A hypothesis is a first try at explaining why you think something is the way it is. You might see flashing lights up ahead on the right on the highway and hypothesize (from your past experiences in similar situations, plus application of logic) that someone has been pulled over for speeding. As you get closer, you can perform observations to confirm or deny your hypothesis (it might be an accident instead, or even road workers doing clean up on the shoulder).
A hypothesis is made very early in the process, and its purpose is to phrase the question in a way that gives you something to prove or disprove -- a way to test whether you are right. In some experiments you do several cycles of improving your hypothesis before doing a final collection of data and analysis.
Scientifically, to be properly formed a hypothesis must be falsifiable -- there has to be some concrete result that, if you got it, would show you that your hypothesis is wrong. Some hypotheses cannot be falsified using labs, observations, and experiments ("There is a God" being a classic one), and therefore they are not useful when attempting to scientifically prove or disprove a phenomenon.
If a particular hypothesis has been rigorously constructed and thoroughly tested over many years by many experimenters, science begins to call it a "theory".
Colloquially, it's not a word very often used; people tend to use 'theory' in ordinary speech for both what formally should be the hypothesis OR for what is formally a theory. This drives scientists up a wall, because people use "theory" to mean "half-assed off the wall silly explanation" when it should mean something much more venerable and tested. This leads to things like anti-evolution activists saying, "It's only a theory! It could be wrong!" and intending their statement to come off as derogatory.