> Also, just because it didn't get read right away doesn't mean it's a failure. Ask some gray haired academics how often they've seen (first hand or otherwise) an obscure paper from decades ago become key in solving a modern problem.

True, but don't tell me that all of nearly a million unread papers published this year (accepting the Smithsonianmag's figures as true, for the moment) are going to solve a key problem of science in 50 years...

Unread and uncited are very different things. It appears that the 50% number is some sort of arbitrary (short?) time scale "has it been cited yet" metric. Citing something is much harder than reading; you need to do some research, write it up and go through the publishing process. If 200 people read a paper but not cite it, is it a failure?

I am pretty sure that the vast majority of cited papers have not been read by the people who cite them.

I've often tried to find the "canonical" paper to cite for a certain fact, and found that it barely even mentions the relevant fact at all. It just somehow became canonical because everyone needed a citation for the same thing later on.

Very likely not all :) How do you figure out which of the nearly a million are the useful ones?