Please visit my conference and seminar notes page for more notes from this seminar.
Earlier this summer, I had the chance to audit a seminar in Dallas with dog trainer Michele Pouliot. If you’re not familiar with Michele, I encourage you to check out her website. Michele is a very talented dog trainer who has won international competitions in the sport of canine freestyle. She also has worked for almost four decades with Guide Dogs for the Blind, training hundreds of guide dog teams. Michele really knows her stuff and it was wonderful to get to spend two days learning from her.
One of the most important concepts in training is shaping, which means using a progressive series of steps to train a new behavior. Shaping allows a trainer to start from a point where the animal can be successful and gradually build toward a more difficult or complex behavior.
However, shaping can sometimes be difficult for both beginner trainers, as well as more experienced trainers. When shaping a new behavior, the trainer must make split second decisions. The trainer often has to make decisions on the fly, and can’t spend time deliberating whether or not to click.
During one of the question and answer periods, Michele discussed when to raise criteria during shaping. This can be a tough decision. The trainer doesn’t want to stay too long at any one step, but also does not want to move forward before the animal is ready.
Many trainers use what is called the “eighty percent” rule. That is, you can move on to the next step when the animal is correctly doing the current step eighty percent of the time. The problem with the eighty percent rule is that most trainers are not actually taking data and calculating out the percentage of time that the animal does the behavior correctly. So, rather than the eighty percent rule, what most trainers use is the “well I think he’s probably doing it correctly around eighty percent of the time” rule.
In any case, Michele discussed that she prefers the “fifty bucks” rule rather than the “eighty percent” rule. That is, Michele often asks her students “Would you bet me fifty bucks that if you give the dog the opportunity to perform the step you are currently training, that he will do it immediately and correctly?
If you would bet fifty dollar, then it’s time to move to the next shaping step. If you wouldn’t bet fifty dollars, then you need to stay at the same step (or perhaps even modify your shaping plan). If the answer is maybe…it’s probably best to not move on yet! Michele uses this rule in her own training and when teaching students and she said it has really helped her and her students.
Also, Michele suggested that if something is not working and the behavior is not improving, rather than continuing to click, you should take a break and evaluate the situation. During training, you always want to feel like the behavior is getting better.
I think that trainers sometimes end up getting stuck and stay too long at one step because their shaping program moves in too big of steps. Often, by breaking a task down into even smaller steps, the animal can master each step faster, and the whole program actually moves much more quickly.
I found this discussion during the seminar very interesting because shaping is one area of training that, in many ways, is still much more of an art than a science. More research is still needed to develop better guidelines to help trainers understand how to shape behavior quickly and efficiently, with minimal frustration for the animal or trainer.
Do you use the eighty percent rule or the fifty dollar rule? Or, do you use another type of rule or strategy to decide when to move to the next step when teaching a new behavior?
Oh, Mary!
If I was to use the “fifty dollar’ rule, I’d be stuck at the very first step 🙁
Being a non-gambler by nature, the most I will gamble is ‘a kiss’ 🙂
And being a strongly anti-gambler from (bitter) experience, I would not even be able to recommend this to my friends or clients 🙁
(On the other hand I think that 4 out of 5 correct responses is still too low to proceed further. )
Cheers,
Jenny
Hi Jenny,
Thanks for leaving a comment.
So, do you have any specific rules or strategies that you use to help you (or your clients) determine when to move forward during training?
To a large extent, I think “when to move forward” can often depend on the animal and what is being trained. However, I think guidelines can be helpful, especially when teaching people who are new to training.
~Mary
Yes, guide-lines are necessary — but in my experience (as an instructor) people WILL interpret these guide-lines to suit themselves.
I remember Ian Dunbar giving the advice to give (I think it was) three treats to a pup for toiletting outside when toilet-training. He said to count then out — one, two, three as you deliver them.’
Somebody asked ‘Why Three?’
The response was, that if he merely advised ‘give the dog a treat after peeing’, half the people won’t bother with treats at all, and the others are likely to be mean with them.”
So my usual advice is to wait for 9 out of 10 correct responses. Then people are likely to at least not up the ante until they get at least a 50% correct response.
