For example, consider an auto mechanic trying to figure out why a car is making a funny noise. He may hypothesize that a part is worn out, or that a coupling is loose, that some parts have come out of alignment, etc. But he probably won't seriously consider supernatural causes such as ghosts, gremlins, or demons. This mechanic is using methodological naturalism. If this mechanic is also a theist, then he uses methodological naturalism in his job, but does not subscribe to philosophical naturalism.
Methodological naturalism and science
Some critics, notably Intelligent design proponents, charge that this is an arbitrary restriction that blocks exploration of potentially fruitful areas of research.
However, science is eminently practical. The reason scientists don't consider supernatural explanations is that there is no way to test them to see whether they're correct. In the example above, an auto mechanic may hypothesize that a problem was caused by evil spirits, but there is no reliable way to either confirm or disconfirm this hypothesis. To put it bluntly, the reason that scientists don't consider supernatural explanations is that they don't work.
Note that it is conceivable that someone might some day come up with a testable explanation for some phenomenon that involves a supernatural entity. If a genuine miracle ever occurred, this rule would have to be revisited. In the meantime, given supernaturalism's track record, it seems safe for scientists to avoid it.