73

I (an undergrad) and 2 other profs from my department have been working on something and we were thinking of submitting our work to a conference (a pretty well known one) this week. The fact is there is no related work that has been done on the problem we are solving and therefore there is no paper which we can cite as a starting point for our work. The only existing things that we make use of are trivial definitions from mathematics (like Riemann Integral) and also definitions that are pretty well known in my area (CS). So again, we do not think there is a need to cite any particular papers for that.

Now my question as stated was - is it acceptable to have a paper with no references? Or do you think we should just mention some papers (we never had to use them) where those definitions were actually proposed?

PS: I'm sure some of you might suggest me to ask those profs. But quite frankly none of them have much experience as far as publishing papers is concerned. So it would be great if some one could give me a good advice and help me out.

Edit: Thanks everyone for the answers. Of course somewhere a little sarcastic and ridiculing, but I can understand since this is a little weird. Actually yes, I had the notion that a paper wasn't worth citing unless I use some results from it (say an algo or some theorem). But I guess that isn't the only reason why I should be referencing as many seem to suggest here. The fact is we had seen quite a few survey papers before we started this out and I was under the impression that there wasn't a need to cite them since anyone could find them. But now, things are bit more clear and I guess I should cite them and I will :)

  • 55
    You're trying to say that you are working with professors that "do not have much experience in publishing papers"?? How... did they become professors then? – penelope Nov 14 '13 at 15:23
  • 21
    Your department has professors who “[don't] have much experience as far as publishing papers is concerned”? Run! – F'x Nov 14 '13 at 15:24
  • 8
    Come on, at least a couple of references to justify your work :-) – dgraziotin Nov 14 '13 at 15:40
  • 17
    I don't know what specific topic you're working on, but "1 or 2 papers a year in somewhat well-known [venues]" is pretty darn prolific for someone in my subfield! – zwol Nov 14 '13 at 18:07
  • 4
    @mchid That's not true.[1] – Roland Nov 15 '13 at 9:53

20 Answers 20

118

Though I don't think there's any hard rule against having a paper with no reference, it seems pretty weird. Note that references are not only for citing other people's results (theorems, algorithms, etc.) which you have used, but more broadly to recognize other's scientific contribution. For example:

  • Has the problem never been discussed before? Who first realized it was a problem, stated it, formalized it?
  • You probably put the problem into the broader context of your field. And if you don't, you probably should. This sure requires citations on recent work on related problems, even if nothing was ever done on the one problem you're addressing.
  • For example, is your problem a specific case of another problem, or does it have generalization?
  • What are the consequences of your results/findings? They probably have some impact on other related problems, or practical consequences on real-life issues.
  • Didn't you or your co-authors ever do any prior work on this issue?

Finally, let's see it another way: you have solved a problem, that no one else has solved, worked on, or more generally discussed. And you did so using only elementary techniques, which have been known for so long that they do not require citation. Stated like that, it may sound like either you're a genius opening an entire field of mathematics, or you're working on a useless problem that nobody cares about. You probably don't want the reviewer to be thinking that way!

  • 41
    There is third possibility: the problem is solved, just under different name, on different notation, or is in fact a simple consequence of a more general thing. If you consulted it with no-who is proficient in your field, the third option may be very likely. – Piotr Migdal Nov 14 '13 at 20:03
  • 6
    @F'x this helped me. Actually when I said there was none worth mentioning I just meant that we hadn't used any previous results (algo, theorem etc.) But yes, there were quite a few survey papers we read before we started on this one and so I think I will mention those at least in the introduction. – anon. computer scientist Nov 15 '13 at 5:25
  • 12
    @anon.computerscientist Yes, if you read review articles on this, be sure to cite them! – F'x Nov 15 '13 at 8:19
  • 6
    ^ This. IMO, this the main use case for citations. Your job as author is to allow a reader to understand how your argument is constructed and to be able to conduct external validation of that argument based on the body of knowledge that informed it. – Matt Nov 15 '13 at 8:57
  • 2
    @anon.computerscientist In that way, it is not uncommon to have a paper that does not directly use any non-trivial non-textbook theorem; yet, the context matters a lot. – Piotr Migdal Nov 15 '13 at 15:27
68

I know of exactly one published paper in my field with no references whatsoever.

