Analytic philosophy too often restricts itself to dealing with assertions, commands, and questions, as if these were the only three kinds of sentences. Grammatically speaking, this is not far from the truth, but grammar often masks the use of sentences. For instance, we can phrase assertions as questions, e.g. "Isn't the weather glorious today?" If we examine the ways we use sentences rather than their grammatical structure, we will find a rich variety. Among the various ways we use sentences are making jokes, telling stories, thanking, and praying.
To say that words are names for things presupposes a great deal about language. A name is simply a label, and a definition of words as names presupposes that we already know what to do with these labels. Pointing to two nuts and saying "this is called 'two'" is a perfectly adequate ostensive definition, but without an understanding of how "two" is subsequently to be used, we might mistake it to mean "nut" or "brown" or "round." And to say "this number is called 'two'" presupposes that we already know what a number is. Ostensive definition cannot be the foundation upon which language is built; it is only useful if we already have language. The Augustinian picture of language in section 1, then, does not describe how someone without language can learn language, but describes how someone who has language already can come to learn a new language.
In pointing to a blue circle and saying, "this is called 'blue'" or "this is called 'round,'" we may make the same outward gestures. This does not mean that there is a mental act that constitutes meaning one rather than the other. There is no single characteristic feature— inward or outward—that constitutes "pointing to the color."
Though talk of non-existent objects can sometimes be confusing, it is often a part of our language-game. We can say "Mr. N is dead," even though there is no longer a Mr. N to talk about. The names of nonexistent objects can play a role in our language-game provided we give them a use.
There are a number of problems with the idea that the world is filled with composite things that can be analyzed into indivisible, undefinable simple parts. First, it is far from clear what counts as simple and what counts as composite. White could be considered simple, or it could be a composite of all the colors of the rainbow. Second, to say something can be named but not defined or described tells us not about the nature of the object, but simply of the role of the word in our language. Third, the process of analysis whereby we break composites into their component parts does not always give us a clearer or simpler expression, but often just confuses the matter. Imagine referring never to "brooms" but only to "broomsticks attached in a certain way to brushes." Our attempt to analyze a composite world into simple parts is confused on many fronts.
Wittgenstein's remark in section 32 gets to the heart of his problem with the Augustinian picture of language, in which words are names of things. It is not so much that words are not names for things—in the majority of cases they are—but rather that this particular relationship between language and the world cannot be as fundamental as it is believed to be. The remark in section 32 asserts that the Augustinian picture can only be true for someone who has language already, but not for someone who is coming into language for the first time.