Training With Affection

At ClickerExpo in Portland last week

At ClickerExpo in Portland last week

Food is the go-to reinforcer for training with positive reinforcement, and for lots of good reasons: All animals behave in some way to get food; it's built in or we'd die. Food is a primary reinforcer, meaning you don’t have to learn to want it. With dogs especially, it’s easy to divide into small bites and deliver quickly and consistently, so that you can get in lots of quick reps in a given training session or day. And it’s an extremely potent tool for creating positive associations.

Yet animals also behave for many other reasons, and we humans seem to be particularly interested in whether they will change their behavior because they like us. Researchers have gone back and forth on whether social interaction with humans is a primary or learned reinforcer for dogs and how to quantify its effectiveness; a widely cited 2012 study found it to be less effective than food. Yet many dogs do seem to greatly value human attention, as evidenced by all the acrobatics they regularly perform to attain it—including a whole host of behaviors many people don’t enjoy, such as jumping, barking, mouthing, pawing, and object stealing. And that’s worth exploring, because especially when you’re trying to replace a bad habit with a better one, there are advantages to being able to reinforce the new behavior with whatever the animal was trying to get with the old one.

In recent years, I've attended several seminars presented by Human Animal Learning Opportunities in St. Louis with Dr. Jesús Rosales-Ruiz, an associate professor in the behavior analysis department at the University of North Texas. Dr. Rosales-Ruiz’s special interests include animal training, particularly clicker training, and he is generous about sharing what he and his students are working on with nonacademic practitioners. At these workshops, attended mostly by dog trainers, he reported on the progress of various projects by his grad students, including one called “Give Them Love" whose goal was to further explore the effectiveness of human affection as a reinforcer in training dogs.

I originally wrote about their protocol for my column for the local rescue One Tail at a Time, but Rosales-Ruiz presented on the topic again at ClickerExpo 2017 in Portland last weekend, so I thought now would be a good time to update and republish that post here.

In the shelter environment, dogs who do enjoy affection from humans are often starved for it—and behave accordingly, sometimes scaring potential adopters and volunteers alike with their wild enthusiasm. Training is thought to improve their prospects for getting and staying adopted (though we're still figuring out how) and can enrich the dogs' lives while they're in temporary care, but shelters have limited time, limited resources, limited staff, and if they’re lucky, a rotating cast of volunteers with varying aptitudes for training.

It was in this environment that the UNT team set out to develop and demonstrate their protocol, working with their local shelter in Denton, Texas. They wanted something that worked quickly and was as simple and systematic as possible so it could be easily carried on by staff, volunteers, and adopters.

To explore the use of affection as a reinforcer, they first had to determine what exactly that would look like. Clearly not all interactions with humans qualify. Some of them dogs find downright scary, and some they find annoying: It’s very common for owners to undermine their own training by “rewarding” a lovely behavior with a well-intentioned but off-putting pat-pat-pat on the top of the head.

The researchers in the 2012 study had decided that a good social reinforcer would be four seconds of scratching around the neck combined with verbal praise. But Rosales-Ruiz says that wasn’t nearly enough: Petting, he argues, is the type of reinforcer that is best delivered continuously as the behavior it’s reinforcing occurs, like music or kisses, and not in little bites, like food. So the UNT team started with a definition of “affection” as calm, gentle, and sustained physical contact.

The training process they came up with had two phases:

1. TEACH THE DOG HOW TO GET THE REINFORCER—JUST AS YOU WOULD WITH FOOD. In other words, make it contingent on something he does.

The students picked five dogs who jumped up on people under certain conditions, and chose two simple alternative behaviors to reinforce instead: sitting and lying down, both of which are incompatible with jumping.

They entered the dogs’s kennels bent slightly to greet them before they could jump, and then petted in a slow, sustained manner with one hand for up to two full minutes—or, initially, as long as the dog kept four feet on the floor. No verbal praise was added.

If a dog jumped up, the person stood up, stopped petting, waited for the front feet to return to the floor, and then resumed petting with one hand. If the dog walked away, the person also stopped petting.

If the dog sat or laid down, though, the person switched to two hands, and again petted for up to two minutes.

Within 5-20 minutes, all the dogs were sitting or lying down to solicit the students’ affection. And perhaps because there weren’t many contextual hints that the students were capital-T Trainers—no treats, clickers, or other special equipment—Rosales-Ruiz says the dogs quickly learned to respond the same way to shelter volunteers and potential adopters. We might guess that from the dog's perspective, anyone with hands might have been able to offer this reinforcer.

2. USE THE REINFORCER TO TEACH MORE BEHAVIOR. When the dogs would remain lying or seated for the full two minutes, the students switched to petting for 15-30 seconds, with very brief pauses (3-5 seconds at first) to allow the dog to choose to remain in position and “ask” for more. They began to use this interaction as a reinforcer to teach the dogs to stay for longer stretches, and while the person stood up, walked away, or entered/exited the kennel. All of these human behaviors evolved into environmental cues for the relaxed behavior. (Think about how handy that would be at home.)

To help clarify when the dog had met the current criteria, they incorporated a marker signal, equivalent to the clicker or “yes” frequently used in training with food to signal that reinforcement is on its way. For this, they chose a hand motion that looked like the beginning stages of reaching toward the dog. This ritualized motion—which civilians would be likely to do anyway as they went to pet the dog—could be consciously used to mark any other behavior the trainer liked, letting the dog know exactly what he’d done right and increasing that behavior in the future.

Some students played with this with their dogs at home to great effect as well, using the protocol to teach them in small, achievable steps to remain relaxed while they added distractions like plucking a harness off the hook by the front door, exercising on the floor, or leaving the room.

When the dogs were adopted, the new owners were offered a class in which the procedure was taught to them. In a HALO seminar, Rosales-Ruiz showed some impressive video of the dogs sitting and lying down patiently in an outdoor ring amid other dogs, adults, and children during class.

The students are reportedly in the process of building a website to share the protocol with more shelters and pet owners. In the meantime, this video is the only thing I've found available to link to. [Update as of March 2018: the Constructional Affection website is live.]

I haven't used this protocol exclusively, like the students did, to train multiple behaviors, but I haven't been at all surprised to find that it works well in reducing vertical greetings that have been inadvertently reinforced by human attention. And it has definitely changed the way I pet, specifically the duration. It has helped me find petting routines that my own dog, who has historically walked away from casual touch, finds reinforcing.

There are also some takeaways here that are bigger than any single protocol: (1) The individual dog decides what is reinforcing, or in other words, what is worth behaving for. Food isn't a reinforcer just because the bag says TREATS, and a pat on the face isn't reinforcing just because dogs enjoy some forms of human touch. (2) If you can observe what a dog is working to get, and you can give him same thing after a behavior you like better, you've struck gold.

This post was originally written for One Tail at a Time. It has since been revised and updated.