Sometimes a duplicate question gets an answer before it gets closed as a duplicate. In some cases this answer is either better than the answer(s) in the "main" question, or at least adds something valuable to them. Such a situation can be very annoying, because it is not clear what future duplicate questions should be made duplicate of, information gets scattered, etc.

Technically, such a duplicate question can be merged into the "main" one; this copies all the answers into the "main" one, as if they were there all along. Only moderators can perform such merging. It is not clear to me what guidelines moderators use to decide on merging, and I would be grateful for explanations.

Here is one example:

I flagged the recent duplicate for moderators attention suggesting a merge; my flag has been marked as "helpful", but nothing else happened. What does it mean?

I did not flag the old duplicate, as I am less sure that it should be merged. E.g. the answer there is accepted; I guess in case of a merge, the answerer would loose the reputation associated with having an answer accepted, which is a pity. Still, reputation issues aside, I think it would be more convenient if this all was merged.


1 Answer 1


Merging is difficult (impossible?) to reverse and so is used only when the questions are such close duplicates of each other that all the answers to one would read acceptably as valid answers to the other.

Perhaps more merging should be done, but since the decision in each case requires comprehensive review of both threads, careful thought, and great confidence in the correctness of the action, I usually am reluctant to perform this unilaterally.

Maybe merge proposals should be made here on meta as each case comes up, and debated and voted on by the community, just as tag synonym/merge proposals are.

  • $\begingroup$ Thanks! Frankly, I am not sure I understand the point of non-merged duplicates. The idea behind keeping (and even, as some say, cherishing) duplicates is that they can help leading future readers to the correct thread; in the perfect world, all duplicates would be closed immediately and would never have answers. In reality closing takes time and answers often appear, but I would argue that closing with merge should be the default option, as opposed to closing without merge... However, I see your point. I will maybe try to bring up some duplicate discussions on meta to see how it goes. $\endgroup$
    – amoeba
    Commented Dec 21, 2014 at 22:19
  • 3
    $\begingroup$ After having performed many searches on this site, I have grown to appreciate the value of duplicates: it often is difficult to know which keywords to use (even when searching for an answer of one's own!). Duplicate questions often state a problem in different ways and use different vocabularies. The merging process destroys the questions and keeps just the answers, which would lose that latent semantic information. $\endgroup$
    – whuber Mod
    Commented Dec 21, 2014 at 23:19
  • $\begingroup$ Wait, @whuber, in this case I must have misunderstood how the merging works! I was assuming it leaves the duplicate question intact, and its only effect is to move all the answers into the "main" thread. If what you are saying is true and the merging destroys the duplicate question, then can you explain e.g. what these questions are: stats.stackexchange.com/search?q=[pca]+locked%3Ayes ? I found three "locked" questions with [pca] tag; all of them say that they were "merged" into other questions (with link provided). Isn't it how merging always works? $\endgroup$
    – amoeba
    Commented Dec 21, 2014 at 23:28
  • 2
    $\begingroup$ @amoeba You're correct: I had forgotten that the duplicate question was left behind. (The feedback given to mods when going through the merge process does not include a visit to the affected thread; it's just a lot of text messages documenting all the changes that were made.) That fact must make me reconsider the merging process. It still remains the case that merging cannot be undone by anyone in the community--and might even be pretty difficult for SE programmers to do, too--so we would still want to proceed with caution. $\endgroup$
    – whuber Mod
    Commented Dec 21, 2014 at 23:41
  • 1
    $\begingroup$ I am glad that this misunderstanding was cleared! Still, maybe it is a good idea to bring up some specific merging suggestions here on meta. Actually, I listed two suggestions in the OP here, but have now reviewed them and see that the second one (about PCA imputation) was maybe not ideal: the answer to the duplicate is referring quite substantially to the specific details of the question. I edited my post to remove it. But my first suggestion (about PCA feature selection) still stands; it would be strange to re-post it as a new meta thread. Shall we maybe try to consider it here? $\endgroup$
    – amoeba
    Commented Dec 21, 2014 at 23:59

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .