"any hypothesis that we develop is conditioned by the state of our knowledge at that time, a point which was easily illustrated by reference to swans. At one time any definition of a swan would have included the requirement for white plumage, but along with European settlement of Australia came the discovery of birds essentially identical to swans but with black feathers." Karl Popper (1902 - 1994)
So new data or information, or more close examination of existing data, can trigger the refinement - or even the abandonment - of a previous conclusion.
If something stares them in the face, they observe it. If they observe that it only appears in some circumstances, they note that. If it always appears under those circumstances, then it is likely to be a fact connected to those circumstances. If it doesn't, then it is more likely that it is depends on other circumstances.
If green soiders crawling up the bedroom wall stare them in the face whenever they overdo the booze, then their existence has some connection with the booze. If others (teetotallers) also observe them then they are independent of the booze and the data on them has doubled.
If their appearance is recorded repeatedly, then a prediction can be made of the circumstances when they will next appear. If that prediction proves correct, the observers have the beginning of a statement that green spiders do climb walls, in the circumstances predicted.
Other researchers may subsequently repeat the experiment. If the spiders fail to appear on cue, then the assumption is that some factor has not been taken into account by the previous observers and the connection of the spiders and the circumstances of their apperarance is not a fact.