What can I do to get the maximum out of the academic papers I read?

I have had tremendous luck trying to track down specific answers for stuff that everyone else takes for granted and stumbling into things I didn't know I needed to know. For example - in my assay I use a reagent called TMB. What this stuff does is turn blue in the presence of peroxidase and you get your numbers from how blue it's turned. So one day I was looking for the name of the product of the TMB reaction for a report. I went diving in the literature and found a paper written some time in the early 70's that answered my question. But it also hinted at explanations for some other observations, so I start poking around some more. And of course that raises new questions in my mind and pretty soon I tripped over a big pile of serendipity. An example of this is when I went to check the spelling of serendipity a few seconds ago (Just like it sounds? No way!) and found this: One aspect of Walpole's original definition of serendipity that is often missed in modern discussions of the word is the "sagacity" of being able to link together apparently innocuous facts to come to a valuable conclusion. So I believe that http://xkcd.com/214/ (Of course I'm also the guy who posted http://ask.metafilter.com/156656/Or-maybe-Im-just-seeing-something-that-isnt-there so your mileage may vary.) Don't read a paper just because trees died for it. Read it because it seems to answer a question you have. It may or may not, but while you're in the neighborhood, it might answer questions you didn't know you had.

In addition to the summarizing/synthesizing that you do, I also try to really pick apart the methods section on the first read, making an effort to predict the limitations and try to consider why there are any discrepancies between my list and the one presented by the authors. I read mostly medicine/public health, so perhaps this is not an issue with compsci. Something a little more applicable to your field that might also address the discipline/focus issue you brought up: skim the citations at the end of the paper and look up anything that strikes your fancy. This helps you learn about related contemporary research and can also direct you to older, influential papers. But yeah, most of the time I don't get too much further beyond the abstracts.

Read the abstracts of the papers cited. Rinse, repeat.

I'm curious where you get papers. Do you have access to electronic journal databases through a university or a library? If so, check out the Science Citation Index. It shows how many times each paper was cited and where, so you can kind of chart the evolution of ideas from one author to the next. The citation index is part of the ISI Web of Knowledge database, which has been my go-to resource in my research. It makes it easy to find series of related papers while weeding out "insignificant" work. It's also plain old brain fun, the academic equivalent of flitting around wiki pages or tvtropes.com.

I can't address the "how to choose what to read" aspect, but the "how to use this information later" question basically is about making a personal notes-library, and I can tell you how I've handled that, you can see if any of that sounds useful. What do you do afterwards now? Is this project-related, or just for general interest? Are you keeping/archiving the papers (either in a file cabinet or as pdfs)? If so, how do you organize them, and how are you planning to refer back to them? As you're writing comments, are they in a notebook, in a txt file attached to the pdf, in the (digital or physical) margins, on post-it notes, etc? What was really useful for me in project-related readings has been keeping a digital bibliography, renaming all the files with a fixed format (example - "Smith2008JCP_spectra", referring to Joe Smith's 2008 article on the spectrum of the molecule I'm looking at, in the Journal of Chemical Physics) Then in the comments section of the properties box, a few-phrase summary of why that article was relevant. "rotational spectrum of Molecule at temps from _ to _K". Then I'd set up the folder properties to display that column, and it's not too hard to skim down. I once tried folders for subcategories but ran into indexing trouble with stuff that had multiple possible locations. Whatever works for you, i'm sure there are tools out there. I also had a physical notebook or sometimes a word file with my notes in it, because I didn't reliably have the ability to make margin-comments on the pdfs, though that's the preferable route for me now.

I've found that, in my research area at least, the keywords authors choose aren't very consistent. So, whenever I get a new article I replace the keywords the author chose with ones from my list of keywords (I have about 30 keywords that I use). If you're reading in several areas, that might be more difficult. Once I read the paper I also generate a short summary (about 10 lines) of the highlights. That way, when I'm looking for something I can narrow the search down with the keywords I've added, then skim some summaries to find the few papers that I'm looking for. I also keep a list of the keywords that authors are using, so that I can effectively search for new articles based on whatever terminology is in vogue.

Use a program like http://www.metafilter.com/92758/For-the-Academic-Theorist-Hulk-in-All-of-Us to keep an archive of papers tagged with notes. Mendeley also has some social functions that should recommend papers based on what you've already read.

Also, most journals offer RSS feeds. Subscribe to them, and skim them using the above method. Since you'll never read past the title 90% of the time, this is managable. Even better is using a service like http://www.citeulike.org/ to store your library. It'll find users who have bookmarked similar papers and start making recommendations to you. These are often very useful.

As Sydney Brenner tells it, during the glory days of the Cavendish lab at Cambridge Francis Crick kept a placard above his desk that said READING ROTS THE MIND. Now, I would obviously take this advice with a large grain of salt, but I think the sentiment is essentially correct in that there are often much better ways to spend your time than mucking about with an ever-growing pile of papers. For instance, talking to other researchers helps sort the interesting questions in your field from the boring ones - which is often very hard to do when reading isolated papers that all treat their objects of study as if they were God's gift to humanity. Even more important is rolling up your sleeves and hacking through some problems/experiments of your own. That said, reading the literature is important, but it should be done efficiently. My own approach varies with my reason for reading the paper. The easiest case is when I'm looking for a specific fact or methodological detail - I find it an move on. For recent papers that are directly related to my work, I read the abstract and use it to predict the experiments and figures that are in the paper. I then check myself, and try to figure out why the authors did things differently than I would have. For keeping up with work outside my field, I do essentially what chrisamiller describes above - triage ruthlessly, stop reading as soon as boredom sets in, and don't get bogged down in the details.