Finding the right graphic for an icon is hard — and even if you do find a decently descriptive graphic, it might not be descriptive for long.
For the majority of cases, trying to represent an abstract idea like “bibliography” in a 32-by-32 pixel array is futile, even if you do have millions of colors and an alpha channel. Sure, you might choose a book with a magnifying glass as your icon, but that graphic could mean many things: “library”, “help”, “research”, “index”, “vision impaired”, etc. Any interface that uses the icon would still have to add a tooltip to explain what it means. There is a reason why we have words — it’s so that we can specify one thing in particular no matter how complex or abstract the thought.
Why make the user go spelunking for the information they need? Just give it to them.
It came to my attention recently just how fragile the connections are between the iconal representation of a concept and the actual concept. Here is the Microsoft Word icon for “to save”.
It’s a floppy disk. There is only a tenuous connection between saving and a floppy disk even for those of us who know what a floppy is (and at the moment most of us remember them), but floppy disks are on their way to becoming as unknown as Charles Yerkes. Don’t know who I’m talking about? That’s my point.

Iterative design isn’t design by trial and error. Iterative design is a process of continually improving not just the design, but also the problem your design is trying to solve.