14

Before publishing, I found a minor mistake in a study that took a month to do. The mistake will not affect the main conclusion of the work, it just shifts a curve on one of the plots ever so slightly. My advisor says it is not worth the time to redo the work to correct the mistake, because it will not make a big difference in the results. (This is a month-long project). My field is engineering.

Is it unethical to publish the results with the known mistake?

10
  • 2
    Why would you need to "redo the work"? Isn't a quick re-write enough?
    – Buffy
    Mar 24 at 10:26
  • 11
    One thing to consider: if you end up wilfully publishing a mistake, you may end up in a situation where you have this niggling doubt at the back of your mind for the rest of your life whether you did something unethical and whether you should have done things differently. If you are seriously considering this, be sure that you have properly factored "your own peace of mind in the future" into your decision making. Mar 24 at 13:57
  • 7
    @MrAsker "Would it be OK if the effect of the error is shown, in that respect, but I don't say what caused it?"If you know what caused it but do not say it because it will make your results look less impressive, that is certainly unethical. Mar 24 at 15:52
  • 19
    If the mistake shifts things so slightly that it doesn't matter, why can't you present the results accurately as the experiment was performed? If you can tell it doesn't matter, won't everyone reading the paper be able to tell that, too?
    – Bryan Krause
    Mar 24 at 19:10
  • 12
    I wish my projects were only a month long! Mar 24 at 19:17

8 Answers 8

52

If you try to cover up the mistake by manipulating data, or omitting critical information that is needed for reproducibility, then yes, it is unethical.

If you include a mention of the issue, and describe any applied correction, then it is fine and it is up for the peer reviewers to (hopefully) pick up on and decide if they are concerned about the impact it may have on your results.

11
  • 5
    The first paragraph here s the right answer. Your advisor is wrong. I recommend the rewrite, though the second paragraph offers a possible way to avoid that. Mar 24 at 14:01
  • Thanks. With regard to omitting information: I've struggled to understand "reproducibility" within my field. Is the ideal to offer enough details so that someone can reproduce the data "well enough" to come to the same conclusion you did? Or would they need to be able to reproduce every plot exactly? If the latter, a single journal article would be a 300-page book in my work. Unless I give them all my source code (which I can't do). The mistake in question is regarding a detail that wouldn't normally be mentioned. My advisor usually favors less detail for readability's sake.
    – MrAsker
    Mar 24 at 15:24
  • 6
    @MrAsker one must distinguish between "the good of science" and "the norms of a field" and "issues of practicality". For the good of science, everyone should be able to perfectly reproduce everything in the paper just by following your methods. If you want to support the good of science, you should aim as close to that as you can, within the constraints you're subject to. This doesn't require redoing a month of work to fix something trivial, but if you can explain the discrepancy briefly in a footnote, future readers will thank you. Mar 24 at 16:14
  • 5
    Of course, for the good of science it should be mandatory that all the source code be published too, since reproducibility is usually kind of hopeless without it; but as my friends sometimes hyperbolically say, we live in a fallen world. Mar 24 at 16:15
  • 3
    if the problem of publishing the source code is the length and not any requirement to keep it a secret, there are archives such as at the university of Michigan that focus on such sharing for research --- icpsr.umich.edu/web/pages/deposit/index.html
    – Mike M
    Mar 25 at 0:08
10

When you publish a paper, you are asserting that everything in it is true. If you know before submitting for publication that something in the paper is false, then (excuse my bluntness) you are lying. Lying is unethical (at least almost always).

As a separate issue, if you don’t fix the problem before publication and a reader spots the error, they may wonder if there are other errors in your paper.

In conclusion, I recommend fixing the mistake before submitting for publication.

6

From the example you provide it sounds like the work that has to be re-done is remaking a figure or re-analyzing the data (since you mention that you already know what will happen: a curve will shift to the right).

This is very different from redoing actual experiments/measurements (even there, one could argue that a month is nothing, unless your contract or funds have run out to prevent you from redoing the work).

If it's the latter, you might want to leave the data as they are and discuss the suboptimal setup in the discussion. In that case I would also argue it is not so much a "mistake" as an ill designed experiment.

