Tuesday, April 20, 2010

Laev Throws Me a Bone

I was sorely tempted to skip Schutzhund practice tonight, after my last post, but I went. And it was a good thing I did.

(Long post, so here's the summary -- 300 Peck rocks, Laev nearly breaks my neck, and I am happy about it all.)

I cut up 3 hot dogs to use on the field. I'm not a big fan of hot dogs -- I don't like the smell or the texture -- but they do count as high value rewards in Laev's book, and I was starting fresh. Some of my very first Schutzhund research included a Gottfried Dildei tape in which he conditioned dogs to love the training field by simply walking them onto it and feeding hot dogs, until they were rabid field addicts. Say "classical conditioning" all you want, I couldn't quite bring myself to throw hot dogs for free :) but I had another plan.

(After all, Bob Bailey tells me that Pavlov is always on my shoulder -- so any wiener-intensive activity would have the same conditioning effect.)

After my NRM article came out this month, I received an email asking about building duration in heeling without using an NRM. I recommended the "300 Peck" program. At ClickerExpo a few weeks ago, I heard Alexandra Kurland explain 300 Peck more accurately than what is often bandied about on the internet. (The name "300 Peck" comes from experiments training a pigeon to peck a lever 300 times for a single reinforcer.) Her version is very useful for training duration -- or, in my case, re-training. I opted to try it tonight.

So I started toward the field, set up, and waited for Laev to give me attention. (I could have asked for attention, in normal training, but one of my goals tonight was to see what Laev was capable of gathering and offering me without any extra prompting.) When she put herself in heel position, cued by my posture, I cued "heel" and took a single step before halting. Laev moved with me and sat. I gave her a bit of hot dog.

Next, two steps and halt. Treat. Three steps and halt. Treat. Four steps, halt, treat, and then when I was shooting for five steps, she glanced away from me to look at another team nearby. Looking away is not a part of correct heeling behavior (we train for eyes-up, attentive work), and so I simply stopped and stepped back a bit. Laev realized she was not only out of position, but had missed her opportunity, and she flung herself back at heel with a sharp bark of frustration. I cued "heel" again and made it our five steps.

(The most common misrepresentation of 300 Peck is that when the learner makes a mistake, the trainer begins again at the count of one -- and this actually reinforces errors by making it easier to earn reinforcement after failure! Instead, I simply started where we'd left off; Laev had successfully completed four steps but failed with five, and so I reset and asked for five steps again, not one.)

Laev didn't make another mistake while entering the field, and then I stopped 300 Peck and doodled for a moment with simple things. When I took her to the setup for the heelwork down the center of the field -- the gunfire danger zone -- I restarted 300 Peck with a single step. Here Laev was a bit conflicted; she was quite vocal, barking sharply if she made a mistake and occasionally just for the heck of it, and there was even some intermittent whining. I ignored all the vocalization; I know it's just a symptom and not worth addressing. It will vanish if I fix the real problem. Still, with all that, she stayed focused and intense in her work. Between reps, she glanced away and sometimes seemed to have a hard time refocusing; I let her work through her own conflict, refusing to prompt or help her. I wanted a baseline, and I'd rather not have her rely on outside help I can't give later. And she always managed to pull herself back to me, giving me attention so I could cue the heel again (and thus reinforce the attention).

As I recall, in the original pigeon experiments, researchers noted that as the ratios became longer (say, 200 pecks for a single feed instead of ten pecks), the pigeons took longer to initiate the behavior (procrastinated) but then worked intensely once they'd started. This was very like what I saw with Laev; her ratios were much smaller, but the challenge was greater at center field than at the side, and she delayed probably 5-8 seconds at her worst there. Once she turned her head back to me, however, she worked well.

Several times I took a break, doodled, and then restarted at lower ratios. We got as high as 20 steps of correct, intense heeling for a single bit of hot dog tonight, a feat I think we achieved thrice. That's not very impressive when I recall that she did about 500 steps of heeling for her BH on this field, but it's an order of magnitude better than Saturday's session, where she simply shut down in the middle of the field and told me she'd rather not play, thank you. So I think we'll be sticking with this plan for a while. We might not ever add the gunfire again, but maybe I can at least get our old level of performance back.

So it was a good session already, even if we'd quit there, but I stayed on for bitework.

I'm trying to give prime training time to the dogs who are actually trialing in a couple of weeks (I have a schedule conflict, which is why we're not doing tracking or protection titles instead of a full Schutzhund title attempt). So Laev went last tonight, and I asked the helper to give her suit bites, something we're playing with as I consider other sports. Laev gets a bit hot on the jacket, as it's a different type of fight than on a sleeve, and when she was glowing in the Tolkienesque forge-fires of glorious battle, my helper suggested teaching her the ringsport Defense of Handler.

