The imagination has made more discoveries than the eye. (Joseph Joubert)
We make a conjecture, a guess. We make tentative proposals, and then see how they work. If they don’t withstand criticism, we try again, either making an iteration on the initial theory, or making a drastic revision that explains why the initial theory was mistaken. If they withstand criticism, this says nothing about their truth or falsity. What happens, though, before the criticism?
Anaximander, the first scientist, said the Earth was unsupported in space. From the modern point of view, we do not understand that at the time this was a highly controversial supposition, one that went against the firm foundation of the senses. It is obvious (and I rarely use the word, for very few things are truly obvious) that this was something Anaximander could not possibly have observed. Here we have one of the most revolutionary conjectures, for while it happened to be true (or at least approximately true), it is the first known case of an individual going against the assumption that the senses were the final arbiters of the truth.
Anaximander deduced that if the Earth was supported on pillars, or the back of a turtle, or four elephants, or water, that the pillars, turtle, elephants, or water would themselves require some support. And this new support would require a support as well, and so on, in an infinite regress of supports. So Anaximander did away with the whole notion of support entirely and said that the Earth did not rest on anything.
The eye and the ear are notoriously unreliable, but more importantly, they do not give us any interesting theories. Cosmology does not come from the senses; cosmology is created from the mind. Here we have the first scientist acting — unwittingly, and centuries before the 18th century — as a Kantian, not an empiricist. Science began without empiricism, for empiricism in science is unnecessary and possibly unproductive.