It's not like you need an educated guess either, if you already have a question or theory to test, then would you need to have an answer before you can prove it?
If you already have a question or theory to test, then you have a hypothesis. I'm not sure what you're thinking of when you say "hypothesis," but I'm pretty sure what you mean by "question" is what xingoxa means by "hypothesis," and what xingoxa means by "question" is something else.
I suspect that what xingoxa and Xonen mean by a question (which is not a hypothesis) would be something like:
"What factors can affect people's pain tolerance?"
Okay, we have a question! Let's test that by...what, exactly? Going and looking at people with different pain tolerances? Trying to notate every single factor
that differs between them? How do we know which factors are related to pain tolerance and which are just incidental?
A better approach is to ask:
"Does swearing increase people's pain tolerance?"
I suspect that this is what you mean by "question," but it's really a hypothesis with a question mark at the end—you are making an educated guess that swearing might increase people's pain tolerance, and asking whether this educated guess is correct.
This hypothesis is quite testable. Get a bunch of people together and have them take some (harmless) pain tolerance test (like measuring how long they are able to immerse a hand in ice water) twice. On one trial, give them some collection of harmless words to shout when feeling pain. On the second trial, let them swear. (Be sure to vary the ordering of the trials.) Then calculate the average change in immersion time.
A hypothesis embedded in a question makes it testable. A question with no embedded hypothesis is not testable. (However, it can certainly be a launching point for collection of empirical data, like collecting all the information you could about people with different pain tolerances. Upon looking at this data you might notice that people with greater tolerances seemed to swear more than people with lesser tolerances. This would then become a hypothesis
—an educated guess based on available data—which you would have to test through a controlled experiment. As xingoxa said, theories don't just drop out of general empirical data. Testable hypotheses do, which you then experiment with to form theories.)