I have been thinking about the meta-model in NLP, for which I have a rather natural distaste (given the presuppositions above). The meta-model is full of grammatical terms and made-up jargon. It jumps and jags uneasily between straight-forward sensory experience and hugely complicated "chunking" of different layers of "abstraction." Like an onion, the meta-model has layers that cause tears.
Tears is not a meta-model violation, although there should certainly be a separate category for ambiguity. Do I mean tears that accompanying weeping or tears like those in a piece of paper or cloth?
I will throw no stones here (are all metaphors complex equivalencies?) but I will say this: there are other ways of training the instinct to recognize and respond to language with considerable acuity. They involve units of thought and patterns that do not exist in the meta-model - and yet which are demonstrably doing the same things that the meta-model wants to do.
I could give you a model with complicated terminology. But I'd rather engage you in a story. . .