If you haven't been to a Kathy Sdao seminar yet, quit reading this blog and go find one. I'm not kidding! I have seen her three(?) times and I think she's fabulous.
Okay, this weekend was an advanced clicker training seminar, and between lectures and dicussion we did dog work focusing on discrimination and adding cues. I had been a little worried about taking Laev to the seminar -- I thought she might not be advanced enough -- but I thought she needed the mileage and it would be good experience. Well, she did better than expected. :-)
First, Laev learned to target a ceramic saucer from among a hat, book and saucer, no matter the arrangement of the array. (Kathy suggested we wear opaque sunglasses to reduce the additional cues from our eyes.) She also learned to touch a plastic bottle, from an array which did not include a saucer. No sweat for either, of course. Then we added a cue for touching the saucer or bottle -- that broke the flow a bit, because she didn't understand my new action, but she pretty quickly caught on to that. I was flashing a matching object as a cue, and at first she tried to target it (it was, after all, a saucer!) and was confused when that wasn't reinforced, but after a single session (10 reps) she got the idea that flashed saucer = go touch saucer.
When she had learned the cue for touching the bottle or the saucer, it was time to put the two hot objects together and see if she could discriminate between the cues. This may sound easy at first -- hadn't she already learned the cues for these objects? -- but of course it's not that simple! All the cue had been 'til that point was permission to go and touch the hot object; it had nothing to do with identifying the object, even though it might appear that way to a casual observer. Of course Laev would go to the saucer when shown a saucer -- why would it even occur to her to target the book, when that had never been reinforced? But going to the saucer when the bottle was also available, and it also had a reinforcement history, now that was challenging.
This is why when clients tell me that their dog already knows "sit," I don't always believe them. :-) Yes, the dog probably knows to plant his rear end, but he probably doesn't really understand the cue like they think he does. Can he distinguish between 3 or 4 or more cues for different behaviors reliably? Then he knows "sit." Otherwise, you're just getting lucky in that you're asking for what he's already guessing.
Laev didn't quite "get it" in our limited training time. I have data sheets showing success rates of 40-100% in various sessions, but she could not consistently stay above 80%. But given the amount of time we spent on this, that was to be expected; none of the other dogs present really nailed it, either.
I'm going to try to continue this work at home and see if we can get a solid "match to sample" behavior under our belts.