There is a very in-depth article about Errorless Learning and No Reward Markers making it’s way around Facebook recently. The article, by internationally renowned dog trainer Emily Larlham, was brought to my attention by a training colleague. There’s a lot of information in Ms. Larlham’s article but, for me, it boils down to a deceptively simple question: How do we know when our dog is “doing it wrong”?
The immediate answer would seem to be that the dog is not doing the behaviour as requested (cued, commanded, choose your word). On the face of it, it seems a pretty easy thing to determine. But if we dig a little deeper, we could ask why our dog isn’t performing as we expected. And this is where it gets interesting. There are a few reasons why our dog is not responding as we would like and I’ve written a series of articles about that. You can find them here in the archives. So it would seem that we need to know what an “error” is before we can tackle what “Errorless” learning might be.
Wrong is wrong, right?
When we ask our dog for a given behaviour, we are making a few assumptions. First, we are assuming that they have a clear understanding of what behaviour we are expecting. And that’s not always simple. When we ask our dog to “Sit”, we know that we expect them to put their rump on the floor. But that might not be the end of our expectations for that behaviour. Do we expect the dog to stay in that position until released? Do we expect them to sit straight (not rolled on to one hip)? Are they supposed to come to our front and then sit? Can they take their time and settle into it or do we expect them to drop their bum immediately? Before expecting a “correct” response from our dog, it is only fair that we make sure that we have taught them what is expected.
The next assumption we make is that they actually recognize our request and know that it means to perform the behaviour we want. Humans talk a lot. Can we really expect our dog to respond to a cue every time they hear it? If so, it would be reasonable to assume that when we ask our guests to “sit down and have a cup of coffee” that our dog will recognize the “Sit” word and comply. But we don’t really expect that. We expect that our dogs will know when we are talking to them and when we are addressing others. While dogs are remarkably good at that, it may not be as easy for them to determine as we would like.
A third assumption we make is that our dog is willing to do as we ask. Whether it is out of fear of reprimand for not complying or in hopes of a reward for their efforts, we assume that our dog is motivated to do as we ask. As trainers, we use our preferred training techniques to instill that motivation to work with us. But the world is a distracting and interesting place. Sometimes my request is not my dog’s highest priority. Is my dog being stubborn or “blowing me off” or is she just not paying as much attention to me as I would like?
The Error in Errorless Learning
So my dog is wrong if she 1) doesn’t understand my cue 2) doesn’t know the behaviour the way I want it 3) didn’t hear or understand that my cue was meant for her 4) is distracted and does not make me her priority or 5) isn’t motivated to do what I ask. Interestingly, there is quite a bit I can do about these situations as my dog’s trainer. That’s what training is all about, after all. But what “errors” in performance boil down to is that my dog did not give me the response that I expected. Put simply, I didn’t get what I wanted.
At the risk of sounding like a Zen Buddhist thought-puzzle, is my dog wrong because she thinks she is wrong or because I say she is wrong? It’s an important distinction. At first glance, the term Errorless Learning suggested to me that you learn without making any errors. Strictly speaking, this is true. But when he introduced the procedure in 1963, Columbia University professor Herbert Terrace meant that there would be no indication to the subject when they were performing a behaviour incorrectly. So trainee might produce responses that didn’t meet the criteria for reward, but they were not responded to in any way for those responses. Terrace’s procedure was “errorless” in the sense that the trainee was never told they were “wrong”, they just weren’t told they were “right” either.
Terrace’s Errorless Learning protocol also attempts to maximize correct responses by setting up the trainee for success. This is done by starting with easy to achieve goals and gradually increasing the difficulty. If necessary, prompts are introduced help the learner succeed. If this sounds like Clicker Training or Mark and Reward training to you, give yourself a reward! That’s exactly what I was taught when I learned Clicker Training years ago.
Helping or not?
It’s been my experience that even the best intentioned indication of failure can be annoying at times. I’ve had coaches tell me “You’ve almost got it, try it again.” And I’ve given up in exasperation. I know the coach was trying to be encouraging but his repeated acknowledgement of my failure just took the wind out of my sails. As a coach, he was giving me information on what was not going to work. That can be helpful. But when it distracts or makes learning difficult, it just gets demotivating. And that’s what a No Reward Marker is – an indication that you didn’t get it right.
Like so many things in dog training, we can believe that our intention to be helpful will make something acceptable even if our dog is clearly showing us it is not. In many cases, I see trainers using various No Reward Markers because they believe it will help their dog arrive at the “correct” behaviour more quickly. And there is some logic to that – if you know what will not work, it narrows down your choices to find the correct choice more quickly. It certainly made sense to me and I started using No Reward Markers for a few weeks while training Tiramisu. The results were fascinating.
I introduced the marker word “Nope” said in a neutral tone when she offered a behaviour that was not what I wanted. The first thing that I found interesting is that Tiramisu was not learning new behaviours any more quickly or slowly than before using the No Reward Marker. Fortunately, I kept track of my training sessions during this trial period and an interesting pattern developed. Over the 12 weeks I used a No Reward Marker, Tira would be eager to start our training sessions as always, but her willingness to stick with it eroded over time. Our sessions without No Reward Markers would last anywhere from 10 to 15 minutes with eager responses and dozens of repetitions and Tira would still look disappointed when we ended our sessions. After 12 weeks of using the “Nope” No Reward Marker to assist in her learning, it was difficult to get a session to go more than 5 minutes and we always ended when Tira showed no desire to respond any more.
So was it an error because Tiramisu thought she did it wrong or because I told her it was by saying “Nope?” In the end, I believe I was just acknowledging that I wasn’t getting what I wanted. My data showed me that using that No Reward Marker didn’t improve Tira’s rate of learning and it decreased her willingness to work with me for longer periods as she used to. It seems that the Errorless Learning components built into Mark and Reward training were actually of greater benefit than I had realized. Apparently there is a difference between being “Right or Wrong” and being “Right or Not-Right-Yet.”
If there is no demonstrable benefit to marking an error, why do it at all? That is the message of Herbert Terrace’s Errorless Learning protocol. And it’s important that we realize that Terrace didn’t mean that the learner didn’t make errors, only that there is no benefit to the trainer acknowledging those errors. I applaud Emily Larlham for her detailed essay on these important concepts. But I think sometimes the devil is truly in the details.
I’m a self-admitted “training geek” and I love the science behind behaviour. But I need to be careful when pulling information from human psychology or canine ethology. It’s easy to cloud the issues with facts. It’s important to know where this knowledge comes from but I think it’s even more important to implement what we know effectively. I know from experience that too much explanation can be confusing and distracting. My dog is the most important thing. As my No Reward Marker experiment showed me, just because a concept is interesting doesn’t mean it will work for my dog. Understanding that Herbert Terrace’s Errorless Learning meant that we should not mark errors and not that the learner didn’t commit errors is also important. It puts the responsibility on the trainer and not the dog.
Until next time, have fun with your dogs!
Photo credits -