Thursday, July 16, 2009

Principles of the Adventure Drill

[The following essay was posted to three lists this morning: DogTrek, PositiveGunDogs (PGD), and Animal Reinforcement Forum (ARF).]

For those of you not following the online training journal I maintain for Lumi and Laddie's retriever training (see signature line), we have been experimenting for several weeks with something I call the Adventure Drill (AD). It simply means running land-water-land (LWL) retrieves in the most difficult terrain I can find: thick and tangled underbrush, logs and boulders as obstructions, fast-moving streams, steep embankments, and so forth.

When I began the AD, my goal was to habituate Laddie, my younger dog, to distractions, especially water, which historically have tended to divert him on returns during events. It was a difficult problem to solve because Laddie didn't display those distractions in private training, and of course you can't suddenly transition into training mode in the middle of an AKC test. But it occurred to me that if I made our private training conditions significantly more distracting than at events, perhaps I'd be able to trigger the behavior and then use management, +R, and if necessary -P to address it. The retrieves I was using were so arduous that when I first introduced the AD while out training with both dogs, I didn't run Lumi, my older, more fragile dog on the same retrieves, but made hers easier. Lumi didn't have the distraction problem and I didn't want to subject her to risk of an injury or associating unpleasantness with retrieving.

It turned out that almost no "training" was needed to overcome Laddie's distractibility during the AD because even in those conditions, he showed little evidence of it in private training. On one occasion he did stop to swim around in the creek and I was able to immediately approach and get him going again, and perhaps that did help. In scores of AD retrieves since then, he's never shown the behavior again. More hearteningly, he also performed well in the only Senior Hunt Test he's been in since we started the AD, repeatedly ignoring the exact kinds of distractions that might have diverted him in previous tests.

That's all nice, but it's not surprising. That's a typical clicker-training procedure, called proofing. It's also common in all sports to train at higher criteria than you'll need to compete at, in order to compensate for what Alice Woodyard calls the Event Discount Factor, which means that dogs (and their handlers) don't tend to perform as well in events as in practice, sometimes going so far as to behave diametrically opposite to their normal practice behavior.

But something else happened with the AD that I didn't expect. Laddie had the remnants of another problem besides distractibility when we began the AD: Ever since he was about a year old, he's had problems getting into the water while carrying an article. That of course comes up during the return of every LWL retrieve, and since many events incorporate LWL retrieves, has stalled Laddie's competitive career. We worked on it all last year, and largely repaired it, but it still came up occasionally in several tests this spring. Of course we were continuing to work on it in various ways, and we would have applied those to the AD as well.

Except for one thing. It stopped happening when we started the AD. And it also didn't happen in the test Laddie took after we began the AD.

So here was a behavior that was NOT fluent at one criteria level, that BECAME fluent when we raised the criteria, and then REMAINED fluent when we later went back and tested at lower criteria again. I tried to think what precedent I had for that in my experience, either in my dogs or in myself. I couldn't think of one.

I considered the possibility that it was just coincidence. Maybe Laddie's LWL problem was on the verge of being completely solved anyway, and we just happened to get into that territory at the same moment that we began training with the AD. But that's not what it felt like.

After another couple of weeks, I suddenly had another thought. Lumi also has a problem we've been struggling with for some time: She tends to dawdle on her pick-ups. Recently, that problem had taken a turn for the worse, when, in her last two events and also in one group training session, she actually dropped her shoulder and began to roll in the grass for several seconds before finally getting back up, picking up the bird, and completing her retrieve. She still got a ribbon doing that in one event, but the next time, the judges dropped her for it.

Once again, despite a great deal of work on the problem, Lumi was not fluent for fast pick-ups at relatively easy criteria. Did it make any sense to dramatically raise criteria at that juncture?

Well, I tried Lumi out on the AD, and it worked. Lumi wends her way thru the difficult terrain, gets to the bird, perhaps shakes off or looks around a little, then picks the bird up and heads straight back. Her old behaviors of flipping the bird over and over, or pick-it-up/put-it-down/shake-off/pick-it-up-again, seemed to almost disappear. Lumi's improvement hasn't been as blanketing as Laddie's -- Laddie seems to have virtually no problem with LWL re-entries any more, whereas Lumi still exhibits a little dawdling -- but the improvement has been dramatic. Also, mystifying. Why should this work?

Oh, the AD also seems to have played an unplanned but logical habituating role in Lumi's development. She seems to be having less and less problem dealing with underwater debris like that encountered in the "stick ponds" sometimes used for event water retrieves.

So what have I learned from this experience? Obviously, I've learned that it seems to be a good idea to run retrievers in difficult terrain at some point in their training, as a way to temper their skills and kind of glue everything down.

And of course the experience confirms Bob Bailey's central training aphorism: If what you're doing isn't working, change your behavior.

But what about the more precise "principle" the AD seems to suggest: If a behavior is NOT fluent at one criteria level, RAISE the criteria level in order to achieve fluency.

No, that wouldn't seem to be a good plan in general. Yet in this particular instance, it seems to have worked, and maybe there really is some underlying generality.

Here's one theory. Perhaps the AD retrieves are not only more challenging, perhaps they are also significantly more rewarding -- hence reinforcing -- than our more usual practice retrieves. So by raising the criteria, I also happened to be raising the value of the reward for correct responses.

Now that might be a useful principle to know. Like if a smart kid isn't doing well in school, try putting him in tougher classes.

Lindsay, with Lumi & Laddie (Goldens)
Laytonsville, Maryland

Field training blog: http://lumi-laddie-test-series.blogspot.com (see "Archive of Video Blog Entries" in right margin)

YouTube playlists:
-- Lumi: http://www.youtube.com/view_play_list?p=BC338082E0B890DB
-- Laddie: http://www.youtube.com/view_play_list?p=9A44913FB240932A

To further explore the frontiers of dog training, join our DogTrek list at: http://groups.yahoo.com/group/DogTrek

No comments:

[Note that entries are displayed from newest to oldest.]