[T]here is a methodological bias in favor of taking natural discourse literally, other things being equal. For example, unless there are clear reasons for construing discourse as ambiguous, elliptical, or involving special idioms, we should not so construe it.
