[gphilip]

I can speak of theoretical computer science: if you have work that is worth publishing (as evaluated by your peers), then yes, it is possible to publish without any affiliation. I would be very surprised if a good result is rejected just because the author is not affiliated.

(Indeed, it is possible to get published even if you have negative affiliation; see: http://insa.nic.in/writereaddata/UpLoadedFiles/IJPAM/20005a2... )

The operative word in the first sentence above is "if". In my opinion, unless you are in the Ramanujan or Erdős class, then it is highly unlikely that you come up with a publishable manuscript on your own. Briefly put: you need a lot of "context" to be able to do publishable work, and you are unlikely to get this context unless you interact with the research community.

[ewjordan]

The corollary to this: most correct work is not worth publishing. And most people bitter over their work not being accepted were not rejected because their work was incorrect, but because it was not notable. So at the end of the day, it's all a popularity contest, and that's what you're really engaging in.