Thanks for your response. Page 55 is helpful, though it does not answer my question completely. The second class of models described on 55, those based on a common function, still appear to be the result of modeling one or more geniuses. In the first description, the example given is the hypnotic model, which is a combination of the Milton model and design based on the variables discovered in that model. The second example on 55 that might apply is the "hybrid model", which would seem to be a model based on multiple geniuses--like the Meta model. On 56 New Code is said to be based in design, though it is still unclear to me how the alphabet game was developed specifically from the parameters discovered in Classic NLP. (This and other passages are the source of my comment about the alphabet game being designed--so I'm afraid that any interesting distinction there is not really mine--just parroting what the book seems to be saying.) My questions about eye-accessing cues is actually sharpened in light of 55. All of the examples there involve modeling of geniuses, and the classification system seems to leave no room for another type of model. But eye-accessing seems to have come from simply intense observation of common behavior. So my question is whether this constitutes a third type of model. Your idea of the 6-step reframe as self-modeling is intriguing, but the pattern seems to have simply been handed to Grinder by his unconscious explicitly (on the blackboard) rather than as the result of some sort of modeling process. What began as a minor confusion for me is becoming a conviction that there is more to the source of NLP than modeling, even though modeling is extremely important and even central. I don't think I would be this clear on that point without your input. Again, thanks for your response.