The Hound and the Mitt

As faithful readers of DogTrek will know, Lumi and I are in the middle of a shaping challenge proposed by Sue Rivers: training our dogs to retrieve an article, cued by seeing the article's mate.

Here's a video of our current stage of training in this challenge, after roughly a month:



As you can see in the video, there are two articles ("targets") on the floor, in this case a mitten and a chess piece representing the Hound of the Baskervilles. With Lumi a few feet away, I hold the mate of one of the articles directly between the two targets and cue "match". The correct response is for Lumi to pick the hot target up in her mouth and "hand" it to me.

THE PLAN
At this stage in the training, I am as careful as possible to give Lumi no information other than the mate itself to help her make the selection. I use no pattern of when I change hot targets -- sometimes after one rep, sometimes after two -- and I try to move the mate on a plane exactly bisecting the two targets, giving no hint in the mate's movement which target is being "aimed" at.

My intent is to click the moment Lumi makes the correct selection, but in order to discourage testing -- Is this it? Is this it? -- I don't want to click if she touches the article without picking it up. In this video, I get trapped at one point by clicking and then realizing she didn't actually pick up the chess piece. It's the usual wrong-click dilemma. You don't want to dilute the meaning of the click by not following with a treat, but you also don't want to reinforce the incorrectly clicked version of a behavior any further by feeding. I chose to wait for the behavior to be finished and then fed.

Oh, you can't see it in the video, but for each rep, I have a piece of cheese in my mouth. A split second after I click, I lean over and Lumi takes the cheese from my mouth with her sweet little tongue and lips. You might not want to see that.

TARGET BIAS
Although Lumi has a high percentage rate of correct responses (10 for 11 in this video), she is still not perfect in this skill, and I don't know if she ever will be. Part of my goal as a trainer is trying to understand what's confusing her at any moment and training past each particular difficulty. In this case, the biggest issue seemed to be preference for one article, what you might call "target bias". We've seen this often before in this project.

What happens is that in any particular session, one of the articles tends to be easier for Lumi than the other. In this video, she was more confident about the mitten, although when the session is over, she's under the impression that I'm going to let her disappear somewhere to chew on the chess piece.

So Lumi doesn't necessarily like the mitten better, she just got stuck on it in this session for some reason. We work with about eight different articles, and I can't always predict which one she'll get stuck on for a particular session. I try to select two targets of nearly identical attraction to her, and the better I do at that, the easier the session is for her.

ROOM FOR IMPROVEMENT
It's interesting, and for me dismaying, to hear the sound of one's own voice. In listening to the sound track on this video, I sound neither as thrilled as I actually am when Lumi gets a match correct, not as indifferent as I actually am when Lumi gets one wrong. Even more disconcerting is the distracted-sounded "I love you" to DW Renee as she's leaving the house near the end.

All I can say is, don't believe your ears. My heart is filled with love for my wife and my dog, my voice just doesn't show it very well at times.

One more thing to work on.