The "References" section reads, in its entirety:

No references on this topic seem to exist and no useful results could be found.

  • Presumably you have read the original print? If not it is entirely possible the references are inline and/or foot prints. Digitization of old papers is pretty bad in some cases. – Meirion Hughes Nov 15 '13 at 10:14
  • 2
    @PiotrMigdal not necessarily in CS, conference papers are a unit of publication there and for top conference the demands are as for good journals in other disciplines. – Artem Kaznatcheev Nov 15 '13 at 13:05
  • 2
    Unless someone publishes in a journal, pretty much. – Suresh Nov 15 '13 at 16:46
  • 16
    This is in fact a full paper; it was peer-reviewed; I have read the original print; there are no inline references; and I have confirmed personally with both authors that they were unable to find anything relevant to cite. – JeffE Nov 16 '13 at 5:09
  • 3
    I think that in 1985, CS was still a new enough field that you have solved a problem, that no one else has solved, worked on, or more generally discussed from @F'x's answer is an could have really been the case. – Bobson Nov 17 '13 at 0:03
35

No, it's not acceptable to have a paper with no references. One of the first things a referee looks for is whether the paper shows proper awareness of the current state of the art in the field and references previous, recent work properly.

What is actually somewhat more commonly encountered (although not in any decent journal) than a paper with no references is a paper in which none of the references are appropriate. For example, anti-relativity kooks will self-publish papers in which all the references are to papers from the 1920's, textbooks, their own work, and the work of fellow kooks.

In fact, one of the quickest ways of detecting that a paper is a kook paper and not worth spending much time reading is if it has these characteristics. The fact that the professors you're working with are unable to place your work in the context of current work in the field suggests very strongly that they are not competent, and this is reinforced by your description of them as being unfamiliar with publishing. You probably want to stop working with them in order to avoid embarrassment.

Becoming associated with someone who's a kook or publishes incompetent work could be the kiss of death to any future academic career you might have been hoping for.

  • 7
    This answer doesn't seem to address the question's premise that there is no related work and instead goes off on a tangent about kooks. Although it is important to be able to identify sloppy self-referencing work, that's not what the poster is asking. Instead of merely assuming incompetence, there should at least be some advice about how to search for relevant work or broaden what is considered relevant or something. – Rex Kerr Nov 15 '13 at 1:01
  • 1
    Rex, it seems that this answer assumes that there is related work and posits that the professors supporting the project have failed to understand or investigate the sub-field. The OP actually replied back to one of F'x 's comments above that there was material to reference. He/She had a misunderstanding in what should/should not be referenced. – Matt Nov 15 '13 at 9:02
  • @Matt - That the assumption was correct doesn't make it a good answer, though. "All legitimate areas of inquiry are already fields with published literature" is what it is implicitly stating. I doubt that is true or should be true, even if it is almost always true that work does fall within an established field. – Rex Kerr Nov 16 '13 at 1:02
  • @RexKerr - I agree that it is invalid to suggest all subjects that will ever see published research already have published material. I disagree that a yet-researched subject will come to be researched without any influence or inspiration from something that exists. – Matt Nov 19 '13 at 11:27
18

In principle there is nothing wrong with that. But, I have a hard time seeing that there is nothing published whatsoever on anything that forms the basis of what your research concerns. The problem must have some origin and there is probably literature that forms the basis for your work by for example indicating that your problem is a gap in knowledge. So even if it would theoretically be possible to publish a paper with no references, it seems so unlikely that the problem should be sought elsewhere, such as indicated by the comments.

The result of submitting such a paper would probably lead to rejection, even if your work in itself is sound.

  • Yes, this is one of those things that is not technically against any rules, but the circumstances giving rise to it are so vanishingly unlikely that one is unlikely to ever truly have literally nothing they can cite. – Columbia says Reinstate Monica Jan 21 at 3:08
16

I've actually been working on a project which is similar in that nobody seems to have done any work that is directly comparable in the area of application. The work uses standard Bayesian techniques (this is an applied statistics paper), but these techniques are not at all used in the area in question in the way I've done it.