The complete exercise is this: The dog and handler heel forward and encounter a couple of decoys, one of whom greets the handler and shakes hands. They move on, and the handler cues the dog to heel backward and keep an eye on the suspicious characters. One of the decoys sneaks up behind the handler and strikes him, and the dog is to bite in defense. Key points are, the dog may NOT bite the non-threatening decoy as he shakes hands, waves his arms, etc., and the dog may NOT bite before the handler is struck, even if he sees the decoy coming.

Laev doesn't know how to heel backward (well, actually, she has a lovely heel in reverse as I walk backward, but she doesn't know how to heel facing the rear as I travel forward), so we started her just in a sit facing backward. The helper (R---) came up behind me, raised his arms, and then struck me with both hands. I shouted, and Laev launched in a gorgeous black-and-mahogany fireball of avenging fury. It was perfect.

Except for the fact that Laev logically went for R's arm on my back. And en route, she grasped a large hank of my long hair. So as she stuck the decoy and knocked him backward, fighting furiously, she yanked me around rather violently by the neck.

Once I got free, I teamed up with Laev to beat on the helper for a bit :) and then collected her for another try. Laev grasped the game very quickly. On her third or fourth rep, she tried moving early when R started sneaking, but he simply stepped backward and I just blocked her with the leash so that she couldn't reach him, and we reset. It was an honest question -- "what's the actual cue for biting, the approach or the hitting?" -- and she did not make the mistake again. We did it probably 8 times or so, I didn't count, and the last time R tried to psych her out with some feints toward me, and she remained coiled like a cobra until he made contact with me. Good girl! and we both had a lot of fun.

That was it for the night, and we packed up and started home. It wasn't until the drive home that I realized how much I hurt from getting yanked in a downward spiral by my head -- and that was only a half-hour later. I am gonna HURT tomorrow!

But, even if I do, it was worth it -- we had a good obedience session and a rip-roarin' bitework session. Laev will be easier to live with for the next couple of days, and I saw some progress toward regaining what we'd had. Yay!


Anonymous said...

The gunfire situation is not working as you are missing an essential part of what she is seeking and that is relief. You need to implement successful negative reinforcement for her. What she is seeking after gunfire is relief and she needs you to implement it. This is not a case where trained self relaxation protocol will hold up its too situational. On the field the dog is programmed to play your game, follow your commands, to earn reinforcement. She is trapped by her commands and as such the stress builds. What you need to do is release her from work and take her to a safe area (crate, mat, etc) one step if heeling, one second if down after gun fire. You can impose relief on her successfully using neg reinforcement. If she knows you will get her out of the stressful situation she will trust you. You can then work on building delay to the release. Two steps heel, two seconds on the down, etc etc. This dog can be successful in the sport, you have to step up and be her team leader out there. When you take the field you are in charge (at least in her mind) she knows that you control her reinforcements. You need to implement relief. I hope you dont take my post as a criticism its not. I can be contacted if you would like through www.rcadogsports.com

Laura said...

Rebecca (I went to look you up), thanks for your comment! Your point re negative reinforcement and building up to the relief is right on. I actually have done this, and that is how we built heeling with the cap gun in my hand. I can't leave the field promptly enough, due to size and layout, but I did heel, pop the gun, and then send her to her relaxation mat. It worked well enough to get us to our high point!

It is very clear that running is negative reinforcement for her, dumping the pent energy and stress. She isn't fleeing the gunfire and me -- she never ran back to the car or elsewhere, she just ran a large loop and returned. It was just a laaaaaarge loop! I needed an alternate R-, and so I used the relaxation mat.

Again, I think a big part of the problem is there is NOT always predictable relief when she's exposed to gunfire, as I can't always know it's even happening, and so she never quite totally trusts in the coming relief -- which upsets me on several levels.

Crystal said...

Wow, I'm so glad I stumbled on your blog.

First of all, that NRM article? Was amazing. I read it about a month ago, and loved it. My dog and I experimented with NRMs awhile back, and we both found it incredibly frustrating. We dropped them, and fast. I only wish I'd found that article sooner (but I suppose it hadn't been written when I was looking for it).

Second, this:
The most common misrepresentation of 300 Peck is that when the learner makes a mistake, the trainer begins again at the count of one -- and this actually reinforces errors by making it easier to earn reinforcement after failure!

Is also really helpful. Starting over again at one step never really felt right to me with 300 peck heeling. My dog is too smart for me most of the time, and she would definitely figure out that making mistakes was in her favor.

lol accounts said...

For anyone who're not familiar with Kairosoft it is just a troupe which releases city direction model video games a number of them having a flare pass of overcome others with out carried out in any blithe precious pixelart format.In the case of Ninja Town your task is usually to jumper cable anyone Ninja Group and bring together Japan.Today for those who are informed about Kairosoft you'll be thrilled to know it comes after to a greater extent carefully to be able to Heroic Astro Narrative in addition to Kairobotica subsequently say Ultra Mall Tale. lol accounts na
lol accounts cheap