A friend asked me how much do dogs remember if they only train something once per week. The reference was to our weekly agility trainings. The answer, as Facebook would say, is complicated.
First, I would like to point out the 2011 study The effect of frequency and duration of training sessions on acquisition and long-term memory in dogs. While the study was focused on how would different training schedules impact how long the dogs would retain the knowledge afterwards (the answer: once behavior was learned all groups retained it for four weeks regardless of training schedule used), the interesting point for our discussion is that not only have those Beagles that trained only once or twice per week remembered their training, they learned faster than those who trained every day.
Whoa! What are you saying here? Should I just train my dog once per week for fast learning?
Well, it depends on how you define ‘fast’ and I think it also depends on what you’re training. In the study all groups trained the same behavior. Would results differ if the behavior trained was more complex? Who knows.
Here’s the graph showing how quickly each group progressed through stages of learning the task of jumping in a basket. One session = 6 repetitions.
W1 – trained once or twice weekly for 6 reps
W3 – trained once or twice weekly for 18 reps
D1 – trained daily for 6 reps
D3 – trained daily for 18 reps
Dogs that trained once or twice weekly for only 6 repetitions (W1) reached training level 7 in 10 sessions (60 repetitions) while those that trained daily for 6 repetitions (D1) reached the same level in 18 sessions (108 repetitions). Obviously the group that was only trained once or twice per week had no problems remembering their training sessions. They even learned faster! However, for group W1 those 10 sessions took 7 weeks while for group D1 their 18 sessions took less than 4 weeks. If talking about treats delivered, group W1 learned faster. If talking about how long it took, group D1 learned faster.
Another interesting tidbit from the study’s discussion: “Results of Maslovat et al. (2004) suggest that interference of an extra task might be as beneficial to acquisition in humans as extra practice on the initial task, which suggests an important effect of the intervening interval.” If this is also true for dogs it seems that we could safely practice other things during the week and they would not interfere with learning in our once-weekly sessions.
From this study it seems that if I had a dog that could eat only a handful of treats per day (or would only be interested in a few treats or whatever) and I wanted to teach him several behaviors it would be best to practice those behaviors once per week. Even if the dog would only get 6 treats per day I could train 7 behaviors simultaneously. Each behavior would take a little longer to complete, but I would maximize the number of behaviors learned over a month or two.
I would like to note one thing though: those Beagles were only learning one behavior and they were always brought to the same room with the same equipment. Dogs are very sensitive to the environment in which we teach behaviors, so for my above scenario to work I would have to practice each behavior in a different location and/or with different props.
Now back to our original question on training agility and to my very unscientific observations of what is going on. This was a laboratory experiment and the environment in which our dogs train agility is very different, so we need to take this into account. The Beagle study made sure that the experiment room looked the same during each session and there were no distractions that could interfere with learning. This is crucial when teaching a dog new things: there must either be very weak distractions present or the dog must have learned to cope with those and stay focused on the task regardless. Our agility environment is quite a bit different: we train outside, there are other dogs present, new smells each week and other distractions.
The behavior we were discussing with my friend was sending the dog through a tunnel which is in a way similar to the behavior the Beagles were training. They both work by building the value for an object – a basket in Beagle’s case or tunnel in our case. The more value the object has, the more our dog is going to gravitate toward that object. The value is built as we feed the dog on the object or reward looking at, walking toward, touching and walking on/through object. Running through a tunnel is not an intellectually difficult task (at least not until the dog needs to find an entrance from weird angles). It could be scary or uncomfortable for some dogs, so it could be emotionally difficult, but once the dog has run through the tunnel a couple of times, there is nothing intellectually difficult about it. It’s all a question of whether the dog finds it worth his while. Is running through the tunnel worth more than sniffing or visiting the classmate? That is the real question.
Value building is an interesting thing to observe. If your dog is one of those who is usually slow(ish), try to teach your dog to go to a mat and lie down using his dinner kibble. Observe the speed with which he moves. Better yet, record a video. Next day use very very good treats – dinner leftovers (lots of dogs think that people food is the best), baked liver treats, chicken gizzards cooked with asparagus (Ruby loves those), cheese – whatever your dog loves above all else. Of course the dog will already know how to go on a mat, but watch the speed. Any difference? I’m betting he was faster this time! Now this doesn’t conclusively prove anything because he could be faster simply because he already knew the behavior, but in my experience this isn’t the case. I was actually working on this exact task a few years ago using kibble for several sessions and Ruby was going through the motions, but without much enthusiasm… And then I used really good treats the next time. Wow! I wasn’t expecting such big speed difference!
Caution: do not try using better treats in the middle of a session if your dog is slow – you would only be teaching him that moping around will make you produce better treats. Instead, end the session and wait at least 10 minutes before you go and get those better treats. You don’t want the dog to learn I’m slow -> she cuts up better treats -> then we train using better treats! Score! Being slow pays off.
In the example above you can determine how much value you have built for the mat by how fast the dog approaches it. Another way to see the value is to take the mat into another room: is he still just as keen? How about if I take it outside? Now we’re testing the value of the mat against the distractions of the outside world. What is hard is not the intellectual knowing that he’s supposed to go lie on the mat (though for dogs with poorly developed generalization skills the ‘knowledge’ part will be difficult, too) – it’s the decision to say “no” to distractions and “yes” to the mat. The thing with the higher value wins.
Back to our original question. By now I hope it’s obvious we’re not really teaching the dog to run through the tunnel as much as we’re building the value for running through the tunnel. How can we build this value faster? As we have seen from example with the mat the object gets bigger value if our rewards are better. If I reward running through a tunnel with a piece of kibble the dog might decide this is not enough and sniff next time instead of running through. If I reward it with a piece of meat, it might be like putting 1 EUR into the tunnel’s bank account. Three pieces of meat + me running and praising excitedly might be an equivalent of 5 EUR. And a Two Toy Game after the tunnel might be worth 10 EUR. Each dog has their own value system.
What motivation looks like
So if my friend runs her dog through a tunnel ten times rewarding with a piece of meat she will put 10 EUR into the tunnel’s bank account (if the dog is even willing to do 10 repetitions in the row – this might not be the case with some dogs). Then I run my dog through the tunnel three times and play Two Toy Game each time. I just put 30 EUR into the tunnel’s bank account and it’s very unlikely my dog has quit on me on just 3 repetitions (plus, as you saw in the Beagle study, less repetition might actually be better).
Distractions can decrease the value you put into that tunnel, because the dog must work harder. So if my dog’s friend is near, my dog will need to use self-control to stay and work with me instead of going to play with his friend. I will not be just paying for the tunnel, I will also pay for that self-control, so from my imaginary 10 EUR only 5 EUR will go toward the tunnel (but 5 EUR will go toward self-control, so his self-control will grow stronger with time).
As all analogies, this analogy with money is not perfect or scientific or even original (I think picked it up from Susan Garrett, but then it grew in my head to explain more and more events). But I find it helps me predict dog behavior, create training plans that work, and explain what is happening to others. Not too bad for an imperfect analogy 🙂
I hope I have showed you that “remembering” is not just about memory, it is also about motivation. And motivation is not about how often you practice and how many repetitions you do, it’s about how excited your dog is about rewards you give for those repetitions and about the distractions that compete for his attention.
The source of Java’s motivation
Thanks to Helena Mesarič and Monika Pleterski for photos!