The whole idea of science is this. It's not enough to simply have a really neat idea about how something works. That idea has to undergo rigorous testing through good experiments. So, for example, the reason that most scientists accept Einstein's theory of relativity is not because it's a really neat idea, and not because it was put forth by EINSTEIN, but because his ideas have been confirmed by experiments. An example may help.
The "really good idea" in particle physics, or one of them anyway, is largely the work of three physicists named Weinberg, Salam, and Glashow. These guys came up with a theory describing the weak force, and its very intricate relation with electromagnetism. There are many consequences to this theory, many of which have been verified, and none of which have been contradicted so far. I'll focus on one.
One feature of the Weinberg-Salam-Glashow model (also known as the Glashow-Salam-Weinberg model, the Glashow-Weinberg-Salam model, etc.) is that there must be an even number of quarks. Why that is, is a long story. Basically, the math just doesn't work out with an odd number of quarks. (Well, I guess it wasn't so long.) On the other hand, for many years physicists only knew about five quarks: up, down, charm, strange, and bottom. So here lies a way to test the theory. The theory predicts that there must be a sixth quark. So if you find a sixth quark, that's an indication that the theory may be true. If you can't find a sixth quark and there are only five, then the theory is wrong. So you look for a sixth quark.
Particles that are too unstable to exist in ordinary matter can be created in collisions between stable particles in a particle accelerator. The heavier the particle you're trying to create, the faster you have to collide things. (E = mc2, you know. The more m you want, the more E you need.) The top quark was believed to be a very heavy particle, and so the particle accelerator for the job was the Tevatron at Fermilab. This is a machine that collides protons with anti-protons at speeds very very very very (keep going) close to the speed of light. The theory predicts that in very high energy collisions between protons and anti-protons, one will, from time to time, produce a top quark.
This picture, which I've shamelessly stolen from the Fermilab website illustrates one example of what a top quark might decay into.
Here's what's happening in this picture. The proton and anti-proton collide, annihilate, and produce a top quark (above) and an anti-top quark (below). The top quark decays into a b quark and a W+. The W+ decays into a u and d quark. Now remember, quarks never occurs on their own, but always as parts of other particles. So as the quarks pull away from each other, they either have to snap back on each other, or break apart, producing other quarks in the process. The quarks in this decay are moving so fast that the breakup produces a whole spray of particles (such as protons and pions and neutrons) called a "jet." That's what's denoted by the little spray of lines coming out of the quarks.
The top anti-quark in this process decays into a b anti-quark and a W-. The b produces another jet, while the W- decays into a muon and an anti-neutrino. So what you do is you see if your detector picks up any collisions that result in a muon and some jets. (The neutrino will escape undetected. As for the jets, some may overlap so you won't necessarily see four.)
Of course, it's not quite this easy. There are other ways to get a muon and jets in a proton/anti-proton collision. So what you really need to figure out is whether your experiment gives you more "events" with a muon and jets than what you would expect to get if there wasn't a t quark. (More, by the right amount, I might add.) This, incidentally, is what makes the decay shown in this picture better than most other decays. A muon won't turn up in most of the more ordinary processes. Anyway, if you do get more events, then you look at the energies of all the decay products in these events, and that will tell you the mass of the t quark. This all took place a few years ago at Fermilab, and so we have further experimental evidence supporting the Weinberg-Salam-Glashow model.
I've simplified things a bit, but I think this gives you an accurate flavor for how particle physics is done. You start with a theory. You examine the theory in detail, looking for a way to test it in a particle accelerator, and then you collide the particles, analyze what comes out of the collision, and see if the results match your theory. There are other kinds of experiments you can do in particle physics, but most of the big results of the past 30 or more years have come from particle accelerators.