Tatze has always been a "show me the money" girl. It's because of the way I have trained her. Gypsy wasn't very food orientated and would do anything for praise - so it wasn't an issue for her, but it will be with Twiglet if I am not careful. My supervisor gave me this advice, so I thought I would share it with you. Teach any new exercise by rewarding every time your dog is successful. As the behaviour becomes reliable and predictable, phase down but not out the reward. Do this by varying the frequency of the reward. Begin by rewarding very second, then every third time she performs successfully. Then begin to switch to a more varied frequency, for instance 3 repeats, then back to 2 then up to 6 etc. Keep it varied so that your dog is unable to predict the frequency but is convinced it will come if she keeps offering the behaviour. Keep her guessing. This is called an intermittent reward schedule and is used to strengthen your dog's newly learnt behaviour. It's like switching on the compulsive gambler button in your dog and makes training much more fun with the buzz of when that reward will come.
What do you mean "show me the money" - you mean she has to see the treat? I don't think that's about whether you fade out treats or not - that's about making sure you are not bribing her. Or maybe you mean the presence of treats has become part of the cue? I think different trainers definitely think different things about fading out rewards (broader than treats). I don't fade out rewards in the way you say, as the trainer I use just does not think the gambling effect works "in real life" when the dog is surrounded by things that it could find rewarding. I ask for more, and I have heard some people speculate that cues themselves become rewarding if rewarded enough (sort of a tertiary reinforcer - cue - action - reward), but I don't not reward particularly if repeating an action because the absence of a reward is a clear signal to Charlie that he didn't get something right. It can actually distress him a tiny bit. So once Charlie has learned something, I'll ask for that and then something else he has learned, and then reward. Eg go to placeboard, then turn left - then reward. So I build up things into chains. But I don't stop rewards and I don't use fading schedules.
Thinking about it - Homer is definitely a " show me your money " kind of guy. Will do anything you ask if I thinks there's something in it for him, no treat no game.
If you get this result, you need to change how you use food in training. http://totallydogtraining.com/are-you-bribing-your-dog/ As this article explains: If your dog won’t sit unless he knows there is a food reward within the immediate vicinity, sadly you have not trained your dog to sit at all. What you have been doing, is bribing him.
I use exactly the same approach as Julie. Sometimes my dog has to do more for a reward, and sometimes less, but he always gets one. The 'unpredictability' lies in my dog not knowing how much he has to do this time, and so he keeps on going till I say 'that's enough' (by marking and rewarding). If I ask my dog to do a chain of things then it's things he already knows well. For learning new things I reward every small attempt/improvement. I too have doubts about the effectiveness of the 'gambling' thing. It works in the lab when there's nothing else to do, but I don't think it's enough in the real world where there are competing and guaranteed rewards (sniffs, other dogs, other people, stuff to eat on the ground....).
I wonder if the fading of rewards is more specific to food rewards and trainers who are offering other rewards was such as retrieving or the promise of retrieving in a little while.....
When we do anything in a new location, or with new distractions, we go back to a reward every time then very slowly fade them down, but never out
Am I right in assuming once a guide dog is working it won't get treats for doing what it's asked or not often anyway ? If so maybe that's why your supervisor uses the phasing out method because eventually the dog won't get a treat every time even if it does a massive chain of learnt behaviour like crossing the road at a pedestrian crossing. Guide dogs are amazing.
