14

In our lab, we hypothesized that a technique T1 should be able to solve some problem with high performance. As per our hypothesis, we got an excellent result. We started writing a short paper to be submitted to a conference for the last date is due in four days. The paper is complete except for some proof-reading. We have not yet submitted the results.

Yesterday, just for fun, I was applying a different technique T2 to the same problem. Surprisingly it achieved an even better performance than T1.

We were wondering, is it okay to write a “failure” paper stating the hypothesis failed because of so and so [which is tough to analyze given the time constraint on deadline.]? Some of my colleague suggested to not disclose the performance of T2 until T1 is published, so that later I could do a comparative study between T1 and T2. Will it be okay?

Note: T1 and T2 are very different and it does not make any sense to write on both techniques in the same paper. Plus, rewriting the paper now is also difficult.

Update After going through answers, comments and suggestions, we are submitting T1 paper. Thank you very much all the learned academicians here on academia.SE.

  • 10
    I am not sure why you say that T1 'failed'. You got excellent results, you just got even better results using T2. – xLeitix Apr 15 '17 at 6:23
  • 2
    T1 happened to be on a path of work which lead you to discover T2. Sometimes the path or method of work is more interesting than the results. Many types of work in this world are incremental and research is too. Publish the first one first, start working on the other one as fast as first is done. Or at least write down the ideas clearly enough so you can easily pick it up later. – mathreadler Apr 15 '17 at 11:26
  • 4
    You can get at least three papers out of this: (1) T1, (2) T2, (3) a comparative study. Most people would count that as success, not failure! Of course, your comparative study might show that T1 and T2 are "best" for different sub-types of problems - but you haven't done that research yet! – alephzero Apr 15 '17 at 11:48
  • 1
    In addition to the arguments made in the two good answers here, I'll add that the good results you have so far with T2 probably still need to be corroborated with more thorough testing. – aparente001 Apr 16 '17 at 17:32
  • 1
    Although "sir" might seem like a natural way to refer to someone respectfully, remember that not everyone on the internet is a man. – Stella Biderman Apr 16 '17 at 23:27
22

Publish T1. Later, when you have time to write a whole new paper, publish T2. Doing a comparative study between the two sounds like an excellent follow-up decision as well. I'm not sure why you classify T1 as a failure... it sounds like both work and one works better.

That said, yes, it is okay to publish "failure papers." Depending on the field it might be harder to get it approved by a journal because it's less sexy to find a negative result than to find a result. This causes a measurable bias in some fields and depending on the field can be a major problem, because failed research is just as important as successful research epistemically speaking.

  • The paper is in Computer Science. Thank you for your answer. But, somehow, we are reluctant to go with just T1. We feel like we are hiding something to get published. Is it a natural thought? – Coder Apr 15 '17 at 9:09
  • 6
    @Coder You are not hiding anything; you found a technique that works and obviously felt was worth publishing. Preparing a second paper with the new technique, demonstrating its superiority and making comparisons would then be a separate contribution to your field. I think there would be more to learn from that, than simply publishing the second technique. – JNS Apr 15 '17 at 10:19
  • 2
    @Coder First of all, I don't think it's misleading at all to publish the result even if you know another one is coming down the line. You found that result first. Maybe if you found T2 first, then tried to make it worse so you could write two papers that would be sketchy, but what you're doing is totally above board. Second of all, if you don't publish T1, if I come across the same problem in two years I might try T1. By putting that method out there, I would not only know that someone else has tried it and what happened, but I'm also more likely to find T2. – Stella Biderman Apr 15 '17 at 22:33
6

I am assuming you frame the second paper as describing a rejected (failed) hypothesis, because your theoretically informed expectation (=hypothesis) was that T2 shouldn't perform as well as it did.

Negative findings (rejected hypothesis) can be as interesting and publication-worthy as positive findings (confirmed hypothesis). What matters is how strong the underlying theory/model is and whether your research is designed in a way that a negative finding is a good test for a theory/model and not just straw moving in the wind. If the rejected hypothesis is strong and the research casts doubt on a model/theory that is generally believed to be true, than this negative finding is much more important than the umpteenth corroboration.

Even better if your research design aids in explaining where exactly the model/theory fails. And yet more publication-worthy if T2 not only challenges a mainstream theory/model but also performs better than a more standard technique.

Pragmatically I agree with the strategy to publish T1 first, and then T2 after spending some time exploring why T2 performs as it does. To increase chances of acceptance, stress how the 'negative' finding challenges conventional wisdom and emphasize the unexpected and strong performance of T2.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.