Not all knowledge is discursive knowledge

"[We may have] irredeemably nonverbal knowledge. We recall the rowers in Hume's boat… If I am one of the rowers who going to certain rhythm bypassing contemporary convention, I have evidence that we have a convention to row in that rhythm. Our success enrolling in that rhythm for the last two strokes as evidence by which I arrived at my expectation that you will continue to row thus; And that you expect me to go on rowing thus. And it is evidence that you observe the same evidence. I can use such evidence, I can expect you to use it, and so on; but I cannot describe it. I cannot say how we are rowing — say, one stroke every 2.3 seconds — but I can keep on rowing that way; I can tell whether you keep on rowing that way; later, I could probably demonstrate somebody what rhythm it was; I would be surprised if you began to throw differently; and so on. Now there is a description that can identify the way we are rowing. We take 1.4+ or -.05 seconds for the stroke and .9 ± .1 for the return, exerting a peak force of 70+ or -10 pounds near the beginning of each stroke, moving the oars from 32° plus or -6° forward  to 29° plus or -4° back, and so on, in as much detail as you please. But, as we row, we have no use for this sort of description. We can either give it nor tell whether it is true if somehow it is given. We would need instruments, and even if we had them we could not go on rowing as we were while we took the measurements." -- David Lewis, Conventions, pp. 63-64