53

I'm an undergrad and kind of new to this whole research thing. I've been doing research for the past ~9 months as a requirement to graduate with my bachelor's (there's the research track and software development track, I chose research).

I was told to try and submit my paper to conferences/journals (depends on their deadlines) to see if it gets accepted. But, one thing my professor said really stressed me out. If my paper gets accepted, and people find a serious mistake in it (ones that could cause your conclusion to be wrong, etc), it would destroy my whole career before it even began.

Can anyone with more experience go into detail about what could really happen? Assume, that the paper really gets accepted. On one hand, I'm not entirely confident of myself, and on the other, I have found a passion in research and would love to continue on for a PhD in future and this might help boost my resume a little given it is my only research experience.

  • 55
    A professor of mine once told me the story of a man (not himself) who did his doctoral work in differential geometry. It was an impressive and very important piece of work, and he rapidly progressed at high level universities. He also rapidly switched fields, becoming an algebraist. Within a few years it was realized that his results in differential geometry were wrong. Very wrong. "Cannot be saved in the slightest" wrong. But his algebraic stuff was golden, so he was fine. I believe it was characterized as "an honest mistake by someone who just wasn't good at geometry". – zibadawa timmy Feb 13 '17 at 5:20
  • 48
    Calm down. Papers are not to be read, they are to be counted. Your paper will not be read by a lot of people. When someone reads it, they will not pay enough attention to notice any mistake. – user12956 Feb 13 '17 at 13:50
  • 21
    The cynic in me would rather say that it is the opposite, while being a bit embarrassing, it's still a benefit for you. Nobody really checks what people reference in your articles, just that they reference you - anyone referencing your articles in order to say that you are wrong is just boosting your reference count. In addition, you can write a correction paper where you explain what the result really should be, giving you a second article almost for free. In the end, if you get accepted with a severe error, then it's a bigger shame on the peer reviewers and the journal. – Mrkvička Feb 13 '17 at 13:52
  • 21
    It won't pass peer review if there is a serious mistake (or it shouldn't at least), and thus, only you and a few reviewers will know about it anyway. I had papers with flaws submitted - flaws were discovered, paper was rewritten, and then resubmitted with no errors. – Per Alexandersson Feb 13 '17 at 14:31
  • 22
    I discovered a major bug in my second journal paper two years after the paper had been published in a journal, and three years after it had appeared at a peer-reviewed conference, and one year after a followup paper with the same error appeared at a peer-reviewed conference, and a month before I deposited my PhD thesis. The only serious consequence was that my thesis suddenly got 15 pages shorter. – JeffE Feb 13 '17 at 15:49
103

Congrats on your paper. No, it wouldn't destroy your career but it would be awkward and embarrassing. It could potentially hurt your career if the mistake was the result of obvious sloppiness, gross incompetence, and worst of all, outright dishonesty. But for the first two of those, the damage would very likely be containable and if you keep doing research, after publishing another paper or two that had no mistakes, no one would remember this minor incident.

With that said, your advisor is correct that it's best to avoid publishing papers with mistakes in them if at all possible, so do make a sincere effort to check everything to the best of your abilities before submitting the paper.

