I agree that learning and retention are two different things Mark.
I'd be quite surprised if that what I said, retention testing is a pretty common way of checking learning has taken place, they're very closely linked. Transferability, that is the capacity of a learner to apply what they've learned to new challenges, is also a mechanism for checking learning has actually happened. Performance during a lesson and retained learning, however, are very often different, hence the question in the other thread.
It is not safe to assume that because someone can perform a particular cast with instructor feedback in a lesson has actually "learned" how to do it. This means that drills that promote rapid improvements in performance in a lesson may not be promoting retention quite as well. Clearly, as we've discussed previously, there are many factors that are going to influence how an instructor is going to teach but the methods that we employ, whoever we teach, should be directed to maximising retention and transfer. This means, even though it is probably counter-intuitive, that methods that deliver fast results but depend on instructor feedback or intervention to work are going to be less effective than slower methods which rely more on the caster using independent feedback to work.
This often means that an instructor has to teach a caster how to engage with the feedback available to them. To do this effectively both the caster and the instructor have to be able to describe, observe and react to the same things. The instructor is at a huge advantage as you have pointed out in that they already know where to look and collect a huge amount of information from places where the caster themselves cannot look, ie their own body motion. However, as a very basic rule of thumb, if the caster can't use it for feedback purposes, it's probably best not to draw attention to it.