I also think that the way horses react to our body language is a different topic though. It has more to do with learning theory, in my opinion.
Although I do not think it needs to be a different topic, I guess I was a bit confusing when bringing up Impie. Actually the point of my post was not meant to be "What makes a horse react to bodylanguage?" but "When are my signals good and when aren't they?" At least for me my cues do work the more effectively and bring about more subtle results the more they take the horse's reaction into account - that is, when they themselves get modified according to the effects they create - instead of being just predetermined cues.
I'll try to make it a bit clearer with an example: Last Saturday I asked Unico to turn his hindquarters away from me by directing my hip there (while standing next to his belly, facing him). If I saw him moving sideways with his whole body and thus the frontquarters moved as well, I counteracted this by adjusting my posture in terms of taking the hip that was closer to his head backwards away from him. Consequently, I created a slight tendency to draw his frontquarters towards me which made it impossible for them to move away from me at the same time.
However, this can only work if I do attend to the effects of my movements (i.e., the horse's way of moving) in real-time, instead of just giving the "right" cue (just attending to my own movements). Perhaps that's also the reason why we differ in our views on this, because I try NOT to train my horses to react to a certain set of cues. It's more like if the horse perceives something to be a cue, then it becomes a cue or part of a cue for me - namely for the response the horse just gave. That's basically what I tried to explain in the R+ for humans video
and text. So I do not have a certain, fixed sideways cue for example but a general idea of what could make the horse go sideways, which is defined and constantly redefined as our interaction changes, depending on what effects it produces. If in response to my current sideways cue the horse moves forward more than I want, I reward just as much but adjust my position the next time I ask. So I guess there is not very much need for the horse to filter out things.
It's only that I tend to forget to work in this way when I get lazy, and then I just perform the move that worked for the last few horses or what worked the last few times with the same horse. And then
it is that my horses become less sensitive. Otherwise I have not seen much of a change in their sensitivity to body language as they grew older.