If it's the former and you catch a mistake in the writing stage, then by all means fix it. Transparency and your reputation are key.

2
  • Recommend removing the first paragraph or move it farther down in the answer for future askers only. I think OP was clear that the rework required is specifically to redo the experiment correctly. Just because OP knows how correcting the mistake will affect the analysis does not mean that the analysis is the only correction needed. They specifically said otherwise. Mar 24 at 13:57
  • Thanks. The issue is not a plotting issue. It is an error in the data that causes the plot to look slightly different.
    – MrAsker
    Mar 24 at 15:25
5

No, it is not unethical to publish results with a known mistake. As long as you state clearly what the mistake was, what the effects on the results were and give your reasoning why the main conclusion is not affected. Point out the shift in the curve and explain it. All of that is perfectly ethical.

What would be unethical would be shift the curve to match, change the data to match or in any other way prevent the results from being reproducible.

As long as you state something along the lines of

"The first three data sets were collected according to these standards and resulted in these curves; the final dataset was collected according to these slightly different standards and resulted in almost the same curve"

then there is nothing to say against publication.

1
  • If you downvote my post, please add a comment explaining why. Otherwise I cannot improve the post.
    – AlDante
    Mar 31 at 16:23
5

Is it unethical to publish the results with the known mistake?

Of course, it is, unless you disclose your mistake. If you disclose your mistake in a publication, the only way to publish your findings would naturally be a submission to a low-quality venue. Junk results belong to a junk venue. This might be fine with you.

As for the impact of a hidden mistake, in some fields, folks will read your publication and perhaps care or reuse the data, whereas in other fields, your paper would be write-only. It's hard to say what the impact of a hidden mistake would be.

However, ethics is not the main point here. Regardless of ethics, and even if the impact of your mistake is zero:

If you value your own self, in particular, what you say and write, then tell the truth, or at least don't knowingly lie. Lying is not healthy and doesn't do you any good in the long run. Build and maintain the habit of being truthful as much as you can. For your very own benefit.

2

It’s not completely clear what the nature of the mistake is, but from your description in the question and comments it sounds like, in broad terms, the data you have collected do not reflect the experiment you claim in the paper to have performed. That means that publishing the results without redoing the work would be knowingly misleading. This is academically unethical.

It might be worth considering the potential consequences for attempting to publish in this case:

  1. Reviewers may notice the error and require you redo the experiment anyway in order to publish.
  2. After publication, a reader may notice the error. You could be forced to retract the paper, which reflects much worse on you than not publishing at all. At the very least the paper would be badly received by the academic community.
  3. If evidence arises that you knew of the error and published anyway, you could be subject to disciplinary proceedings in accordance with your organization’s academic ethics codes of practice.

Compare these possibilities to a one-month delay before publication when making your decision.

0

Redo the experiments during review.

If you are certain, you could send the paper for review, with a cover letter explaining the problem and stating that results will not shift nowhere close to changing the conclusion of the paper. Then, during the review you could rerun the experiment.

This might result in one of few outcomes: your paper is reviewed and then you fix the problem before publication, editor might reject it before you fix the issue, or one or more reviewers might reject reviewing due to this issue. In any case, you will not be doing anything unethical and will have chance to reduce time to publication.

-1

If there is no time right now or the process is to far, publish an erratum if the error may hinder another team to reproduce your results.

4
  • OP is clear that they haven't published yet, so they definitely should not intentionally publish an error with a plan to issue an erratum later.
    – Bryan Krause
    Mar 26 at 0:03
  • @BryanKrause: theoretically yes, but better that the student corrects the error in judgement of the supervisor than not correcting it.
    – Sascha
    Mar 26 at 20:17
  • If they're not able to convince the supervisor to do differently now it's extremely unlikely they'll get them to agree to publish an erratum, which many people unfortunately view as a bad mark.
    – Bryan Krause
    Mar 26 at 20:22
  • @BryanKrause: People view erratum as bad marks? Can not remember that i ever colleagues speak badly or make fun about errata. I would suppose the supervisor wont care after the publication......
    – Sascha
    Mar 26 at 20:33

You must log in to answer this question.

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