Philosophical Investigations, reading it together. In the Tractatus, he is under the mistaken assumption that every proposition must have a definite sense. That a statement must have a fixed sense is reflected in his analysis, that is, the one-to-one correspondence between a name and an object. He inherited this thinking from Frege, as per Frege's idea that a vague concept is not a concept at all, just as a vague boundary is not a boundary at all. However, in the PI Wittgenstein demonstrates that because a proposition is not clear, that does not mean that it has no function (PI, 71). Sometimes being unclear or inexact (in terms of a statement) is precisely what is needed. The method of analysis, as presented in the Tractatus, forces a view of language that is just mistaken, and Wittgenstein begins to realize this in the very early 1930s.
So, philosophers have a tendency, as did the philosopher of the Tractatus, to analyze language as if one is doing mathematics. This method of analysis rears its head all the time. In fact, when interpreting the PI, as is done in this thread, and in my thread on OC, we are making the same mistake. We are looking for that precise exegesis, which leads to a discovery of Wittgenstein's meaning. It does not mean that the work is all bad, it means that sometimes we are searching for the very thing Wittgenstein is criticizing. We think we eliminate misunderstandings by making our analysis more exact. When what we need is a general idea of his method, the PI method. Wittgenstein criticizes philosophers often for looking for the real artichoke beneath the leaves or layers (BB, p. 125). This criticism only goes so far though, because much of the time we are re-wording his writings to look at it from a different angle.