Friday, September 28, 2012

Shedd Animal Training Seminar: A Bridge to Better Behavior

Ken Ramirez was clear: using reinforcement is the best way to train an animal, and reinforcement is most effective when it is inherently enjoyable and provided immediately. But providing reinforcement quickly can be challenging, especially when the animal is performing at a distance. How can you get a fish to a dolphin when it’s in the middle of arcing through the air or a piece of meat to a dog who is performing an agility obstacle 20 feet from you?

This Shedd trainer uses a whistle bridge for the Aracari.

Well… you can’t. But you can bridge the gap between the time the animal performs the behavior and the time it gets the reward. This is done through the use of a bridging stimulus, or bridge for short. This terminology, while occasionally used in the dog world (we usually say “marker” or even “click”), is widespread in zoological training programs, and you have to admit, it’s a descriptive word. A bridge is a signal that tells the animal, “I like what you just did, and I’ll give you a reinforcer as soon as I can.”

Using a bridge is not essential to animal training- learning will take place whether or not you use one- but it does have several advantages. It is obviously useful when you just can’t deliver that fish fast enough. It can assist with precision by helping the animal identify exactly what part of the behavior it just performed is being reinforced. Was it the height of the jump? The size of the splash? The way it turned its head while mid-flight? The bridge provides clarity. And bridges work neurologically because they’ve been classically conditioned to signify that a reinforcer is coming.

If you choose to use a bridge (and Ken would argue that you should, at least some of the time), there are three things you need to do. You need to choose an effective bridge, you need to teach it to your animal, and you need to be proficient in using it.

When you select something to use as a bridge, there are a few things to keep in mind. First, and possibly most importantly, you need to use a bridge that your animal can perceive. Obviously, an audible bridge is useless if the animal cannot hear it, but you also need to make sure that it is unique to the environment so that it can be distinguished from other sounds. For example, a bridge that sounds like a telephone, doorbell, or microwave beep may not be a good choice for our pets.

The bridge should be practical: if it’s too hard to use, you will either struggle with it (and impact your timing, observation, or other critical training skills), or you’ll stop using it entirely. Most zoological trainers use a whistle that they can hold in their mouths because it leaves their hands free to do other things. Similarly, the bridge should be easy to replicate so that every time it’s used it sounds the same (this is especially important if there are multiple trainers working with the same animal).

And finally, the animal should have no prior negative association to the bridge. Although you can desensitize an animal to a sound they dislike, your training will be better off in the long run if it you don’t need to go through this process, especially considering animals can sometimes have a spontaneous recovery of the negative association.

Teaching the bridge to an animal is a fairly straightforward process. Pairing the stimulus (a whistle, a click, a flash of light, etc.) with a reinforcer repeatedly will result in a Pavlovian type response: the animal perceives the bridge and automatically expects that the reinforcer will come next. We dog trainers do this when we “load the clicker” by doing the click-treat repetition over and over again. Teaching the bridge is usually a pretty quick process. Dogs tend to figure out the click-treat association within five minutes or so. The Shedd staff tend to be a bit more methodical about this introduction, but even so, the animals in their care readily pick up on the bridge.

If the animal you are working with doesn’t figure out that the bridging stimulus predicts a reinforcer is coming, you should look at why. Is the timing off? If too much time elapses between the bridge and the reinforcer, the animal may not be able to make a clear connection. Likewise, if the reinforcer comes at the same time as the bridge or even before, the connection will be tricky or even impossible for the animal to understand. Or maybe the item you are using isn’t truly a reinforcer. Perhaps your dog doesn’t like beef because it makes him feel sick. Or it’s possible that the bridging stimulus you’re using has a negative association you aren’t aware of.

Finally, you need to be proficient at using the bridge. Can you physically operate it, and do so without excessive fumbling? Personally, I find i-clicks easier to use than box clickers (and I have friends who find the reverse to be true). You also need to make an effort to practice your timing skills. Ken showed us a variety of training games: you can train a human friend to do a simple task. You can bounce a ball and click every time it hits the ground (or bounces off a wall). Or you can enlist a friend to play “hand games”- the friend holds up one or more fingers at a time, and you click when they hold up only one, or only when it’s their index finger. Improving your timing will improve your training.

