Training Logs‎ > ‎

Recall Cues A and B


Over on the yahoo list clickersolutions, a lively discussion about poisoned cues is underway. The only research supporting the hypothesis that R- will indeed poison a cue is a single-subject experiment that hasn't yet been published.

Because of my own scientific training (as a behavioral ecologist in a zoology PhD), I'm not convinced that experiments with n=1 can be generalized to other individuals, but that's ok for now, because I just want to see how Vanya responds to a new recall cue trained entirely without R-, versus a new recall cue trained with 20 to 30% trials backed up with R-

The R- will be pressure from a long line when he doesn't immediately respond. (For those who get confusing about operant conditioning terminology, when you remove something (negative: pressure on the line) to reinforce, or increase the frequency, of a desired behavior (running toward me), that's negative reinforcement.

Here's what we'll do: train two new cues (let's call them A and B). One I'll train with R-; one without. Start from the very beginning with each cue. Train each with increasing levels of distractions. 

Then have a neutral handler (known as The Friend) come in for the experiments themselves. My friend won't know whether A or B was trained with R-. She will give the cues, under increasing levels of distractions, and she will record how promptly, how enthusiastically, and how reliably under distractions, Vanya responded to A versus B.

(Then I might train the same cues for another set of time, to see if the responses change with greater levels of proofing. Or I might see if either cue works well as a conditioned reinforcer, useful for backchaining--as Murray 2007 did in her study.)

Nope, this isn't scientific, but it will help me learn what works better for my particular dog.

Equipment: harness for both cue words; long line only for R- cue word (inside fence for both cues)

Training protocol for R+ (similar to other recall cues he has learned, based largely on Nelson's RRR): start by pairing cue with c/t at a distance of 0 ft from handler; then increase distance (10 ft, 20 ft, 30 ft, 50 ft, 75 ft, 100 ft, 200 ft, 300 ft) while holding distraction levels constant, then increase distraction level at each distance level. 20 reps at each level. When all 20 reps are followed by desired response, go to next level (ie, 100% response rate is the criterion for moving to next level of distance or distraction). 

Training protocol for R-: same at early stages. After distance=50, instead to the next level of distraction when he has 80% at lower level, and use line pressure for the other 20%.

Training will be done at meal time, when his motivation is highest. Each day, I will alternate which cue gets trained at which meal (so A will be at breakfast one day, and at dinner the next day). 

Cue words: well, one of them rhymes with his "really reliable recall" cue. Do dogs hear rhymes? Hmm. 

Rewards and motivation: will ideally remain constant throughout. For both cues, each reward is one piece of his regular kibble; motivation is dinner-time (ie, he'll get fed via this training, and he's an extremely food-motivated dog, so this tends to work fine for training new cues). 

If he misses a response for Cue B, he gets P-: no c/t (and I lower the distraction or the distance and repeat at an easier level until he gets 100% response rate). Hmm-but what do I do to keep him from getting his distraction (stuffed kong, friend, etc), and therefore being rewarded for noncompliance? 

If he misses a response for Cue A, he gets R- (a tug on his long line until he begins to come, followed by a c/t when he gets to me). I move onto the next level once he has reached an 80% response rate.
Showing 0 items
R- count?
R- count?
Showing 0 items