I'm afraid I also have to disagree with the concept that the click itself after a while becomes a reward. A reward = something nice and meaningful. I like seing my paycheck envelope because I know what's in it. The envelope itself is irrelevant if empty!
Clicker training is a presicion tool if used correctly and works because the learner is going to repeat an action that it gets rewarded for. The method is based on trust, if the dog hears the indicator noise (click) the reward for doing the action is going to follow shortly. When I am training I do sometimes mark totally the wrong thing. But if my dog hears that click he gets his reward, no matter what. As SheltieChe says, withholding the reward because the action was not what I was looking for will mean the power of the clicker as a marker is diminished and he will start to lose faith that the click ALWAYS indicates a reward. Why is he then going to be motivated to earn that click if it doesn't pay out?
For an example of expanding/enhancing a behaviour, when I'm teaching heelwork, I start off by clicking and treating every little good thing so the dog learns his position is by my side. Once the dog has the idea, then I start witholding the click for actions that are less good (right position but too wide) and only marking the best ones (in close)- that will improve the behaviour because the dog learns that only if he is right on my leg does he get that reward. Once I have the position satisfactory, then I will start witholding the click to maintain it for longer - so to begin with he gets clicked for each step, then every other, then every third and so on - but to keep motivation I will click and treat after one good step every so often (about once in 10-15 clicks). In this way the dog works extra hard because he doesn't know exactly when that payout moment will come. I've actually found with the more experienced dog that he 'ups the ante' if he doesn't hear a click for a bit because he really, really wants to get it then.
It is a balancing act witholding the click long enough to get to the 'OK I'm really trying now' point but not long enough that the dog stops trying and not giving so frequently that he isn't stretching himself. This random rewarding plays on the dog's gambling mentality to extend the behaviour.
My understanding of clicker training, having studied quite a few books and DVDs on the subject, is that it is a means of
teaching the dog a new behaviour or behaviour chain. It is not something to be used as a reward on the completion of a seperate series of behaviours. So I would use the clicker to teach heelwork, to impove the position and to teach the dog to maintain the position over an increasing period of time. But I wouldn't go out and do a round of heelwork with sits, turns etc then click the dog and give him a treat at the end of it. There is no way that every single one of those actions will be perfect and therefore clicking the dog after a couple of minutes performing different actions will be meaningless. Nor would I click after completion of an agility course for the same reason. In these situations the dog is performing a series of seperate behaviours that it knows how to do and therefore I am not going to be teaching it, but I will be rewarding it at the end. The dog in this situation needs to have an expectation he will get a good payout at the end so you build this into training - once the behaviour is correct.
I think it is very sad to see dogs in competition come out of the ring and the handler witholds the reward because perhaps the dog missed a jump or broke the wait on a recall. If he has done the majority of the round well he will have no understanding of why he isn't getting his reward and the disapointment is evident. Where is the motivation for him to work with enthusiasm again if he is performing a complicated set of behaviours but with the liklihood of no reward? If I wasn't paid one month because I had made an error but no one told me that I'd done something wrong then I doubt I would feel like bothering to turn up the next month!
However, that doesn't mean the clicker can only be used to teach one action in isolation as you can use it to build a behaviour chain.
As an example, one of my favourite things to teach is a sendaway to something and lay down. I shaped Epic to do a basic sendaway to a mat 2 meters away in ten minutes with one and a half sticks of 'Schmacko' using the clicker (& backchaining). I will add that at the time, he had no prior experience of shaping (figuring out how to earn the click rather than being lured). So the dog learns to perform several actions (look at mat, run to mat, lay down )to get one click.
The important thing here is that they are all part of one behaviour chain in response to a single cue - away.
What I wouldn't do is add in a recall off the mat before clicking unless I wanted the dog to run to the mat, down and then return off his own bat. My recall is a seperate behaviour in response to a different cue and I don't want the dog to take this action unless I ask him.
So my rule is clicker for teaching a behaviour, enhancing it and incorporating into a behaviour chain. Click is given at reasonably frequent intervals and always followed by treat.
On something like obedience/agility round where dog is performing several
seperate actions on my cues for an
extended period of time he will get a jackpot reward at the end,
but no click.
I think the basic principals of clicker training are quite easy but it takes considerable practice to get the skills necessary to use it to teach difficult/complex behaviours and to improve a behaviour. I've been clicker training for about 8 years now and am a lot better than I was but a long way still to travel.
