chaining working behaviors using a clicker

I usually change the criteria once the dog has offered the behavior five times in a row.  So, if I am started a retrieve, I start by c/t the dog touching the object with his face.  If he does that on his own five times in a row, then I wait until he actually bites it.  When he bites it five times in a row, then I wait until he bites it and takes it.  Then I train the hold in increments of seconds rather than repetitions but overall I just look for the dog to show a very clear understanding of what I want and to offer it several times in a row without messing up before moving on.  It's amazing how much you can backchain.  Nikon went from tapping a dowel with his nose to doing a retrieve on flat, retrieve over 1m jump, retrieve over a-frame, and send out with a down.  He now gets the reward after holding the down on the send out.  It's all one giant behavior chain.

that's what we are doing right now - back chaining our refind and I wanted to put a clicker to it. She gets distracted on the refind, which is the problem we are trying to solve. So, I'm building drive for her kong by giving it and taking it away, giving it and taking it away and also chaining the come to me and bark then go back to the vic

 Here is a great video on YouTube regarding chaining commands with clicker techniques.  There are many others as well, simply search with the search terms-- clicker training chaining .  I hope this helps!

 

http://www.youtube.com/watch?v=6oFO9Z0oHBA

For complex behaviours, Julie, you often will get better results, faster, using backchaining.  It really depends on what you want to teach, though.

Take Kleenex from a box for example. I would do something like:

Put a Kleenex on the floor.  Click/Treat (C/T from hereon out) the dog for looking at it.
Then C/T for nosing it.
Then C/T for picking it up (if your dog has a "take it" or "pick up" cue already, you may be able to skip the early steps).
Then C/T for bringing it to you.
Then C/T for dropping into your hand.

When you have all of those behaviours, move the Kleenex to a new location on the floor. C/T for retrieval.
At this point I would place a Kleenex box on the floor, and place the Kleenex near the box. C/T
Then place the Kleenex on the box. C/T.
Then pull a Kleenex almost all of the way out, but not all the way. C/T
Finally, have the dog pull the Kleenex out normally, and deliver it to hand. C/T.

This process actually can go very fast, many steps you'll fly through. After that, it's just about moving the Kleenex box to a location (like a table), putting it on cue, and practicing!

 

  Honestly, a set number is  something I don't always pin down across the board for all animals I'm working with (and I  say animals and mean both human and other animals). One of the ones that I might use fairly frequently though is that if I come back the next day, and the animal picks right back up where we left off, we're ready to move on if it's something more complex that I might be doing across many days. If I'm going to shape it pretty quickly, I might really not allow for more than 3-5 trials with any particular response requirement. If you go on too long at a certain requirement, they start to think that's the terminal response. To the same end, I won't allow more than that many successive failures. That also shuts the animal down. So, you go back to the last point of success.

If you really want to know exactly when to start getting rid of the clicker, and thinning out your schedule of reinforcement, you can make it a data based decision, though  I never do with my non-human animals. I would say at a more general level, if I have performance that is maintained with no practice over a period of time [could be a day, could be a week gap- I would make my decision based on what type of gap I will be dealing with when I have completed training (if  my animal will do it at least daily, like sit, I use one  day, if he will do it weekly, I use one week, if it's  less frequent I would still stop there)]. Also, if you need the behavior in multiple in environments, and I assume you do and will need it in novel environments,  test that too. For me, once I have good performance maintained over time and across environments, then  start thinning the schedule of reinforcement. I'm thinking for the type of skills you are training, this performance maintenance over time without practice and across novel environments are going to be critical things (I'm not totally familiar with SAR, but I am under the assumption she must be able to perform in novel environments).

Yep, novel environments with every distraction you can think of. I appreciate all the feedback here including the step by step tissue ct process. This has helped me think through our steps, successes and failures. What i have done since posting this is find one non distracting location to run the beginning of our backchain and use the clicker to mark each correct step. She picked up on the clicker mark right away even though up until now we havent used it in our sar training. The first time i clicked, she totally head snapped to me like OH! The sound of the click seemed to help reinforce the backchain. So, we did about 10 in a row at the one location spread over several days. Then we altered the location slightly and did 6 to 8 there. Then we went back to the spot she knew and did 6 to 8 there. That seemed to reinforce it even more that this is exactly what you are supposed to do. Then we moved to a new location again then went back to the spot she knew.. each time she would falter very slightly at the new spot, then get it perfect then when we went back to the old comfortable spot her confidence skyrocketed. She looked forward to going back to that old comfort spot. Btw, whenever she would falter i would stop the excercise immediatly and put her in a down stay for several mins before trying again. She hates down stays, especially in environments she doesnt know. So, that was like the worst punishement ever in her mind. She learmed that if she ran off to sniff she got a down stay but if she did the back chain perfect she got a kong filled with goodies.

Today i took the clicker out and added in the search part of our behavior chain (very short search). I thought we would crash and burn because we are in a highly distracting novel location, i removed the click and added a step all at once. She was full of energy and drive to earn that kong though. It was perfect. To test her overall progess, we later did an actual search. She had one moment where i could see her mind falter for a split second before she made the decision to come back to me for the indication. So that tells me we need to go back one or two steps and do some more reinforcing. Still, it was a huge improvement over a few weeks ago before we started this process. I think we are on to something with the pushing forward slightly then stepping back to the comfort zone, rinse repeat..