  • 20
    I am willing to bet that almost everyone has published something with some sort of error in it. Don't even get me started on basic spelling and grammar. No one even understands the spellcheck/grammar check button any more. Or basic knowledge of formats such as: APA or MLA. – NZKshatriya Feb 13 '17 at 5:18
  • 2
    Thanks for your insights! What kinds of mistakes would you call "obvious sloppiness" or "gross incompetence"? What I'm worried about is making a mistake in my coding because of the lack of information available online and in part, due to my not-so-pro coding skills. This could then lead to "wrong" or inaccurate results, and therefore possibly invalid conclusions since I'll be analysing those results. – User1915 Feb 13 '17 at 6:09
  • 34
    @User1915 but somehow implicate my advisor since I'll have to put his name in the paper too. Well well, you've just opened a whole new can of worms here. Setting aside the question of whether his name should be on the paper or not (which is important, but has been discussed ad nauseam here on academia.se in many questions, search for it), you should know that you cannot "implicate your advisor" in anything - if he chooses to accept coauthorship in a paper he takes full responsibility for its content, so it's up to him to make sure there are no errors if he cares about his own reputation. – Dan Romik Feb 13 '17 at 8:33
  • 4
    @GiorgiMoniava No. The reviewers have a much broader range of responsibilities. Checking the correctness of the results is only one of these. Others are to assess the relevance of the results in the context of the current state of research. However, if authors write a relevant paper containing new results, one cannot expect that reviewers are always able to spot errors that even the authors did not spot. I would guess that when the author is a respected researcher, a reviewer not understanding a particular technicality may even give the author the benefit of the doubt and recommend publication – Earthliŋ Feb 13 '17 at 12:14
  • 3
    @GiorgiMoniava Reviewers are human. – JeffE Feb 13 '17 at 15:50
50

I have a feeling that your advisor is using fear to make you work with extra care on the paper. While his intention is good, I don't like this method at all. He is definitely exaggerating.

Unless the mistake is a scientific misconduct, statistically speaking only one paper can't have much impact in your career, either in a negative or positive way.

Many published papers contain a lot of mistakes, even the most important ones. You should always write the best papers you can. But if errors happen, then just move on.

  • isn't the whole point of reviewers to check that submitted papers don't have errors? – user69307 Feb 13 '17 at 10:16
  • 4
    No, it isn't. The main point is to determine the significance of the results. Finding errors (or lack of them) is just a part of the task. – Martin Argerami Feb 13 '17 at 11:41
  • 2
    @GiorgiMoniava certifying correctness is some, but certainly not the whole, point of reviewers, and that certification is well known to have very limited reliability. Think of it in terms of levels of accountability: if it turns out that a paper had a major error that escaped the reviewer's attention, this is a small embarrassment to the reviewer whose credibility has been hurt with the editor, who knows his identity, but that's the only person who would know. The author on the other hand has been exposed in a major error in front of the whole world. That's a much bigger risk. – Dan Romik Feb 13 '17 at 12:21
  • 5
    ... It stands to reason therefore that the author, who is putting their reputation on the line in a much bigger way, would be proportionately more careful and diligent in checking the correctness of their own paper than the reviewer. – Dan Romik Feb 13 '17 at 12:22
  • 1
    @Giorgi and others: For what it's worth, in many branches of particle physics (which I worked in), reviewers don't check correctness of results. Period. Sometimes they will notice an obvious error like a factor of 2pi in an equation and point that out, but in general, it's just not practical for a reviewer to check the correctness of a paper in what we consider a reasonable time, if at all. – David Z Feb 15 '17 at 15:26
9

Probably not.

The "mortal sins" you definitely want to avoid are plagiarism and fabrication. Hopefully your supervisor has explained to you the ethics of scientific research, so that there is no chance you will commit these.

Everything else is forgiveable. Your most likely mistakes when starting out with research will be leaving out citations to some important previous work, or accidentally introducing bugs in your code or math. There are three lines of defense against such errors:

  1. You yourself should do a thorough literature research, and carefully proofread your paper and test your algorithms;

  2. Your supervisor presumably will check over your work before approving it for publication;

  3. The reviewers of your paper will spot some (but probably not all) potential errors that slip through.

Once the paper is accepted, any remaining errors will be visible for the world to see. But note that you can publish errata or revisions to your paper later on on your own personal web site, which is where most readers will download and read your paper (in the field of computer science, at least). So if you spot a minor error down the line, you still have a chance to announce and fix it.

What if there is a huge, gaping, embarrassing problem with your paper, and the reviewers miss it? That's still not the end of the world, provided that your good work outweighs your bad. Many years ago in computer graphics, a researcher published a fundamentally flawed paper at a top-tier conference (it wasn't obvious at the time, but it relied essentially on the false premise that rotations commute.) This spurred unrelated researchers to publish the report Errors and Omissions in Marc Alexa’s "Linear Combination of Transformations".

