The point of this exercise is to show us that the connection between words and things only makes sense within the wider context of a language. At the end of section 6, Wittgenstein makes an analogy with a brake lever. The lever connected to the rod only w orks as a brake if the rod and lever are attached to the other mechanisms that make up a car. A rod and lever on their own are not a brake, nor much of anything else. Similarly, to say that language connects words with things only makes sense given the co mplex mechanisms of language that are already in place. We cannot take the word-thing connection to be fundamental to, or prior to, the workings of language.
Augustine's description works fine as a description because, within a language-using framework, we can certainly claim that certain words name certain things, and that we can learn by pointing and speaking which words name which things. The trouble with c onstruing Augustine's description as a theoretical account of language is that this construal adopts a position outside of language; that is, it does not take the existence of language for granted, but rather tries to build a model of how language works f rom the ground up. Unacknowledged assumptions are bound to creep into this model because there is no position that is genuinely outside of language for us to adopt.