is usually interpreted not as a statement about what John doesn't think but as a statement about what he does think. He does think this novel is not good. Similarly, the sentence
John doesn't think this novel is good.
normally conveys John's positive desire not to leave early. This phenomenon has been referred to as Neg-raising (??, 19??) and as negative absorption (Klima, 1964).
John doesn't want to leave early.
In the early days of generative grammar, this was as a syntactic phenomenon to be explained by a transformation that raised the negation from the complement clause to the main clause. In more recent years, the phenomenon has fallen below the linguistic horizon, as researchers have judged it to be purely pragmatic in nature. I concur with this judgement entirely, and in fact suggest exactly what kind of pragmatic process is involved. It is, once again, an instance of metonymy, where the coercion relation is provided by the explicit content of the sentence itself.
The logical form of the above sentence includes the following (where tense is ignored):
That is, there exists in the real world the condition e1 of a thinking event e2 by John j not occurring, where the content of the thinking is the eventuality e3 of the novel n being good.
This interpretation is then pragmatically strengthened via metonymy, using as the coercion relation think'(e2,j,e3), coercing the second argument of not' from e2 to e3. The resulting interpretation of the sentence is thus