You know you're in trouble when people start pointing out your "errors and omissions"! But Marc Alexa went on to become a very well-respected professor in computer graphics.

  • 1
    one of the reasons I am not confident as mentioned in my question was that there is a lack of resources for me to confirm that what I did was correct. It's related to machine learning, and I was instructed to use some models from previous papers by adjusting the code to my own dataset. So far, a 2-3 of those models have pretty clear instructions. But then there are those with little or no info, and I'd still have to use them. For such cases, I am completely on my own. – User1915 Feb 13 '17 at 12:43
  • 2
    @User1915 you don't "have to" use anything, and you don't "have to" publish a paper part of which you have doubts about. Just like my other comment about your advisor taking responsibility for a paper he accepts being a coauthor of, the same is true for you obviously. If your concerns are severe enough about correctness of the code/results and the lack of resources, you need to have a frank discussion about this with your advisor and figure out how to proceed. Do not let yourself be pressured into publishing a paper you're not 100% sure you can stand behind. – Dan Romik Feb 13 '17 at 15:24
3

What is research? Finding new stuff that no one knew about or did before! If you get all too worried about not doing mistakes, you will eventually do "safe" stuff and chances are that this will not be the new stuff you were looking for. It's probably best to stay away from people like your professor that take mistakes too serious. Research thrives in relaxed environments but is ultimately doomed to fail if it gets competitive.

  • 3
    This strikes me as an unhealthily lackadaisical attitude. The place for mistakes is in the privacy of your own office, laboratory or department. Everybody makes mistakes there, and lots of them. But, by the time you're writing up work for publication, it needs to be, to the best of your ability, mistake-free. And mistakes will inevitably still get published: it's not a disaster as long as they were honest mistakes, but it's still something one should seek to avoid. If you want people to build on your work, they need to be confident that it's correct. – David Richerby Feb 13 '17 at 18:21
1

Perhaps, but not correcting it will cause a lot more damage to your credibility. Science makes mistakes. But it corrects the record and move on. You know what you need to do, you just don't want to face the consequences. You don't get a pass as an adult.

0

Nothing will happen if your conclusions are "wrong", but were appropriately drawn.

Researchers make wrong conclusions all the time. Conclusions are based on the information available at the time. In a research paper you describe what you did, how you did it, and the conclusions that you drew from what you did. One task of peer review is to decide if your conclusions can be reasonably drawn from what you say you did and the results you got.

In your paper you should describe what you did and how you did it. I can't comment on your specific field. But you should describe what inputs you used and the settings and methods you used. In my field it is fair to say I used a value of "x" for "alpha" based on the approach in "Smith and Jones, 2008" and a clarification like "we believe this value to be applicable to our situation because..." or "this is the only published value available". And in the discussion you can talk about how different values of alpha or different approaches may affect the result and conclusions (providing an analysis of how a different approach affects your results is a bonus, and may be requested by peer review.) You need to frame what you did in such a way so that if the conclusions aren't "true", what you did wasn't a "mistake" but was the logical conclusion drawn from a logical approach. You will have to decide what is the equivalent in your field.

As an undergrad doing research, you should be practicing making logical conclusions from logical choices in method. Research papers don't often have complete answers (the are often too narrow), they take a step in the direction to finding answers.

That being said, conclusions that are "right" (i.e. that are reproducible) are always better.

  • 3
    Sorry, I know this answer is well intentioned but it simply isn't applicable to mathematics - most of the key words and phrases you use like "settings and methods", "appropriately drawn", "inputs you used", "believe this value to be applicable", "reproducible", and even "conclusions" don't make any sense in the context of a math paper. A mathematical proof can be correct or incorrect, and that's it; there is no way for a result in math to be wrong but "appropriately drawn" (in fact the word "result" means something quite different to a mathematician than to other scientists). – Dan Romik Feb 13 '17 at 19:10

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.