TITLE: When should a result be made into a paper?
QUESTION [95 upvotes]: I recently posted a short (6 page) note on arXiv, and have more or less decided that I should not submit it to a journal.  I could have tacked it onto the end of a previous paper, but I thought it would be somewhat incongruous -- it is an interesting consequence of the key lemma unrelated to the main result.  I really liked the concision of the paper and didn't want to spoil it.
This brings to my mind several philosophical and/or ethical questions about the culture of publishing that I find interesting, particularly at this point in my life since I am nearing the point of seeking a permanent position.

Two things are clear: It is to one's advantage, especially in the early career stage, to have many publications.  It is also to one's advantage to have a strong publications.  So take a result that is not very difficult to prove, but is interesting mostly because I think it may be useful as a stepping stone to another as-yet-unknown result.  Given that I have already put it on arXiv, is "because I think it could be published" a good enough reason to publish it?
One good reason to submit a paper to a journal is to have it in the refereed scientific record.  I was tempted to claim the result at the end of the previous paper without proof as a remark, but in the end thought better of making such a claim without giving a proof.  How bad is it to make such a claim if the proof (or even the truth) is not obvious, but can be proven as a reasonably straightforward generalization of someone else's proof of a different result?
This is a more pragmatic and frank question.  Should young researchers be careful to avoid the impression of splitting their work into MPUs (minimum publishable units)?  In this instance my reasons for not including the result in another paper are somewhat complex and non-obvious.
As a counterpoint to 3, how should one balance the purity/linearity of the ideas/results in a paper, versus including as many related results as possible?

I realize that these are very soft and subjective questions, but I am interested to hear opinions on the matter, even if there is no universally true answer.

REPLY [8 votes]: I would say that concerning the overwhelmingly large number of mathematical papers that appears, the main question (for the mathematical community, your career is something else) is: will the publishing of your paper add more signal (being MathReviewed, if someone needs the result, she or he is likely to find it) or more noise? For example, concerning a slightly different question (when should one split a paper in two?), I tend to think that if splitting does improves the probability of finding any of the results by a MR search, then splitting is actually good.