I've been reading a lot of research papers recently about imagery and how it affects computational speed - the ability to quickly perceive the idea as compared with sentences. This is a pretty interesting article: Why a Diagram is (Sometimes) Worth a Thousand Words.
While this is not new, research shows this to mostly be the case only when the image itself is representative of existing knowledge - i.e, when you see a picture with a bottle of wine, the bottle means "alcohol, party, flush feeling" rather than "storage container". when the image does not fit your mental model, it might actually slow you down, because you have to question the meaning of that image in the new context.
This brings up the topic of how much time us visual communicators should spend researching what are images that are commonly understood. I was trying to figure out if there is a central repository somewhere - couldn't find any, but did find an interesting logo repository.