Once you’ve chosen an effective bridge, taught it to the animal, and are satisfied that you can use it well, you’re all set to bring out better behavior in your animal. In my next post, I’ll tell you about the different ways the Shedd staff do this. But for now, I’d love to hear from you. Do you use a bridge (or maybe more than one)? Why or why not? If you do, what bridge(s) do you use? Please comment with your experiences!

8 comments:

Elizabeth said...

What was the general consensus about using a bridge/marker during a stationary or continuous behavior that you want the dog to continue while receiving the reinforcement and afterwards?

Best example I can think of is a stay. So you want to feed 30 seconds into your 1 minute stay. Do you use a marker and then fed? Or does the marker signal the end of the behavior so they would be right to get up and receive their reinforcement? Or can your marker mean, keep doing what your doing, you earn a bonus reward, if you continue to stay as I bring it to you?

I've heard several different trainers with several different opinions on this. What is your take?

andrea said...

my bridges are yes and clicks generally .. my husband uses nice .. I am amused by how quickly dogs figure out all three ... for precise things I like the clicker - but i find for things like stay it tends to mark the end of the behaviour (as it should) so I prefer a long yes then I have time to release - just sharper timing ...
Read an interesting article ages ago about using a longer bridge for things like dog walks... author used 'excellent' as she could really draw it out - a little like warm/hot she felt it let the dog know it was on the right track to reward ... never tried it myself but thought it made sense:)

Chris and Mike said...

Click (box clicker) or Yes! (said sharply, distinct from all the other yesses used during a day). The click is more precise, but the clicker is not always in my hand.

For duration behaviors: Habi and I have been practicing heeling across all streets, now that we are fairly good at loose-leash walking (for those who don't know her, both steps are miracles for our reactive border collie). If she's somewhat distracted when we reach an intersection, I say "Follow Me", then click/treat after two or or three steps and immediately cue "Follow Me" again. Other times, when she's more focused, she gets lots of Go-o-o-ods and Wo-o-o-ows (it's amazing how you can draw out and modulate a single Wow) as we cross, then a jackpot on the far side.

Chris and Mike said...

Oh - a great website for improving reaction time (= clicker speed): http://www.bbc.co.uk/science/humanbody/sleep/sheep/reaction_version5.swf

It's humbling, and very, very fun.

Kerry M. said...

I use a couple - a very chirpy "good" and a clicker.

I'm also experimenting with a new one - the dog's name. I currently have 3 dogs most days (co-puppy raiser so I go from 1-3 dogs depending on the day), and I find one of my dogs seems hopeful/expectant whenever I say "good" even when I'm clearly working with the new puppy. My other dog just gets it, whoever I'm working with has earned the treat. Rather than devaluing my marker word for my foolishly optimistic dog, I am trying to use the puppy's name as an individual marker, which works for right now since I'm treating all name game responses. Not sure if it will work well long-term or not, but I figure it will help to get me over the next month when I have intensive puppy training.

Anyone else struggle with markers for multiple dogs?


Sara said...

I use both a clicker and a verbal marker, along with a visual marker (the thumbs up signal) with all of my dogs.

Crystal (Thompson) Barrera said...

Elizabeth-
There wasn't any conversation about whether the bridge/marker ends the behavior. From what I remember of the observations, though, the bridge did end the behavior for the animals at Shedd.

I, too, have seen and heard of trainers doing it both ways. I personally prefer for the bridge to end the behavior, but people seem to be successful doing it both ways.

Kerry-
I don't have multiple dogs, so I haven't struggled with that issue. But! There will be an upcoming post on training multiple animals. :)

ThanksandGig'Em said...

These are wonderful posts. Thank you so much for sharing them. I shared one of your pictures and provided a backlink to this post. It can be found brittanymead.com