The whole idea of "observation" in a field particle physics is quite different from the classical way we think of observation, e.g. using a microscope to look at a cell.
First of all, the idea of observing something visually, with light, has a fundamental problem. The main problem is that optical instruments (lenses, microscopes, etc), even if they were built to perfection, are "diffraction limited". Since light is a wave, it will interfere with other light waves. This interference makes it impossible to use light to accurately resolve things that are very small.
To get around this issue, experiments like the Large Hadron Collider do something completely different. They generally will have an area set up that is surrounded by detectors. They then attempt to create the particles they want to study in this area. In the early days people used cloud chambers and bubble chambers in which the paths of particles could be directly seen from the trails they left. The type of trail (curvature in magnetic and electric fields...) would tell a lot about the particles.
Now faster detectors able to keep track of far more particles quickly are used. When the particles hit a detector, they leave a kind of subatomic "fingerprint". All observations in particle physics are reliant on the interactions between the particle we want to observe, and some part of the experiment.
Physicists can analyze this fingerprint and find out all sorts of interesting things about the particle. This is how they "observe" the particle's properties. Since these detectors are incredibly sensitive, they tend to pick up a quite a lot of "noise": readings that aren't the particle's fingerprints. The best way to alleviate this problem is to measure billions of fingerprints. When you have more fingerprints, your data are more conclusive, and you can be more certain that your observations of the particle's properties are correct.
(published on 03/29/2012)