I am beginning to really like Ricoeur:
Let us look once more at the functioning of ordered polysemy, which we considered earlier with field theory at the level of language. Then it was a question of limited polysemy; ordered polysemy is properly a meaning effect produced in discourse. When I speak, I realize only a part of the potential signified; the rest is erased by the total signification of the sentence, which operates as the unit of speaking. But the rest of the semantic possibilities are not canceled; they float around the words as possibilities not completely eliminated. The context thus plays the role of filter; when a single dimension of meaning passes through by means of the play of affinities and reinforcements of all analogous dimensions of other lexical terms, a meaning effect is created which can attain perfect univocity, as in technical languages. It is in this way that we make univocal statements with multivocal words by means of this sorting or screening action of the context. It happens, however, that a sentence is constructed so that it does not succeed in reducing the potential meaning to a monosemic usage but maintains or even creates a rivalry among several ranges of meaning. Discourse can, by various means, realize ambiguity, which thus appears as the combination of a lexical fact — polysemy — and a contextual fact — the possibility allowed to several distinct or even opposed values of a single name to be realized in the same sequence.
I’m picturing this thought as a venn diagram. The first word in a sentence is a vast circle of possible meanings, but as more words are spoken, more huge circles are added to the diagram, and the overlap shrinks. With each word, the overlap is eaten into until the overlap is no more than a point. Or… even better one finds no overlap at all, or multiple convergence points, and the listener/reader is forced to revisit each of the word-circles to see if one has neglected a dimension of meaning which is the key to understanding. This is why, when reading I always read with a dictionary at hand. I do not gloss over unfamiliar words or attempt to grasp the gist of their meaning contextually. It is precisely the unfamiliar words (and the familiar words used in an unfamiliar way) that challenge the assumed context by which one understands.
Coming to understanding something someone says means changing the very context one by which one understands. And, just to make it personal, understanding what someone says is the same as understanding that person. The ability to merely describe repeating behavioral patterns (even if those patterns make a person’s behavior predictable), or even to explain behavior in psychological terms (even if those explanations make people’s behavior in general explicable or even predictable) is not to understand them, but to bypass understanding them as people and instead to understand patterns or explanations. Most people find the latter much much more comfortable, because it fits neatly within an existing understanding, and therefore is a science and not a philosophy.