Models versus experiments

In modern science it is usually said that the ultimate test of any model is its agreement with experiment. But this is often interpreted to mean that if an experiment ever disagrees with a model, then the model must be wrong. Particularly when the model is simple and the experiment is complex, however, my personal experience has been that it is quite common for it to be the experiment, rather than the model, that is wrong. When I started doing particle physics in the mid-1970s I assumed—like most theoretical scientists—that the results of experiments could somehow always be treated as rigid constraints on models. But in 1977 I worked on constructing the first model based on QCD for heavy particle production in high-energy proton-proton collisions. The model predicted a certain rate for the production of such particles. But an experiment which failed to see any of these particles implied that the rate must be much lower. And on the basis of this I spent great effort trying to see what might be wrong with the model—only to discover some time later that in fact the methodology of the experiment was flawed and its results were wrong. At first I thought that perhaps this was an isolated incident. But soon I had seen many examples where the stated results of physics experiments were incorrect, either through straightforward mistakes or through subtly prejudiced analysis. And outside of physics, I have tended to find still less reliability in the results of complex experiments.