Pragmatism in philosophy is much like behaviorism in psychology. So I like to think of this in terms of a behaviorist bird box where a pigeon gets food for learning to peck a given button under given circumstances.
Do you honestly think the bird attaches meaning to the buttons, and thinks of itself as communicating the wish for food through the button? The odds are that the bird is doing something less conceptual. And in fact, in responding to language, so are we. But we have a built-in sense of meaning that reinterprets what we are doing in terms of representation because we as humans have learned a theory of mind.
We think of language in terms of representations because we experience life in stories, so that we can share it more readily. But we know from physiology experiments that the representation comes alongside rather than before the act. We start acting before we are done forming the representation. The two processes happen in parallel, rather than representation happening and then meaning being ascribed afterward. Therefore, our activity is not a result of representational manipulations. Unless it is really complicated, it is basic response, like the bird's. This means there is something below the level of representation that makes language work.
We do still form and store the representation, anyway, describing our own action to ourselves in case we need to share it or reprocess it later. But the representation is really just about storage and communication, and is not a basic aspect of meaning.
This means that our words 'attach to reality' not to interpret the world to ourselves, but to render it communicable. And again, we learn communication and interpretation by success and failure, and not by parsing. We impose that later, and it is a separate refinement process. Think about how many turns of phrase you use that make no literal sense. If you needed to have the language parse out right and find representations in your mind for the respective parts in order to use it to listen or speak, those would disappear from the lexicon.
And from an "ontogeny recapitulates phylogeny" point of view, look at how languages evolve from more complex grammar (like Sanskrit and Attic Greek) to more analytic grammar (like English and Chinese). This tells us that the process of rationalizing and representationalizing language progresses over time from a muddled parallel process (where the meaning is distributed over many words via endings or combinations) into a more linear one (where words more and more stand alone and take the same consistent, uninflected form). Languages get richer in vocabulary and connotation over time, but less complex and nuanced grammatically. If we really relied upon representationalism to base our use of language, we would need to start simple and build.