Speaking of Language

If we read an old poem with insufficient regard for change in the overtones, and even the dictionary meanings, of words since its date--if, in fact, we are content with whatever effect the words accidentally produce in our modern minds--then of course we do not read the poem the old writer intended. What we get may still be, in our opinion, a poem; but it will be our poem, not his. [...]

Each new speaker learns his native language chiefly by imitation, partly by those hurried scraps of amateur lexicography which his elders produce in answer to the frequent question 'What does that mean?' He does not at first--how should he?--distinguish between different senses of one word and different words. They all have to be learned in the same way. Memory and the faculty of imitation, not semantic gymnastics, enable him to speak about sentences in a Latin exercise and sentences of imprisonment, about a cardboard box and a box at the theatre. He does not even ask which are different words and which merely different senses. Nor, for the most part, do we. How many adults know whether bows of ships and bows taught by the dancing master--or down (a hill) and down (deorsum)--or a boys' school and a school of porpoises--are accidental homophones (like neat and neat or arms and arms) or products of ramification?

A child may, of course be philologically minded. If so, it may construct imaginary semantic trees for itself. But it does so to explain the usages it has already learned; the usage is not a result of the theory. As a child I--probably like many others--evolved the theory that a candle-stick was so called 'because it makes the candle stick up'. But that wasn't why I called it a candlestick. I called it a candlestick because everyone else did.
~C.S. Lewis, Studies in Words, Introduction, (1960)