Wednesday, March 23, 2011

Experiments in reinforcement

Strum's dogwalk isn't getting consistently better so I decided to play around with his reinforcements.  He doesn't seem to care if he misses and in fact I'm convinced that after all this time he's still not clear on what's expected of him.  So in order to try to make things clearer for him and give him some motivation to try to figure things out I'm making his reward vs no reward more obvious.  If he gets a hit he gets treats from the treat gizmo and I come over and give him even better treats from my pocket then I break out his special squeaky chicken that makes him mad with joy and I play with him for a bit.  If he misses I mark it with a verbal 'no' as I've been doing all along and I go sit down on the sidelines for 20-30 seconds.  This does cause him a small amount of stress, probably because he doesn't know what's going on and uncertainty is a stressor for him.  He had a lot of lip licking going on and trotting around as a displacement behavior.  I mostly ignored him while I sat, I did have some eye contact mostly because I wanted to see how he was responding.  I'm not a fan of this type of thing and the original suggestion from the running contact seminar I went to was to take him by the collar and lead him back to the start of the dogwalk when he misses.  I did this at the seminar and it took a lot of reps. for him to finally get a hit.  I don't have it in me right now to try that in practice at home.  It'll suck the fun right out of practice for the both of us and that's not what agility is supposed to be about.  It may be bad dog training but a running dogwalk is not a necessary life skill and as such I'm not willing to start going down that road.  So for now I'll stick to just ending the action for a bit.  If nothing else the little break is good for breaking up a string of misses.  Sometimes he gets into a groove of missing and I can't get him out of it short of ending the session.  He didn't have any 2 misses in a row for yesterday's session, we'll see if it holds up for more sessions.



Stats for yesterday's session:

8/11 (73%)

5/8 hits (62%) were for full dogwalks, the other 3 were back chaining

Set-up was dogwalk-treat gizmo

He's focusing too much on the treat gizmo so I'm going to introduce a jump back in next time. 

Went out to the training field this a.m. and unfortunately the dogwalk had blown over in yesterday's hurricane force winds (it was calm in the morning when we practiced).  No way I could right that thing on my own so it may be a few days until I can get back to training.  Guess it was a good idea to cancel yesterday afternoon's bike ride.  It felt good to have a rest day anyway.

5 comments:

  1. He's just so long-strided! Very interesting to watch and see what he needs to figure it out.

    ReplyDelete
  2. He is long-strided! He mostly gets only one paw in the yellow.

    ReplyDelete
  3. I know, the long-strided dog is my bane. You'd think I would have learned my lesson after the first one. And training Lola was like training a Shetland Pony.

    ReplyDelete
  4. Yay for no back-to-back misses! I love the idea of using his favorite toy to reward him as well :) And even though I'm not a fan of "quiet times" like you describe, I see how it might be necessary, say, if he got rewarded with his favourite toy and missed the next contact because of over excitement - he would benefit from some time to get his head straight again.

    An idea for getting Strum focused ahead: Ruby got nicely used to the fact that toy always gets thrown after the jump, so now he will run ahead even if I'm a bit behind him (and there's nothing on the ground after the jump). What finally worked for us was using two toys - one that flies well, but is lowest on Ruby's list of toys, and one with really high value. I would throw the low-value one and then the high value so he learned not to get too crazy about the first toy, because a better one was comming. This was necessary so that he wouldn't get all crazy and jump the contact when he saw a toy. Then we took this lesson to the plank: I threw the low value toy over the jump every time, but only threw the high-value when he hit the contact. Now he runs over the jump even if I'm late with throwing.

    My current problem is exactly opposite than yours: I want him to extend his stride a bit. It's never quite right, is it :)

    ReplyDelete
  5. Ah well, Strummer lives his life in extension. If only our dogs could find the happy medium : )

    The 2 toys idea sounds good. I'm a little worried about him chasing toys mostly because of the way he slams of the brakes when he gets to them but now that I think about it he does the same thing with the treat gizmo. I think I'll give it a try, not much to lose at this point anyway unless the goofball hurts himself.

    ReplyDelete