One of last week’s blog posted caused a bit of a stir amoungst some of you so I will deal with a couple of comments that were made.
AWESOME!! Love the picture of Feature on the Teeter. I do have some questions, after this “exercise” with Feature and the DW, what is next? Would you continue with your current training, change / add something? Would you know test it in competition before changing a thing?….
Yes Nelci competition was the next step. I actually went to a local trial yesterday and let Feature compete in 3 classes. Her contacts were
awesome. Fast and accurate even though it was pretty old equipment that was completely different than anything she had ever seen before. Since the dogwalks here in Canada do not have slats she has one one issue not realizing the seesaw was a dogwalk and therefore put on the brakes too late. I lied her down, for 10 seconds and then repeated the seesaw (which was brilliant). In Canada we can “train” in the ring, so it was a great place to test her understanding.
How do you know that Feature was thinking I screwed up the dogwalk therefore I don’t get to go swimming? If you subscribe to the clicker training idea that the dog relates the reinforcement/punishment to the action preceeding the r/p, then Feature could have been thinking I am going in the crate cause I wagged by tail, panted, spun around, layed down, etc. I think there was way to much time between the missed contact and the final R/P. The fact that he stuck it the next time was probably a learned behavior that if I hit the contact in a particular way I receive a reward. He may have lost some of that original enthusiasm and therefore was more careful of his performance.
Good bit of controversy for the blog Paul, thank you. Let me say that rather than subscribing to a “clicker training theory” I subscribe to the laws of learning. By that I mean the laws of classical and in this case, operant conditioning. As a scientist I hate to quote studies without naming the source but here I go (any scientists reading this that can add as source, please do). There have been studies in operant conditioning where the time lapses in the delivery of reinforcement and punishment have been evaluated. I can’t remember the specifics but one study in particular had the rat perform a task and then it was held in a small area for set duration of time before the reinforcement was delivered (eventually tested well over 1 minute if I recall). Yes, it still worked and the rat was able to be shaped.
Having said that yes, the absolutely hands down best transfer of information to the dog is that which happens immediately. The ideal application would be at the moment the choice-point was made in the brain of the dog. By that I mean, perfect timing would be before the neurons transmitted the signal to the muscles of the body for the dog to leave the dogwalk. . . if that could be arranged by any of us! So being held to that standard, in fact, we are all “late” with every application of reinforcement or punishment we apply, aren’t we?
Bob Bailey told me about a blind lady came to one of his chicken camps that managed to shape a chicken to peck at a target entirely based on sound. Yes the timely would be incredibly late as she would have to wait until she “heard” the peck, then react and click and reward. However the chicken was able to be trained to do the behaviour.
How I can be secure in knowing Feature isn’t pairing the punishment with a tail wagging or other behaviour is because of the extensive history or reinforcement for the discreet behaviour of nose targeting. Knowing she started nose targeting at 4 months of age I can tell you she has had 22 months of reinforcement for this behaviour prior to her first ever time out. The tail wag, although I am sure paired with reinforcement of my attention in the past, is certainly not the same black and white predictor of reinforcement from me, as it has never been discreetly isolated as a response to produce incredibly high value reinforcement for her. A great nose tap on the end of a contact will always be rewarded with either a reinforcement from me or the chance to do more agility.
Now lets look at the “application” of the punishment. Was it really that poorly timed? Take into account what Feature sees under normal conditions when she is performing a contact correctly, she would wait for me to give a release cue. I may be moving or standing still, far away or close, but one thing is consistent and that is the release cue. The release cue will always be part of the “reward process,” usually the last step. I would either deliver a toy, a treat or allow her to move onto the next obstacle.
Feature would have seen none of that when she self-released. Had I kept running even though she didn’t perform correctly, it would have established a new avenue for her to earn reinforcement and given her another reinforced behaviour to “try” on her contacts. So the lack of the “normal” picture the moment she self released would have been the first thing that communicated to her that “something is different.” My actions that followed immediately after, (that of me taking away the reinforcement value for her with the time out) would then communicate that “the different is not good.”
Finally, although I have no video to show you, you can trust me when I tell you, the second performance of her dogwalk had all of the enthusiasm of the first one (possibly more as the other dogs were at the pond having great fun without her!).
Whew, that is a lot of info, perhaps more than is necessary for many of you but I think worth going into detail either way.
Today I am grateful for a new fountain we had installed in the dog’s pond yesterday. Very cool. However Buzz is slightly confused, he keeps staring at it wondering how can ANYTHING make that much noise and splash that much water in a pond. B-U-Z-Z-Y! Clearly he has never seen himself swimming!