My understanding is that a hypothesis is not scientific unless there is a way to sufficiently test it.

Is it ever acceptable to present a hypothesis that cannot be sufficiently tested?

I can imagine a situation where one has a hypothesis (A) that can be sufficiently tested via available research methods (e.g. a questionnaire) and another (B) that can only be sufficiently tested via other research methods that you do not have the resources to conduct (e.g. face-to-face interviews).

Would it be acceptable to partially test for hypothesis B and collect some preliminary data for it in this fictitious questionnaire, despite knowing that this would be an imperfect test? But then, if the testing method is imperfect, would this truly be considered a hypothesis or something else?

  • 1
    You mean like the Riemann hypothesis? Sep 11, 2016 at 15:03
  • @NicoleHamilton No, I'm talking about far more down-to-earth hypotheses that one cannot test sufficiently given the resources (time, money etc.) that are available but which one can nevertheless collect_ some_ data on. Sep 11, 2016 at 15:08
  • Please don't forget to change your title from "fully tested" to "sufficiently tested".
    – Nobody
    Sep 11, 2016 at 15:13
  • 1
    It sounds less like you're offering an hypothesis and more like you're posing a possibly interesting question and an inexpensive way to decide whether it'd be worthwhile to investigate further, which sounds pretty reasonable to me, depending on your audience. Apologies for poking fun. Also, you might revise "this fictitious questionnaire" to be simply "a questionnaire". If you do it, there's nothing fictitious about it. Sep 11, 2016 at 15:40
  • 1
    As scientists, we seek the truth, not simply to confirm our own biases. If you honestly can't know if something is true without extensive testing, your position as a scientist should be that you don't know, not that you think you could prove your opinion given enough money. To be a hypothesis, there has to be already be some evidence to suggest there might be a relationship, exactly the kind of evidence you might collect with your questionnaire. Right now, it sounds like you're in the "I wonder if there's a relationship" stage, not "here's what I think the relationship is". Sep 11, 2016 at 16:03

4 Answers 4


Bose-Einstein condensation was predicted in 1924 and only actually seen in 1995.

Gravitational waves were predicted by Einstein in 1916. It was proved that they followed from general relativity in 1958. They were observed indirectly in 1974, and observed directly in 2015.

If "science" had been restricted to hypotheses which could be tested in the near future with available methods, then these predictions would not have been considered science, and it would have set physics back by decades.

I would agree that hypotheses which cannot be tested in principle are unscientific.


This question seems to be related to Popper's falsifiability criterion. Note, however, that the question there is (certainly for Popper, anyway) whether a hypothesis can be tested in principle. Whether or not you have in fact the means to perform such a test is immaterial.


In statistical terms, you cannot partially test a hypothesis, you either test it or not. Also, there is no such thing as a "perfect" test.

  • You are right. I should have been clearer. In my example, I'm still testing for hypothesis B, it's just that the data collected is not adequate due to resource limitations. Sep 11, 2016 at 15:31
  • Inadequate in what respect? If the data collection method is inadequate, then your statistical analysis will be of little or no use.
    – Samuel
    Sep 11, 2016 at 15:34
  • 2
    Given that the OP explicitly mentions qualitative research methods (interviews etc.), I am pretty sure that he does not interpret "hypothesis" in a purely statistical sense. I assume this answer is technically correct and totally useless.
    – xLeitix
    Sep 11, 2016 at 15:36
  • Inadequate in the sense that one is unable to use the most appropriate research method to test it due to resource limitations. You are right to say that this means that statistical analysis will be of little or no use but could one claim that this still has qualitative value insofar as you're getting some data, despite the using a research method that is not ideal? Moreover, would this be considered a 'hypothesis'? Sep 11, 2016 at 15:37
  • @xLeitix understands my point. Sep 11, 2016 at 15:38

I believe the standard term for what you are talking about would be "preliminary results," and there is indeed a good place for work of this sort in the scientific literature.

The test for whether something is publishable is not whether it conforms to the idealized scientific method. The test is instead whether it makes a meaningful contribution to the pursuit of knowledge. For work that can be (usually retrospectively) fit onto the idealized scientific method, this test is fairly easy to pass.

There is also, however, a place for many various types of preliminary work, such as:

  • clearly formulating a problem and explaining why it is interesting or important to solve
  • developing a method for attacking an unsolved problem of interest
  • reporting a study that shows suggestive results, even though these are not definitive, this helping to motivate a more conclusive study.

All of these can be valuable in a number of ways, and all of them demonstrate incremental progress towards one's goal. The question of whether a particular such piece of work is publishable then boils down to whether the partial result is significant enough to attract interest in the venue in which one proposes to present it.

For this reason, "preliminary" work is less likely to get published in a field's top venues (though it most certainly can be, if enough people care about the problem)---the bigger the problem, the more that people will likely be interested in chipping away it little fragments of it (for example, think about all the preliminary work published on hunting the Higgs boson or chasing P=NP). There are generally also more informal venus like workshops, symposia, etc., which are quite well suited for discussion of preliminary work.

In short: if you think the partial results are interesting, then it's entirely reasonable to try to present them in the appropriate context.

You must log in to answer this question.

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