John Grinder asked (about algebra): Can you imagine an appropriate application of statistical (especially, aggregates and averaging) operations to reveal something about that set of formal rules? " Using statistics when there is a clear pattern already and this pattern is accepted by the scientific community does indeed make little sense.  However, here are some uses that make sense (even in algebra): 1) if we can't give a mathematical proof of a pattern, if we cannot describe the pattern correctly enough, statistics may be of help.  Statistics (averages & aggregates) doesn't help to analyse or describe the pattern, but helps us prove there is a pattern. (as argued elsewhere, lack of proof doesn't always prove there isn't a pattern)  2) If we want to "sell" our pattern to other scientists (e.g. in psychology), backing up patterns we cannot prove (using accepted rules) with statistical evidence makes sense - it's a matter of building rapport. 3) some statistical techniques (such as linear regression, factor analysis, ...) do help to find patterns.  Not that I'm entirely satisfied with it, since (unfortunately) outside the field of mathematics the biggest problem after doing factor analysis is making sense of the results.  I agree with your caution in another tread that statistical analysis may fail to detect the fine distinctions ("little" differences that make a difference).  To me that is caused by using a coarse-toother comb.  I refute that this would imply that "statistics may destroy suble distinctions" - statistics doesn't destroy anything except "credibility" of those that do not want to fit inside this paradigm or cannot develop a comb which tooths are fine enough to prove their distinction is a difference which makes a difference.  Once we know the "suble distinctions", I believe we can design an experiment in order to deliver statistical proof of the destinction. When using statistical analysis on known "examplars" using known "distinctions", statistical analysis can help to detect which distinctions are distinctions that make a difference in that case.  I agree with the point of view that this method may be considered a coarse-toothed comb, but considering them a "waste of money" or "waste of heat and time" is something else: applying a fine comb that one must use manually may be less efficient to move a load of sand than using a bulldozer.  While the metaprogram application "iWAM" may be a bulldozer compared to the precision NLP modeling you advocate, I would like to see an epistemology for a field including both.  For me, I think that we still haven't dealt with the points being made in the discussion (which started in the 1950s) between "clinicians" (seeing each patient as a seperate case) and the "statisticians" (looking for averages and cut-off points).  A good reference book for that discussion is Paul Meehl's "Clinical VS statistical prediction" (1954). The intention of this mail is of course (a) to invite for an integration of several types of modeling that are practiced in the group of people calling themselves NLP practitioners / (b) to align that group behind an epistemology which is compatible with other branches of cognitive science. Patrick www.merlevede.biz