Good article, though he makes it sound as if there is an actual optimal amount of detail that you can put into an icon, while in fact it's very dependent on the context.
For example the right-most "house" icon looks very housey to me, because I could gather from the line of icons that it was supposed to depict a house.
When building interfaces it's funny (or not, depending on the pressure) to find that users interpret the icons in completely novel ways. For example: trash cans are interpreted as brushes because someone expects the paste button at that position.
When building interfaces it's funny (or not, depending on the pressure) to find that users interpret the icons in completely novel ways.
Which is why I always think - do I really need an icon here, or should I use text?
The worst example I have come across was a package we took over where there was an icon of a dog. Incredibly, this was supposed to be "Fetch from database" (dog, fetch, geddit?). Even after days of using it my brain still locked up looking at the damn dog - imagine my poor users who only had to interact with the software occasionally. Switching the icons to text cleared up any confusion.
Unless screen real estate is really at a premium (eg iPhone), I prefer to see text, and any application that gives me the choice (eg. Firefox) I immediately configure to do so.
For example the right-most "house" icon looks very housey to me, because I could gather from the line of icons that it was supposed to depict a house.
When building interfaces it's funny (or not, depending on the pressure) to find that users interpret the icons in completely novel ways. For example: trash cans are interpreted as brushes because someone expects the paste button at that position.