However, I still had a bunch of references. If you are using people's data they like you to cite it. I had a couple of cites for the Bayesian techniques which I'm using - they aren't trivial. The broader area has been much worked on, but much more by biologists than statisticians, and their approaches are very different, though their aims are similar. So, I cited a couple of these papers, out of very very many possibilities. They weren't particularly relevant to the paper at hand, but I couldn't think of anything better. (I actually thought of posting here, but wasn't sure what specific suggestions a non-specialist could give, and I'd already tried asking specialists with no useful result.) Additionally, I used some quasi-mathematical derivations which I also had a cite for. So, generally these things do pile up.

So, I suggest backing up a step, considering the more general problem which you are trying to solve, and perhaps give some references to papers that have worked on that more general problem. Then you could say these papers are examples of the more general/related problem.

Other people have, I think, adequately addressed the issue that not being able to find relevant papers to cite means there might be a problem with the project. I.e. it is something nobody has worked on because nobody cares about the problem. This isn't necessarily the case, but I agree that you should look into that as a possible issue.

Doing novel work isn't necessarily a bad thing. I was offered a chapter in a book for a paper just published, not because the work was particularly good, but because, as the editor put it, "not much work has been done in that area". And apparently he thought the area was worth representing in the book.

13

No.

You say that your work is entirely new. How do you know this? Have you read any survey papers? If you have, cite them. If you haven't, go read some.

Has there been any other work on related problems? Cite it. Explain why their problem is different from yours.

Why should I care about your work? Has anyone expressed a need for someone to do what you're doing? Cite them.

  • 2
    Thanks for that. I was initially thinking there was no need to cite the survey papers I had read. But I guess you're right. I should and now I just did :D – anon. computer scientist Nov 15 '13 at 5:19
12

Your work needs to be self-contained. That means your idea needs to be built, first and foremost, on a solid foundation. You reference other people's work to establish that foundation. With one sentence (and reference) you can establish all the knowledge that your work AND paper is based on; including the language you use.

For example; pick any recent paper in your field. Now examine its references, and pick the oldest paper there; continue doing that and you'll eventually get to papers written in Latin that establish the very basic ideas of things you might take for granted as being fact.

When you reference a recent paper (which in turn references other papers), you are including that full body of knowledge of hundreds of years of work.

So I would say the answer is No, unless you're willing to dedicate thousands of pages to reinvent and reestablish everything your work is based upon.

Any notion that your work and ideas are unconnected or non-derived from any previous research endeavor is delusion.

  • 1
    Solid argument. Though, particularly for an undergrad Sophomore, rather than delusion, the notion is a lack of or mis- understanding of what references are actually used for. – Matt Nov 15 '13 at 9:07
  • 1
    Certainly, but if the author's supervisor (professor or assistant-professor) is saying this, there is no denying it is odd. – Meirion Hughes Nov 15 '13 at 9:45
  • Certainly. A charge of delusion on the part of the professors may certainly be warranted, I just meant specifically the OP. – Matt Nov 15 '13 at 12:17
12

Although my context is mathematics rather than computer science, this seems like an absolute no-brainer: "No". A paper without citations is a crank paper. Basically immediately and irretrievably rejected.

First, ultra-practically, referees will not easily believe either that your work occurs in a vacuum, or that there was no "prior art".

Second, if you think there was no prior art, I'd bet against long odds that you are mistaken. The referee may know the specifics ... and then you look awfully bad.

Third, unless you can explain why anyone should care about what you've done, that is, give references to give context, why should anyone care?

Citations are by far not just about what one thinks one's work "depends on", ... and even that should not be appraised naively.

I am shocked that "professors" would contemplate letting a paper out the door with no citations at all, ...

(And it would benefit you to refine your perceptions of their "publication rate" and/or "research activity". I realize it is hard for a beginner to appraise these things, but your description seemed very strange, as evidenced by some of the comments...)

