If you do not count defining them, and giving examples of them, as "speaking of them directly", I think you must have a nonstandard understanding of that phrase.
Good, so far we are in agreement.
Ah, there is your problem, right there. You still don't understand what a model is! Nowhere in my definition, or examples, of "models" did I ever say, or even intimate, that a model has anything to do with imagining "wires and pulleys", or any equivalent. This is all in your head. Of course you would not understand the usefulness of models if that's what you think they are!

Physics never includes "wires and pulleys" in its models, for if it does, then the wires and the pulleys themselves become the model, not the things they attach to and pull on. If wires and pulleys were not models too, then how would you know what to count as a wire or a pulley? You are modeling the actions of those things, and using that model to say what a wire is and what a pulley is. So if your understanding of a model is correct, wires and pulleys would need wires and pulleys too, to make them "work".

A theory cannot be anything else.
I do not posit a "mental version" of a theory; theories are mental descriptions, there isn't any other kind of theory that I could be talking about a "version" of.
I need to show no such thing, the theories obviously do that. It sounds to me like you are saying I need to show that a car can drive before I'm allowed to get in and drive it. I'll just get where I'm going and say "I did not need to show anything, I just did it."
Are you ever going to do anything other than purport a particular way of looking at connections between scientific theories and human practices? Because if that's all you plan to do, I can say right now with no hesitation that scientific theories have a connection with human practices. That's obvious. The point I have argued against is your claim that there is no point in thinking about scienitific theories (or models) in any way other than as a type of human practice. I have said, and will say again, that this is a particularly useless and ill-advised limitation to place on one's own view of what scientific theories are, because it fails to maintain a suitably tight focus on what one is thinking about. In the process, it resolves no paradoxes without introducing worse ones, and I generally so no particular value in it at all. That doesn't make it a wrong thing to do-- what is wrong is to claim that it is a better approach.

I'm certain that what Einstein had for breakfast on the day he had his most important breakthrough in general relativity in some small way influenced that breakthrough. Maybe he wouldn't have even had that breakthrough that day if he'd been sleepy from too many carbs, or unsettled from too much bacon. I'm equally certain that had he ate something else that day, the theory of general relativity that we use today would not be different by one iota. So this is what I mean when I say that although it is obvious that "human practices" interface with scientific theories, to require that we see them in that light is to miss a great deal of what is important about scientific theories, and science itself. Science is about what you get to throw out, it is not about why you need to include everything that interfaces with it, and your understanding of science needs to embrace that scientific principle. That is what models do for us.

Your position here is just patently false, there's not much more I can say. You have a completely wrong idea what models are, and you are demonstrably incorrect that they "don't accomplish anything". It is perfectly obvious that models are widely used and are vastly useful, and that you think otherwise can only be attributed to the clear fact that you don't use the correct definition of what a model is, despite quoting that definition in your last post.
I'm sorry, but I find that quote singularly unenlightening in regard to the practice of science and the use of models. It is making up its own problems and saying, hey, this isn't a problem. No kidding, I never thought that was a problem! I see no need to claim that man is a "ghost in the machine", nor to deny it, I see no relevance whatever to the entire concept. It certainly has nothing to do with making models, unless one takes that action way too literally (as is precisely the problem with the incorrect wires and pulleys approach).
It has no such implication, that is simply not correct logic. I have never claimed we can "fully understand" a model rather than the real thing, I have simply said that models are the only things we can understand because understanding is a mental relationship and a model is a mental construct. There is nothing we can understand but concepts that we construct in our minds, obviously, but that in no way claims that we understand these concepts "fully". I'm sure every human has many times experienced what it is like to understand vaguely or fleetingly some concept formed by our own mind.
You have simply chosen to define "understand" in terms of behavior, but it begs the question of what a toy is. If you define a toy by how someone behaves when they have one, you clearly cannot also say that we understand toys by that same behavior, or the word "understand" is synonymous with "define". In short, it loses its value as a concept separate from a definition. That is not a useful way to use "understand" in the context of scientific models, as it really misses the boat badly.

Sure you can define words any way you like, including behaviorally. You can define pain by the grimace on a face, or love by a pucker to kiss. But these are frankly quite silly ways to define concepts that we all know quite well simply from experiencing them, or if we have not, no amount of behavioral observations will remotely convey the same meaning as experiencing them. The same may be said for "understanding". I just see a really empty and useless landscape for application of a purely behavioralist approach to everything. More often than not, one is forced into a clunky and awkward way of talking about concepts with meanings that are much simpler, more elegant, are more useful, when thought about in a more direct manner. And no clearer example of that is our whole discussion about "models" and their absolutely central role in doing science. Is it not obvious that the definition I offered is something that is commonly used, indeed inseparable, from the science you have seen? What does that tell you?
Says you. I see that as quite a clunky and awkward way to think about how understanding is "constituted". For example, generally when I understand something, I can tell right away that I understand it, without waiting to see if I am "proficient" in practice. Other times I am surprised to discover I don't understand as well as I thought when the practice of something extends additional challenges, but that's just because experience is a great teacher. I still see no need to use behavioral proficiency as a way to define understanding, I still see understanding as something that is happening in my mind and is fully testable entirely within that mind.

Indeed, you cannot test it anywhere else. If you think that your behavior tests your understanding, you are mistaken, because it is always your mind that must pass the value judgement on the matter, regardless of the behavior. This discussion, for example-- is it a behavior that you are hitting certain keys and imagining that it all means something, or does it actually mean something that is coming from your mind? Behavioralism is pretty barren most of the time, presenting its own paradoxes even as it claims to resolve others.
And why do you think you understand the point you are making to me here? Is there some behavior you have engaged in that I'm not aware of that gives you that sense? Did you just create these thoughts here, or have you "vetted" this precisely identical argument elsewhere and noted the "behavior" of those who heard it, judging that as somehow encouraging to your behavior of making the argument? Or if you did that with a similar, but not exactly the same, argument, what makes you think that there is indeed a similarity, if not by "pure introspection"? Don't you see the paradoxes you are introducing simply by claiming that "pure introspection" is paradoxical? Thinking is different from behavior, get over it!

And that logically requires that words not be replacements or pictures for something, because there are many forms of these pictures and replacements? I don't follow that logic.
Obviously we learn language from the context of use. That's always the way we form our models of everything, from testing against experience, that's "context of use".
Of course a hammer is not a replacement for a nail, but the word "hammer" is certainly a replacement for the thing you might use to pound in a nail. You can only pound in nails with words in certain cartoons.
We can say that, if we want to replace a wonderfully concise and elegant concept of a model for an incredibly clunky and awkward list of behaviors surrounding it. But it would be a bad idea. That just sums up everything that you are saying about models, and that is precisely why no one who works with scientific models uses that awkward and clunky approach to them. It simply misses what is most important about them, I can't say it any more clearly. Instead of thinking that you have an insight that will help everyone who uses models, don't you think it's intellectually stubborn to fail to recognize that no one does it that way for a reason? Or are we back to the trained pigeons again?
Again, you feel you are solving paradoxes, but you are actually introducing them. How will you define "simple" in your behavioral approach? Will you only use it whenever more people are able to use something successfully? That's not a definition of simple, simplicity is an inherently mental construct, and is not synonymous with "easy". By replacing it with something more akin to "easy", you have simply lost its meaning in science.

__________________
Logic is the grammar of truth.

Meaning and absolute certainty are incompatible, and profound meaning and absolute certainty are profoundly incompatible.

The only thing intelligence is capable of is recognizing itself.