I have to agree with Jenny H. I would never bet $50 on my horse’s behaviour. But modifying the $50 rule to a $1 rule may help in my case. I will try this and see how it works for me.
I am also not a gambler, and the though of losing 50 dollars is quite aversive to me. 🙂 However, I don’t think that was the real message. I think Michele (and Mary) were telling us how to use a different rule of thumb, one that can help us if used in the context of shaping behavior. If you aren’t comfortable with the $50 bet (as I am not because it would lead me to be too conservative), then choose another ROT. I like Jenny H’s suggestion (a Kiss) or Christel (a dollar).
One last thing I would like to say… I really wish we had a program (app?) that could help with in the field data collection and on the spot calculation. That way, we wouldn’t have to use heuristics. Maybe an app that would allow us to calculate percentiles very quickly? Did anyone read Galbicka, G. (1994). Shaping in the 21st century: Moving percentile schedules into applied settings. ?Journal of
Applied Behavior Analysis, 27, 739-760? The contraption they suggest would still be too burdensome to use in the field, but maybe an updated version?
Great post, Mary! Thanks for sharing the info!
Hi Valerie,
Thanks for leaving a comment.
I am not a gambler either and do not intend to start now! You are correct, I was not implying to actually bet money and Michele was not either. I’m sure we could come up with another metaphor or saying though, that would say the same thing without a reference to gambling.
The point, I think, is to make sure that you’re sure the animal has really learned the current step, before moving on. I think a lot of trainers run into problems because they get greedy and try to move to the next step too fast.
I would love to see more people doing research about shaping. I think this could be so beneficial to the training community, as well as to our understanding of how behavior works.
And I definitely agree with you, that we need more people inventing and researching better systems for data collection in the field. I actually have a friend who’s working on an iPhone app that should be pretty cool, but it is still very much a work in progress.
cheers,
Mary
Ah, I know that it was ‘not a real bet’ — but I feel that the analogy iis NOT good.
To some people, a bet of thouands of dollars is jus small change (eg The Packers) to other people $10 is serious money (old age pensioners).
So, different people will interpret the message differently. Compulsive Gamblers might be willing to bet on a ‘once in five’ correct response — certainly Lottery Ticket buyers are happy to bet on one in thousands chances.
Other of us might be far more conservative — wait for a less than one in a hundred chance of losing.
I am not going to get caught in this sort of discussion with clients. I would prefer to give them a far better idea of the sort of reliability they should be aiming for.
(OK — I am a literalist. Not the best sort of person to give such advce to ;-(
Hi Jenny,
I certainly see your point and where you are coming from.
And I agree with you, clear, unambiguous guidelines are preferable to things that can be open to interpretation.
As you mentioned in your other comment, though, many people also create their own interpretation of the eight out of 10 (or nine out of ten) rule.
So, the challenge for us trainers, I think, it to come up with ways to teach people so that they see the benefits of waiting until the animal really knows the behavior before moving to the next step.
~Mary
Hi Mary, I love this concept of Michelle’s. Thank you for explaining it so well. I agree, the fifty dollar idea is in the head, not in actual currency.
I can see how it can really make us stop and think.
Then, of course, we could bet another fifty dollars on whether or not the horse will execute those favourite behaviours that s/he loves to perform to get us clicking even though we have in no way suggested it!
Hi again, Mary. May I share this with the Horse Agility people and Alex Kurland’s group, with a link to your blog? Ta. Hertha
Hi Hertha,
Yes, of course, feel free to share!
I would just ask that you include the link, but it sounds like you’ve already thought of that. 🙂
cheers,
Mary
I´ve recently found this blog, and it´s really interesting and entertaining! Nice writing!
The first time i heard about the 80% rule was at Bob Baileys Chicken Camps where he recommended to increase criteria when you have at least an 80% success on 2 consecutive sessions. I believe that last part makes a big difference!
You also have each sessions “sound”, how does it sound like? click, click, Click, click rather than click…………… click,……….. click…………. click
Experienced trainers develop some intuition about this moments, having trained the same behaviors a number of times with different animals.
Never heard about this 50 dollar rule, but it deserves trying it. Maybe betting a carwash? 😉
I also really like to think about the “sound” or the rhythm of the session!
You can tell a lot by listening to the sound.