12

Let me add my two cents in there: the main point of the references is to give context to your work (as suggested by multiple answers already). It's also about justifying the validity of techniques you used and justifying the evaluation process you used for your results.

There's several "different" references you usually want to include in your paper (and while I do admit that sometimes you don't have many in some of the "reference groups", I think it's impossible to have none at all):

  • general references about your (sub)field

    This would be used in the introduction, to set a broader context for the paper, especially if you're submitting to a general conference in your field instead of a small subfield-focused conference.

    As an example, I work in Mathematical Morphology, which is often used to perform image segmentation, object detection and image filtering, and belongs to the field of computer vision and/or image processing. So, when I submit to an CV or IP conference, I have to put it in the context because not all the referees/attendees will be experts in Mathematical Morphology and some general references will be useful here.

  • general references about your problem

    This is actually very similar to the previous, and would be put in the introduction as well. It stems from the same reasoning: not everybody will know that what you are attempting to solve is even a problem.

    Again a personal example: in my field of Mathematical Morphology, there are hierarchical structures used to hold information about an image in a tree structure. I recently published a paper about reducing the size of those trees. Why should anybody care? Unless I give good motivation, nobody will. I had to give references to various applications of the tree structures I was working on (including a tiny bit of motivation for using them at all in the first place), and then give references and elaborate why the size is an important factor and what applications will benefit from my proposed technique.

    Oh, and the "core" of the paper was using something I used in advanced high-school programming papers: DFS. I still had to give context as to why my specific application was interesting.

  • references about the structures/techniques you used (even if they're simple or not common for your field)

    I guess this would go somewhere between Introduction and Related work. But, it's important to justify the validity of what you're using, and you're not going to prove it from scratch in your paper. Maybe it's also interesting to _ elaborate on the connections and similarities_ between the standard usage of the technique and common usage.

    I'll give two examples here: There was a recent paper in my field presenting powerful but simple techniques based on Kruskal's minimum spanning tree algorithm. Of course Kruskal was cited (even if most computer scientists learn about that in their high-school mathematics / undergrad algorithms courses). Also, there was an older paper in Computer Vision about image-by-image search applying techniques from text processing. It was explained in quite some detail how parallels can be drawn while processing an image as a text document, e.g. what is the equivalent of "words" in an image.

  • specific references about your problem and possible previous solutions

    Who attempted to solve it before? Maybe somebody attempted to solve a partial problem? A similar problem? The same problem in a different context? Who was the first one to pose it as a problem? This would go to Related work.

    Even if your contribution is not directly based on any of those, you have to put your solution in to context. If there's a previous partial solution that's justification that your contribution is important. If there's a similar problem solved, comparing similarities and differences might help somebody to one day make a generalized solution. Maybe somebody got results that are as good (as fast, as complex, as precise...) as yours using different techniques? Good, compare pros and cons!

  • references about the testing framework/techniques you used

    How did you evaluate your approach? How will you justify your evaluation to the reviewers? Why are the statistics you presented actually relevant to evaluate the validity/impact of your contribution? Discussion/Results section.

    For a lot of things you can produce a lot of useless statistics. You have to justify your choice of how to present your results. Is the testing framework widely used? Cite it! Is it new or an upgrade on an existing framework? Cite pros and cons related to the "standard testing approach" and argue why your approach is better suited to represent the pros of your contribution.

  • references about the data you used

    Similarly for the testing framework. Why this particular dataset? When did it first appear? What makes it suited for your problem? Where was it used before, and how does that justify your choice?

  • references supporting your claims about the importance of your contribution

    In addition to giving motivation, you should once more emphasize the impact your paper will make, and the potential applications of your contribution.

    This is definitely different from "what you based your paper on". It's almost the opposite: you're listing papers/authors that might benefit from your work and base their future, extended work on your paper.

See, it's not just about citing the mathematical/algorithmic bases of your work. It's about proving that you did your proper background research, giving motivation for your problem, explaining the benefits, and convincing everybody that your work is unique, awesome and useful. And, sorry to say, if your supervisors couldn't tell you most of this, you should think about changing them (especially if you have an interest in academia).

