From a human/usability standpoint, most icons used in user interfaces are failures.

Not sure you believe me? Quick! Name five icons you often use in a piece of software you use regularly: your browser, IDE, spreadsheet, graphics package, whatever.

Now name five more.

If you’re like most people, you probably came up with the first five easily enough. Back, refresh, stop, home, forward. Save, print, bold, cut, paste. But by the time you were getting toward ten, were you starting to struggle?

Now consider that many of these applications line up rows of 10, 20, or even more icons, which all look rather non-descript and blurry at typical sizes anyway. Some applications have several such rows, some at different sizes, some horizontal, some vertical. That’s an awful lot of screen space for something that most users won’t use, at least not without doing the mouse-hover-tooltip-no-it’s-not-that-one-maybe-it’s-the-one-over-there thing.

Oh, and here’s the kicker: of the icons that do get widely used, only a tiny number are truly iconic pictures that are recognisable! A lot are just an arbitrary marker in a predictable position, and it’s that context that the user recognises.