you wrote: My dictionary defines Epistemology as “the theory of knowledge" well, these dictionaries nowadays.  If you can't trust that they all tell the same thing, why pay for them?  So I consulted the free website "encyclopedia.com"...  Here is what I found: "epistemology [Gr.,knowledge or science], the branch of philosophy that is directed toward theories of the sources, nature, and limits of knowledge. ... one of the fundamental themes of philosophers, who were necessarily obliged to coordinate the theory of knowledge with developing scientific thought." (of course, there is a full entry on that website, but if you need more, go and look for yourself.)  As is often the case, NLP trainers may simplify things a bit...  Seen from a philosophy of science point of view, the NLP statement “How we know what we know.”  misses the last piece, which says "in such a way others believe that we know" (in other words, I can believe something like things written in the Bible being "true" (just an example, Jesuits taught me that the bible may not be true, that you had to read it as metaphors.) but scientists won't accept the bible as knowledge and only teach the biblical version of the creation in stead of the big bang when an American Judge tells them to do so. When I took the "modeling & Epistemology" class at NLPU in 1997 (taught by Dilts, Deloziers et al), I (and several other participants with Master degrees) were disappointed by the apparent lack of understanding of the field of philosophy of science by our "faculty" (boy, did they have a hard time, they changed the name if the program the year after us and added a week to the course) - anyway, I think that's where I got the idea that there was a real PhD in this... (the official working title for my PhD is "The Epistemology and Methodology of Neuro-Linguistic Modeling").  Until now, the closest I got to "philosophy of science" in an NLP context was when John Grinder cited Kuhn's book "The structure of Scientifc Revolutions" or when Robert Dilts said: "The epistemology of NLP is probably the closest to that of structuralism" As Dilts and Delozier write in their "Encyclopedia": NLP emphasises modeling and flexibility as the core of its epistemology".  So in my opinion NLP III is when we are discussing how to make a model and what makes a valid model "valid" - it's literally writing down the epistemology of NLP.  For me, that's different from doing modeling and from using other's applications.  If I understand "Whispering" correctly (see p.10) Carmen & John are collapsing the 3 levels I propose, sometimes even into one "thing".  In stead of looking for the epistemology of NLP (as Dilts & delozier try), Carmen & John say "NLP is epistemology". For instance, they write: "We propose that NLP, both in its core activity, modeling and its applications can be usefully understood to be a higher order operational epistemology." Before you fire all meta-model questions to this sentence, relax & be patient, they set out to explain that statement, but unfortunately then leave us a bit in the cold, saying that the full argument will be the topic of their next book, which should have appeared in 2002.   Anyway, until that book appears, the working hypothesis I'm using for my PhD is that the statement of Bateson "shoddy epistemology" still holds. you also wrote: to paraphrase John, if you call these 2 levels, explain the ordering principle - I'd rather call it 2 modeling methods.  A third modeling method (for me) is: take 3 people (a least) being considered the experts or the top performers for a certain job, and throw some jobEQ questionaires (such as iWAM) at them, and then figure out how they differ from the "bottom performers" - Nope, that's not "NLP modeling" as it is currently known (I take the responsibility for living on that planet). the Steve Andreas example you give: I would call making those models NLP level II - if you go to a seminar to learn to apply them, I would call that NLP level I. the video-tape example: another approach of NLP level II - I agree that unconcious modeling will go faster - my explanation for that is that the rule 7 +/- 2 does not apply to our conciousness.  But for me, unconcious modeling will only work well *after* we learned to look for the distinctions: several recent experiments in cognitive science make me believe that if our neural network (aka Brian) is not trained to distinguish things they just won't "detect" them.  (just as new-born babies are said to be "blind" because they didn't learn to process forms yet) Hope this answers some of your questions and clears some of the confusion - keep asking questions, it helps me to better structure this when I put it on paper (that's what's good about having a book circulate for months or even for years before really publishing it). Pardon my cynicism at some points in this mail, it's late at night and it has been almost a week since I tried to figure out what happened in "the NLP community" between 1978 and 1982 so that we ended up with something that is now called metaprograms.  Maybe it was the wrong week to ask, but at this point I'm not sure I'll ever find out (for those that "believe" in metaprograms, proactivity is one of my drivers, and sometimes lack of patience (the related metaprrogral) is one of the downsides).  At least I can derive (cold) comfort from the idea that "for a researcher is that the fact of not getting an answer or doing an experiment that fails is also noteworthy." - but I'll persist at least till the end of january before I give up... I learned by now that many scientist have low scores for the proactivity parameter when they fil out iWAM, and I learned from that "model". Patrick www.merlevede.biz