Greetings RobMan Wow, where to begin! 1. You wrote, "Even though the NLP based models are binary (either present or not) that is still technically a statistical analysis 8)" Please name me one of the "NLP based models that is binary" - remember, if they are not present, they do not exist. Try arguing the anchoring (a NLP based model) is binary. 2. You wrote, "I also think that statistics could be usefully applied to the analysis of what makes a model elegant and sufficient.  e.g. what level of simplification is utilised across the collected models and what impact if any this has - a statistical model of elegance." Help me out here, RobMan - could we get an simple example of how "statistics could be usefully applied to the analysis of what makes a model elegant and sufficient"? I am having a complete failure of the imagination here trying to appreciate how statistics could be usefully applied to the analysis of what makes a model elegant and sufficient. I am particularly exercised by your comment,  3. You wrote "...what level of simplification is utilised across the collected models and what impact if any this has - a statistical model of elegance." It is difficult enough to imagine how the NLP based models I am familiar with have non-overlapping coding vocabularies, with their application to distinct area of expertise, that represent models of distinct logical types of transactions but when I think about the issue of simplification - that is, the mapping from the set of exemplars of a pattern to some explicit model - I completely boggle. It seems to imply that there is some measure of complexity/simplification between the process experiences being modeled and the explicit product that represents it and further that this measure is independent of the processes modeled by the various models involved. In attempting to consider what such a measure might be, I fall speechless. 4. You wrote, "While the variables in an algebraic equation are simplifications that may represent some number of real world complex objects the simplification is well defined...whereas the simplifications involved NLP modelling are still largely an expression of the skill of the modeller." You are certainly correct that the variables in an algebraic equation (you called them simplifications) may represent some number of real world objects but to claim that these simplifications are well defined is altogether another matter. Clearly, the manipulation of the algebraic equations is driven by a set of processes (commutative...) that are well defined but the entire artistry of scientific research depends precisely on the skill (and luck) of the person making the initial mapping between the real (sic)world objects and the variables expressed in the equation precisely as the effectiveness (learnability and efficiency) depends on the skill of the modeler in an NLP based piece of modeling in assigning explicit vocabulary to the elements in those sets of interactions they purport to model. All the best, John