This is the seventh in a series of posts based on my notes from the 2020 Art and Science of Animal Training Conference that was held in Hurst, Texas on February 22-23, 2020. To learn more about the conference, you can visit the conference website.
While I try to take accurate notes, it is possible that there are errors or that some detail is lacking. I did include some photos from Mary’s presentation, but they are slightly crooked (oops) since I was sitting off to the side. Please note that this article is based on the speakers’ slides and my notes. If you choose to quote from it and want to know if it is a direct quote or not, please contact me. If you have a comment or question, let me know and I can try to clarify or provide some additional information. Many thanks to the speakers and organizers who allow me to share.
Mary Hunter: Planning your shaping from a non-linear perspective
One metaphor for shaping
When trainers are planning their training, they usually write out the steps they need to follow to achieve their goal. These are often referred to as shaping plans and the trainer starts with a small piece of the behavior, selects out repetitions that meet the criteria for the next approximation, and continues on until the animal has learned the final behavior. This is compared to climbing a staircase where each step directly follows the preceding one and you take a linear approach to get from your starting point to your goal.
The staircase metaphor:
- A linear path from start to finish
- Useful for duration, distances – in these cases, you are asking for “more” of the behavior with each repetition.
- Examples would be trailer loading where you start with the horse some distance from the trailer. With each approximation you ask the horse to step closer to the trailer.
While the staircase analogy can be useful, Mary went on to describe another way to think about shaping.
Another metaphor for shaping – Building blocks
Rather than starting with one behavior and changing it incrementally through successive approximations, another approach would be to find the pieces you need, use them to build small components, and assemble the components to form the final behavior.
The idea that behavior is composed of small units dates back to Skinner, who wrote:
Stokes and Balsam (1991)
“…the form of the behavior itself may be best described as a discrete unit composed of a series of subunits. Skinner (1938) thought that there were minimal units of behavior from which behavior could emerge.
We suggest that the emergence of these new types occurs when the minimal units are reorganized into new sequences.”
Example: Teaching the components for trailer loading – (Pohjola 2011).
Jaana broke trailer loading down into many different components and taught the horses the individual skills before she combined them for the final trailer loading behavior. By doing this, the horses learned more quickly and with less stress (lower heart rate and fewer fear related behaviors). Some of the behaviors she trained were:
- Go over a raised platform (with angled sides to simulate ramps) – she had the horse go over forward and backward.
- Go through a wire frame
- Go through a wire frame covered with fabric
Shifting your thinking from staircases to building blocks
Mary had three suggestions for shifting your thinking to the idea of using building blocks.
- Focus on actions
- Minimize variability
- Start with success
Focus on Actions
Linear shaping programs often focus on outcomes. The trainer may have a picture of the end result, but not of the individual actions that the animal needs to do in order to achieve this result. However, if you think of the outcome as made up of individual actions, then you can:
- Identify the component actions
- Design your shaping around these actions
Going back to the trailer example, there’s a lot more to getting on a trailer than just taking a step closer each time. Jaana’s approach was successful because she identified all the different components, which included new sensations and skills that needed to be learned.
Example: Dog jumping in a car
Mary was working with a service dog who needed to be able to get in and out of her car. She had been told he had no trouble getting in cars, but when she asked him go to get into her car, he would put his front feet in and then stop.
How could she teach him to get all they way in the car?
She could reinforce him for getting further in (once he’s stuck) – that would be a linear progression. But, it’s actually harder for him to get in if he’s stuck half way. A better option would be to look for some way to change the set-up so that he can do the behavior she wants.
What does that behavior look like? She wants him to jump in. She found a bag of animal bedding (large bag of shavings) and placed that next to the open door so he could use that as step until he learned how to jump into the car. She had him practice for several days with the bag, then removed it, and he was able to get in without it.
In one model of shaping, the trainer observes the animal and starts shaping by selecting a behavior from what the animal is already offering. Then, over time she changes the criteria to either select or drive the behavior in the direction she wants. This is similar to starting with the open end of a tunnel (gathering lots of behavior) and then narrowing the focus through shaping.