11

No, it isn't acceptable, and is probably a sign that you should back up a bit and make sure your result really is completely unrelated to other results. If you think that you have something totally unrelated, which you obtained from elementary-ish mathematics, it probably:

  • Is considered trivial or irrelevant by others in the field. For example, I could very well write a paper on the derivation of the drag of a Batman Curve-shaped object. This uses basic principles and doesn't need to call upon advanced prior results. But it's not too useful, and is probably not considered a worthy topic for research. This, of course, was an exaggerated example, but you may be in a less pronounced situation. Make sure that your paper is something that constitutes publishable work.
  • Has already been published before, just with different terminology. Usually, the simpler things get snapped up first; if you feel this is the case then you should dig deeper; and possibly show your results to other researchers in the field asking if they know of similar results.
  • The third possibility is that you have truly made a breakthrough or opened up a new avenue of research or done something radically new. Such results are very rare, especially when they build from basic concepts only.

Either way, one way to beef up the references is via the introduction. I've noticed that a some physics papers have a lot of references in the introduction, and they skimp on references in the rest of the paper because the mathematics may be straightforward or otherwise not worth referencing.

If your paper is publishable research, there probably is some motivation behind it. And some ideas you may have built upon (If not, you can always find similar ideas to "pretend" to build upon). Write an introduction detailing the motivation, and cite papers there. Look at the introductions of other papers in your field to get a better idea of how to write one and what sort of references are at home there.

  • I guess that even for Batman Curve, in principle it could be made into a publication (given it is not just a single example, but something more general (=reusable)). But then, appropriate references are still needed - in this case not much pure math, but things related to vector graphics, image compression, graphical object rendering, etc. – Piotr Migdal Nov 15 '13 at 11:53
10

I have always imagined that the reason for citing references in a paper is not to give credit, but to keep all research grounded in other research. Something like, if someone cannot understand the result in your paper, or what motivated it, then either the result is wrong, or that person did not understand the result of one of the references, or what motivated it.

If you are working on a specific problem, for example, you should reference other papers which also work on the problem. If you are using a specific tool, you want to reference some other papers that have used that tool, and certainly the result for which the tool was developed.

It's about giving your research context, and by extension, giving other peoples research context, which is why results are published in the first place, instead of kept secret.

10

I published a paper regarding a new technique for printing iron microparticles. In my research I found a long list of printing methods, but ultimately the technique I invented was unique in the field (that's according to the reviewers). Yet the references I used, although they were not directly related to what I was exactly doing, helped me devise the final idea. I think that probably you can put some references, just to give some context on something that might be related somehow, rather than risk yourself to appear as an amateur. In today's world, where there are papers and inventions in the millions, there should be something related to your idea. Or maybe you are like Einstein who published with no references, but even he was exploring and ultimately refining Newtonian physics. So strictly speaking he should have cited Philosophiæ Naturalis Principia Mathematica.

  • 4
    Einstein's "ON THE ELECTRODYNAMICS OF MOVING BODIES" does not have formal references but it is more a question of style. He does refer to earlier work. He says "as shown by Mr. Planck" at least once! – Jim Nov 15 '13 at 16:48
6

Feynman's Forces in Molecules has no reference and was published on Phys. Rev. http://www.cce.ufes.br/jair/mq2grad/PhysRev.56.340_Feynman_Forces_in_Molecules.pdf

  • 5
    As for Einstein's work, Feynman is a historical example which is not entirely suitable for comparison with modern expectations in academic publications :) – F'x Nov 15 '13 at 20:14
5

I agree with the general consensus here: find relevant work and cite it. You can cite textbooks or reviews for well-known things. You can also cite vaguely related problems or just the most exciting problem in the field. (This takes some experience, which you lack, but try your best.)

However note, that 'conference paper' means different things to different people in different fields. A regular paper (article or letter in scientific journal) never goes without References, never. However, depending on the field, some conferences call for abstracts or summaries. Those are much shorter (between 1 paragraph and 2 pages) and can include some references, but not as many as in the full paper due to space constraints. You can probably submit a 1-page abstract without References if the topic is interesting and clear.

