitious behavior. And it turns out that I am really good at creating those!
B.F. Skinner first described superstitious behaviors in experiments with pigeons in 1948. He set a feeding mechanism to trip at variable intervals that had nothing to do with the actions of the pigeons. The pigeons nonetheless started repeating behaviors that had been “accidentally” marked and reinforced by the feeder.
Richard Dawkins explains it here in a short clip. There was actually some subsequent debate about Skinner’s interpretation of what was happening. But the term “superstitious behavior” remained and is now used informally to refer to any behavior that is accidentally reinforced. A couple of the behaviors in this post stretch the definition. But even if they aren’t technically superstitious, they are nonetheless accidental or at least poorly trained on my part.
The following is a behavior that would have been very difficult to teach, had I intended to do so.
When teaching Zani agility weaves using the two by two method, I would tend to mark with a “yes” the moment she committed to the last weave. This was the moment I was absolutely sure she was going to complete the behavior correctly. That’s a natural time to mark. Except in the beginning, she did a few little jumps through the last pair of poles. I marked, and you can see what happened.
But what is the most fascinating is that she only does it when I am on her right side. When I am on her left and she does an “off-side” entry, she doesn’t do her thing between the last poles. I speculate that since she is very spatially sensitive, she is less likely to go hurling out of the weaves when I am over there in the area where she will emerge. Or perhaps I just didn’t mark the exit as much when we practiced on that side.
These first two are pretty cute. This upcoming behavior of Clara’s that I accidentally reinforced is rather unfortunate.
Clara has always been pushy. When she was about three months old I started a training project of reinforcing her for backing off when I was interacting with another dog. I started off with the other dogs in crates and was very systematic about it. I drew lines on the floor for my own benefit so as to keep consistent criteria. We did lots and lots of sessions of this.
It would have been better if I had taught a default down or Go to Mat, or at least thrown the treat away from our immediate area. What I ended up doing was unintentionally reinforcing a circling behavior. She would walk a few steps away, turn and reorient at the desired distance. I marked the turn (way too often) and treated her when she faced us again. What I wanted for her was just to back off. But that is not a well-defined behavior, so I ended up clicking a definable moment, and that was when she turned back to me. Dang.
Unfortunately, the circling either morphed into a stress behavior or it was one already. But I have seen a lot more of it ever since those sessions. She tends to do it when I do not mark a behavior that she expects to be marked. She will immediately whirl around (usually counterclockwise), then often retreat to a mat.
It is impossible to tease apart how much of this is due to it becoming almost a default behavior because of all that early reinforcement, and how much is that it is a natural stress behavior for her. I do wish I hadn’t trained so many 180 and 270 turns when she was young. When I set out to teach her spinning as a trick, it was dead easy. I gave that a second thought and decided not to use that trick.
I even taught Cricket a superstitious behavior. I tried to train a paw lift as a trick (wave), but then she started lifting the paw in her “sit” position. Once it started, I kept accidentally reinforcing it. I didn’t know anything about stimulus control and not much about cues so I had no idea how to get rid of it.
The way I taught the behavior in the first place was not great either. A friend had suggested holding a treat in my hand and clicking her for pawing at it, then fading the hand and treat. Such a bad idea in so many ways. Teaching an enthusiastic digging terrier to paw at my hand? Great! And it worked. It worked so well that she almost never put her left front foot down when she sat again.
How To Avoid Training Superstitious Behaviors
I wish I could give some succinct, pithy advice that could help other newish trainers not to do this. I think it takes experience to learn to predict the ramifications of teaching a behavior. I didn’t have a teacher to ask at the time of most of these (Clara’s behaviors being an exception). Even when you do have a teacher, you don’t always think to ask about everything. But here are the questions I can think of to ask oneself.
- Is there a persistent side effect that is happening when I train this behavior?
- What’s going to happen if it sticks around?
- How can I get rid of it?
- This trick I am deciding to train—what if it becomes a really strong behavior? What might it interfere with?
If you don’t have a teacher, you can learn a lot by videotaping your training sessions, and if you are brave, showing those recordings to online friends if you don’t have a teacher or local training buddy. People can give much better counsel if they actually see what you and the animal have been doing. Most of us humans could use a lot of work on our observation and description skills. Cameras do a lot better job for a lot of us.
One behavior that I got right: I started training that “backing up the stairs or wall” trick that was going around a while back. Zani just loved it and started getting good at it. It was great for hind end awareness.
But then one day when we were practicing our two on, two off agility contacts, she overran them and happily backed up into position. That would be a fault in many agility venues. I immediately stopped training the trick. A more experienced or patient trainer could certainly have both behaviors, but sometimes I realize my limitations. The risk wasn’t worth it to me.
Here is one more of Clara’s. This is an example of something that is cute when a puppy does it, but can get pretty tiresome in a grown dog. Of course it’s still cute, but who wants their fingers licked Every. Single. Time. They go to open a crate door?
OK folks, please tell me I’m not the only one who trains silly behaviors by accident. Does anybody want to say what they have done? Or are you all perfect?
Discussions coming soon:
- More dogs enjoying petting
- Comparing licking and tongue flick behaviors
- “Errorless learning”
- Canine Cognitive Dysfunction
Copyright 2012 Eileen Anderson