I print research papers and other interesting writing fairly regularly at work, and read when I'm in the john or when I'm compiling. An awesome side benefit is that I've started handing papers to the folks I work with who like to socialize more than actually work. When I want to get something done and they don't, I mention something like "Hey, didn't you once say you studied distributed systems in school? Here's a paper for a peer-to-peer name service - would you mind giving it a read and telling me if it's really as awesome as I think it is?"
It seems to work, because now when these same folks come by they are talking about papers I handed them instead of how their cat needs its claws filed or some other trivia.
I've (sort of) given up trying. I let all of my IEEE and ACM journal subscriptions lapse, and I've gone to a "pull model" where I just search for things I'm interested in, and hope I can find a preprint on Arxiv or find it in CiteseerX. There's just too much stuff coming down the pipe to digest all of it, especially while working on a startup.
That said, I still visit machinelearning.reddit.com and semanticweb.reddit.com, etc., and expect that anything terribly important will show up there (or even here at HN).
It seems to work, because now when these same folks come by they are talking about papers I handed them instead of how their cat needs its claws filed or some other trivia.