2

References help make your work discoverable

Apart from the many intrinsic reasons that have already been mentioned, there is one selfish reason why it's good to cite all the relevant previous works in the field, and it's that citing those works will make it easier for people who read them and want to find the latest research on the topic. In the same way that you should chase up all the papers that cite the reviews you mention, to make sure that the problem is in fact still open, citing those reviews ensures that your paper will appear in the Citing Articles section of those reviews.

Even better, in many fields it is common for researchers to have citation alerts on such papers, in which case they will receive a specific notification of your paper. Depending on the system, this will often include the authors of those works, so this becomes a non-intrusive way to 'push' your paper to people who might find it interesting.

  • This is an excellent point. – smci Apr 27 '15 at 5:54
0

I would say even if Newton was going to write a research article related to "law of universal gravitation" would have found related work. Related work gives a broader view of research topic e.g, if newton was going to write article related to "law of universal gravitation" might have given a reference to Aristotle who believed that there is no effect or motion without a cause. Now science and technology is so advanced you must find related work. Or might be you are genius invented new field in science and technology.

0

Here is a paper with no references:

Fisher, R. A. (1950). "The Significance of Deviations From Expectation in a Poisson Series". Biometrics. 6: 17–24. https://doi.org/10.2307/3001420

-1

I'm referee for various IEEE and other Journals, generally speaking citations give me the idea of the behavior of the article, it is really not acceptable for me few citations (7-10 at least for conference); despite, my view could be rought, other referees may finalize the same conclusion.

I suggest to add at least 7 citations (at least on the introduction and conclusion) in order to deceive superficial people like me.

You can refer to book, manual and technical worksheet.

  • 13
    This seems like terrible reasoning. "Add 7, just to make it look like you tried" sounds like a strategy a high school student might employ. – xdumaine Nov 14 '13 at 21:54
  • 2
    I would imagine most (like me) look for specific references to important papers, rather than the quantity. – Meirion Hughes Nov 14 '13 at 21:58
  • Given the large amount of publications it is very difficult to accept no references. Anyway, journals provide to referees and editor tool to check related paper using keyword and references, in order to find plagiarism or related works. When the editor cannot find related paper he cannot contact appropriate referee and may reject the paper as "not interesting". ... the system is not perfect but that is all. – venergiac Nov 14 '13 at 22:16
  • You can all criticize all you want, but this is how the publishing world works. Devoting time to reviewing literature is in part placing your work in context, in part showing familiarity with the field (proving you're not a crank and have done reading) and in part genuflecting and ego-boosting the incumbents in the field. Whether the world should be like this is entirely another matter. – smci Apr 27 '15 at 5:53
-1

There are no general formal rules for this but some conferences may have such constraints and most reviewers will probably sort out papers with few citations anyway. As it is regarded as good academic writing style to cite from primary sources, it might be a good idea to cite from the original papers of the definitions you use [i.e. (Riemann, 1868)].

  • If someone cites a 19th Century paper, unless for a very specific reason way beyond a definition, it tells me that the person is not aware of the state of the art. – Martin Argerami Nov 15 '13 at 11:45
  • It depends from the behavior: on medicine context is a common practice to cite old article (Yarbus 1969) as must – venergiac Nov 15 '13 at 18:58
  • 1
    Most of the papers in quantum cryptogoloy (or more precicely QKD) I read cite Einstein or Heisenberg for explaining the basic principles (and so do I). These papers are from the 20ies to 50ies of the last century. The principles simply do not really change, much like the Riemann integral, so these papers are still state of the art. For me, it seems suitable to cite from the original papers instead of derived works. Of course there should be newer references, except the work does not use any other sources than fundamental works. – Marste Dec 12 '13 at 9:14
-5

It's not a problem, you can submit your work with no references if nothing related to your work is done before. You and your team may get a Nobel Prize for the work you guys are trying to do, so keep it up and do great things.

References are used when we are going to extend the work which has already been done or we want to do something related to that. In your case if you have no references, then you can go for that Conference.

protected by Nate Eldredge Nov 16 '13 at 0:09

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

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