ep 91 responding to errors in training
===
[00:00:00] If you train, compete, teach, or simply love the sport of dog agility, you're in the right place. Hey there, I'm Megan Foster, creator of Fostering Excellence and Agility. Join me as we explore training, communication, handling, mindset, and all the little details that help build stronger and more confident teams both in and out of the ring.
Whether you're just getting started or working toward big competition goals, this podcast is here to support your journey with your individual teammate. Let's get started.
Hey there, this is episode 91, and today I'm going to talk about responding to errors in training. I don't think that there is one right answer or one right approach to this question, but there are some important things to consider when you are deciding how to respond to an error. Because I aim to teach using errorless learning procedures as much [00:01:00] as possible, I sometimes don't think too hard about how I'll respond to a mistake because it happens so infrequently that my response will have little to no impact on the behavior.
I totally heard myself just now. That approach is super oversimplified, and it probably isn't very helpful to any of you listening, so I am going to dive into the nuance next.
First, when we're talking about responding to mistakes, we have to acknowledge that our goal is to reduce the number of errors and s- and how, and reduce the number of times we see the unwanted behavior. Our goal is to see the wanted behavior happening more. We can't know how our responses are impacting the behavior until we see the data.
Does the unwanted behavior [00:02:00] decrease? And if so, you have effectively punished the unwanted behavior. Punishment as a procedure means providing consequences for a behavior that decrease the strength of that behavior. Punishment is the opposite of reinforcement, which as a procedure means providing consequences for a behavior that increase or maintain the strength of that behavior.
Both of those things are neutral. Because learners decide whether a consequence is reinforcement or punishment, we can still see behaviors increase and decrease even when we are not intentional about the outcomes. So that's the science. We're not going any deeper into other meanings, types of reinforcement, types of punishment.
We're just leaving it there, that [00:03:00] punishment means providing consequences for a behavior that decrease the strength of that behavior. So whatever approach you have to mistakes, if you see that the unwanted behavior is decreasing, you are effectively punishing it. So let's get into the art of this and the fact that every situation is unique
We have to consider the phase of learning the dog is in when deciding how to respond to an error. During the acquisition phase, I want my loops tight and the reinforcement rate very high. If the dog makes a mistake, that is on me. They don't know the behavior yet, and I will continue the loop as best I can to maintain motivation to keep trying.
I take responsibility for the behavior happening in this first phase of learning. So you could really just say that I [00:04:00] ignore errors, I maintain clean type lo- tight loops as best I can, and the dog doesn't really notice anything has changed. As I move into the basic generalization phase and beyond, I do want to start seeing that the dog understands the cue in a variety of situations, and that they aren't attaching the behavior to a specific context that I do not control.
This means that I am likely to see an error at some point, and I do want there to be a different outcome for the dog when she responds correctly to the cue. This is where things can get a bit messy. Most behaviors in agility are a series of behaviors linked together on one cue. For example, cuing a backside slice includes committing to the backside, taking off for the jump, cleanly jumping, and then following the handling upon landing.
If the dog breaks commitment, no piece of the behavior was [00:05:00] completed correctly. But if the dog commits to the jump, takes off for the jump, knocks the bar, but follows the handling, three out of four pieces of the behavior were completed correctly. Can you respond to both of those errors in the same way?
You could, but if this is happening chronically, you might see evidence that you've punished the entire behavior and not just the piece that was incorrect.
This is why I base my decisions on what my goal is for that training session. Was I working on keeping the bar up during backside slices? Is there a way I can break that down more to isolate the jumping effort without using the backside skill? What if I'm just working on maintaining connection and flow state within a longer sequence or course?
Should I stop to tell the dog that knocking the bar was incorrect? I wouldn't in most cases. Another example is a stopped [00:06:00] contact. If your criteria for your teeter is four on, and your dog speeds across the teeter and stops on two on, two off, what would you do? Is that the same mistake as flying off the teeter?
Because most of agility is like this, I really take into consideration the individual learner. So if you were hoping for a really clean flowchart, I'm sorry, that doesn't exist in my world. When we're talking about the individual learner, I'm curious about their learning history with this specific behavior.
How high is their overall motivation to do this behavior in the first place? Have I been clear and consistent with my cues and criteria for this behavior in the past? Am I retraining a behavior or maintaining one that is working well? My goal is to balance the need for clarity and precision with the dog's confidence and motivation to perform, [00:07:00] and that does often require a different approach based on the dog and the task that I'm training.
With all of that in mind, while I still want to focus more on how I set them up for success, I have a lot of ways of responding to a mistake. One, I can ignore it and continue on as if it was correct or just didn't happen. I can usually do this in the acquisition phase and in a lot of sequence or coursework situations.
Two, never wrong, sometimes more right, a la Sarah Stremming. This means that I am paying correct responses with a high-value reward and incorrect responses with a low-value reward. I really love this approach for maintaining motivation to keep going for dogs that are more unsure or are having a harder time with something physically or mentally.
Or when the behavior was mostly correct, and I want to reward the effort that was given, [00:08:00] even if the whole behavior wasn't perfect, like my stopped contact example Three, withholding reward for the incorrect behavior, but paying the dog for resetting to me and transporting back to the starting point to try again.
I tend to use this a lot for situations where I can see the dog trying really hard to solve the problem, and I wanna skip past the mistake quickly. I don't want them to stew on it. I don't wanna put pressure on the mistake. I want them to get back quickly and try again while they're still in a good head space.
Four, I can remove the dog from that rep and inter- interrupt the behavior and withhold the reward completely. Now we're getting into situations where I know that continuing will lead to a misunderstanding in criteria, like skipping a weave pole or missing the weave pole entry. I want to pull the dog out of the behavior the [00:09:00] moment the mistake happens and withhold reward versus letting the dog finish the weave poles, usually correctly, and then not rewarding.
So by not removing the dog, I would create more confusion in that situation. And five, I can remove a dog from the training session or the run completely in response to an error. This is an extreme use of withholding reward that I don't generally use, but can if the situation calls for it. Clear as mud, right?
I know it isn't a clear flowchart to go down, and you can absolutely take a very black-and-white approach to responding to errors. You can withhold reward for every single mistake that you see, and you'll probably be okay. But I have found a lot more joy in having a variety of ways of [00:10:00] communicating with my dog, and over time, the right response starts to come pretty naturally.
I don't have to think about it too hard if Sprint makes a really uncommon mistake in the last obstacle of a sequence. I can usually laugh it off and toss her her toy anyways, and it doesn't impact her overall performance I can also be in the middle of a very precise skill training session and know when I should reward the transport and when I shouldn't.
It just comes naturally at this point because I know her so well. Bottom line is that progress isn't made by how we tell them they're wrong, but how we set them up to be right. I'd love to hear your thoughts and questions, so join me over in the Fans of FX Agility community and let's talk about it.
Details on how to join [00:11:00] are in the show notes
Thanks so much for listening. If you enjoyed this episode, I'd love for you to leave a review, share it with an agility friend, or connect with me on social media. Your support helps more handlers and teams find the podcast and be part of this community. You can also find all of my training programs, coaching, and resources at fxagility.com.
Remember, excellent agility is built one training session at a time, and every small step forward matters. Until next time, happy training
