I think I need to explain a little more:
Faultless learning means to making the right behavior very likely to occur, and the possibility of a wrong behavior very much zero. So a "wrong behavior" is not likely to be a choice.

I want you to learn to pick on a green circle on the screen and then get rewarded. If I have a red square on the screen also that is just as big as the green circle, you will choose. Then you will find out which one it is by getting the reward for picking on the green circle. The down side is that the red square to be picked on, is in your database. It got in there during the learning process. It has been considered during learning. If now reward is not forthcoming for picking on the green circle - you will then consider choosing (trying ) the the red square. You will make a mistake!

It is not really that I am saying operating is wrong, but I am all for "stacking the deck in my favor"

All I’m saying is I want to train so that my dog can handle as many variables as possible. So if he has to make a choice, he has the best chance I can give him to make that choice. Tasks vary.

Some, like heeling, are more conducive to the development of a habit ... the activity almost automatic.

Some jobs are more complex and involve a variety of tasks and may require a decision on the dog’s part. Therefore, I want my dog to be able to make a good decision. This means I want him to encounter variables; to get used to distractions; to have the freedom to make choices, mistakes, and choose the best decision that gets him what he wants.

Remember that dogs do things for their own reasons and not for ours. That means I want to set the stage so that the decision he makes is the one I want him to make. It also means his “pay”, the reward coincides with his reason for working for me. For my pup right now when we are tracking, his pay is the goodies on the track and especially the “to die for” ones at the article.

So, when I am teaching him complex activities---like tracking---I want to set the stage for him to have to make a decision. If he wants to leave the scent line, then let him. Let him learn the goodies aren’t there. Let him learn that the quickest way to the goodies is to follow the human scent line. That is the right decision. Following the human scent line is the way to pay.

And at some point he must know that he will ALWAYS get paid. As learning progresses rewards become random, further apart, and often involve completing a bunch (chain) of behaviors--including making choices (like a maze). But in the end: my dog will get paid. At a trial the treats or tug or whatever the reward is WILL COME, probably at the car. But, he will learn they will come. And if you want to apply the random schedule premise: pay might not come until the end of the day; or the weekend; or the next work session; or next trial. But it will come. And it will be worth his time spent working and waiting. i.e. after a long wait, a little niblet of Van Patton won’t do it; but a big fat slice just might.

But, I'd never make my dog wait past the end of a trial or a session.

If I teach my dog without the possibility of faulty behavior, why would he even consider another option (leaving the track)?
And if he really would consider at one point - that option of faulty behavior might be very little and unattractive.
I think it would be a dangerous process for him to find out : leaving the track then going back and get rewarded is a chain of behaviors!?
Chains are easily created accidentally.
Dog heels - looks away- looks back at you= gets rewarded. This example might be easy to comprehend?

About the reward :-))
I much rather have a chain build where the dog understands that if he is allowed to perform the next behavior within a sequence, is the reward. If you do it right , you can continue. That is most likely possible if you build a chain backwards. So we are not actually depending on random reward, or "surprise reward". If he never knows when the reward is coming, I think you create uncertainty. This could help to create a very active performance, yet, it will lead to huge and expensive mistakes, because anticipation is very high.

Again, I think there are some tasks that dogs have to do that should require very little thinking on their part ... the task is simple and repetitive in nature. The handler makes the decision when the action should occur and the dog follows the command. Absolutely. Heeling, retrieve, down on command, out, etc. And I think these are better taught by a consistent, habit forming, behavioral approach.

On the other had, some tasks that dogs are required to do, take thinking or require that they make a decision. Tracking, some service activities, detection, etc. I think these kinds of jobs are better taught in a way that integrates some experiential learning into the picture. I want my dog to learn how to form an appropriate conclusion under variable circumstances in a particular kind of situation. Therefore, I want him to “leave the track” to figure out that what really gets him to his pay the quickest is to follow the human scent on the ground. This experience gives him practice in figuring things out and making the best decision for himself. How to make a good decision is something that has to be learned.

I disagree, the decision on the command "sit" whether to lay down , stand, or move, - or for that matter to sit, is to be made as well....in tracking , you say " such" and the dog has to do it, he will then decide whether to follow a track, or do something else.I understand what you are saying, I still believe faultless learning in tracking is very helpful.

I didn't say it wasn't helpful ... I just mean it isn't the only card in deck in my opinion. And in fact may be a lower value card.

Good discussion. Now what do you mean teach the track with no possibilty of faulty behaviour? Do you always train with short lead? And what about loosing the track?

Loosing the track during "teaching phase" should not be an option. Yes I am talking short leash- be right there. Never the less, the treats are so close to each other that the dog would be a fool not following the track flawlessly. I am not talking about correcting a behavior. I would want the behavior be taught without making mistakes. The reason is that one might think "oh he is off the track" - he doesn't find his treat thus he is not reinforced.....well maybe he is off the track and enjoys himself sniffing something else! So he is reinforced for his wrong behavior, and you thought he wasn't!?