Simply put, if user is about to click/tap/press a button on a screen or keyboard, it is going to take him less time to do so if button is bigger and it is physically closer to his mouse pointer/his fingers.
The referenced Wikipedia article does not however mention whether there have been any tests conducted with expert users - i.e. people familiar with the system, who tend to carry out actions faster and "chain" those automatically, without too much thinking.
Let's take, for instance, a common action, requiring different keystrokes on two different platforms: opening a file. On Mac OS X, in Finder, user is needs to press ⌘-O, on Windows, in Explorer, Enter.
While there probably is a difference in how quick this operation can be carried out by an untrained user, with "Enter" key being the fastest scenario due to the obvious reasons (only one key to press, key itself is generally bigger than "O" etc), I wonder if this difference becomes insignificantly small when tested on expert user base.
Are there any studies exploring the area of Fitts law applicability to expert user base and what are the results?
Also, as a side question: why on OS X ⌘-O has been chosen for such a common action as opening a file, and Enter for rename operation, which user needs to carry out way less frequently?