Is a funnel the best metaphor?
It may be useful in some circumstances, but it means you are starting with a lot of variability and the animal may not offer a behavior that is a good starting point. The funnel metaphor makes it more likely that you will:
- pick the wrong starting point
- pick the wrong environment
Rather than starting with a lot of variability – and hoping to weed things out later, it would be better to start with a narrower focus so you can minimize variability. .
Start with success
If the environment or starting point are not good, then we don’t get the approximations we are looking for. Instead of using a funnel we should:
- Select the right environment and starting point
- The first approximation should be super easy
- Later, extend or combine behaviors
Example: Dog wearing a service vest
Mary was working with a service dog that was reluctant to wear his vest. He would hide under the coffee table when she held out the vest so he could “get dressed.” Instead of starting with the vest, she worked through a series of approximations, using objects that had some of the same characteristics as the vest.
- Can you stick your head through a belt?
- Can you stick your head through something wider?
- Can you stick your head through something wider and heavier?
- continue with various objects until
- Can you stick your head through your vest opening?
Regardless of which approach you choose, it’s normal for there to be some setbacks or moments when you have to stop and regroup. If you are thinking of shaping as a linear progression, you might go back a step or two or add some intermediate steps that help you to regain momentum, clarify some point, or make some other adjustment that will allow you to continue toward your goal. But, if you are thinking in terms of building blocks, you may look for missing components instead of just reviewing the same shaping steps.
Example: Giving a horse a bath.
Mary shared this slide to illustrate how something like “giving a horse a bath” can be broken down into many different behaviors.
Sometimes we get hyperfocused on one component and forget about the others, so it’s a good idea to list and evaluate your horse’s ability to do all the component behaviors. For example, it’s easy to assume the biggest challenge is teaching the horse to stand while you get him wet, but maybe it’s not about getting wet – it’s about the hose moving around his feet. Or some other component that you haven’t thought to train separately.
This may seem like a long list, but don’t let the list overwhelm you. If you train a variety of behaviors, you will start to see a lot of overlap. Many of the behaviors you need may be ones you have already trained. And, those that you do have to train will probably end up being useful for other activities.
- Start by identifying building blocks
- Then, teach the components by;
- Focusing on actions, and
- Teaching in the right environment
- Troubleshoot by finding missing components
This talk was a nice follow-up to Ken’s talk on adduction because it continued the theme of combining individual behaviors to create more complicated behaviors. In Hannah Branigan’s talk (which was after Mary), she used the same idea of building blocks and described how she trains all her obedience behaviors out of a set of foundation behaviors.
With horses, I am constantly using already learned behaviors as component behaviors for new ones. A lot of husbandry behaviors share common elements (hold a body part still, accept contact, accept different sensations, maintain duration, etc.) and I find this is true with groundwork and riding too. If you’ve never thought about this before, it’s worth taking time to write out the behaviors you have trained and all their individual components to see what they have in common.
I have to confess that I laugh whenever I hear someone say that training being a linear progression. I think we’ve all seen the cartoons of how we expected it to go (a straight line) vs. how it went (a tangled mess or at least some ups and downs). But that aside, I loved how Mary explained the advantages of the building block approach. Not only is it more flexible but I think it encourages the trainer to be more creative.
One thing I wanted to mention is that I think people often get stuck in linear thinking because of how they write their training plan. When they list the actions that an animal needs to do (or learn to do) for a specific behavior, they tend to list them in the order in which they occur. But, that doesn’t always mean we have to teach them in that order. Yes, some will be dependent upon others, but there will be many that can be trained as independent behaviors. These can be taught separately and practiced to fluency so that when we are ready to add them, we are asking for something familiar and reinforcing, not adding a level of difficulty.
I’ve been blogging about the ASAT conference for several years now. You can find a list of them in the articles page on this blog, but some older ones are on my website, www.equineclickertraining.com.
If you are interested in learning more about how to clicker train your horse, check out my book, Teaching Horses with Positive Reinforcement, available on Amazon.