What kind of research failures are publishable?

Question

Suppose that a physicist has discovered that the presence of X causes Y in a certain kind of systems. So they go a step ahead and try to work on a similar idea of X being the cause of Y in a little different kind of systems and it takes them months to arrive at the results.

But lo! That relation does not hold true here.

Is this result publishable? In general, what kind of failures in research are not failures from the point of view of writing papers about them? For example, if someone starts off hoping to prove the Riemann Hypothesis and discovers that it is not true, that is 100% publishable since a lot is riding on the RH.

Is a lot riding on it the only thing at work in such situations?

Answer 1

A lot depends on the message which is conveyed by the paper and how it is formulated. In your example, essentially, researchers take system A, use theory A' to explain the experimental results observed in it, publish 1st paper, then take system B, use the same theory and find that the theory does not work in the sense that it cannot describe the experiment there. So, a paper saying "system B shows interesting behaviour which is not captured by theory A', see how" has higher chances being published than a paper saying "we tried to apply theory A' to system B and it does not work".

Further, it is very common that if you achieved a "failure" (which is not the best word if applied to research), that is, a negative result, you can reformulate your message to make it sound "positive". In particular, you can analyse why exactly your attempts did not work, what was lacking in your theory (theories usually have assumptions and/or model simplifications) and how further theories can be improved to actually do the job. This is valuable information and often is perfectly publishable.

Answer 2

What is publishable is up to the publisher with advice from some reviewers who, ideally, know the field and what is important.

But, since research is supposed to lead to knowledge, the fact that something is not true can be as important as knowing what is. In particular, this situation that you describe gives the limits of what you call causality here, which could be important.

But the only way to know is to get feedback from others in the field and submitting for publication is a good way (normally) to get that independent feedback.

Not everything is as fundamental as the Riemann Hypothesis, of course, nor does it need to be. Lots of things with less "impact" are published every day. Several times a day, in fact.

---

And, don't interpret negative results as necessarily failures. A failure is when the methodology doesn't support the conclusions or errors are made along the way. A negative result is just a result.

Answer 3

As with all things in science, you must be explicit with what you mean by the word "failure". A null or negative result is not in general a failure, and are often just as valuable or even more so than positive results. For example, in the search for neutrinoless double-beta decay, only null results have ever been published, yet these results are important inputs to many theories of particle physics and universal evolution.

A "failed" experiment is not one that does not observe an expected effect, but one that cannot objectively answer whether the effect was observed or not. If some background completely obscures your signal so that you can't perform any analysis whatsoever, it may be a failed experiment. But even in this case, it is sometimes useful to publish, as a "cautionary tale" to help other researchers avoid that background in the future. In that way even the "failed" experiment is still contributing to the sum of human knowledge.

Pretty much the only completely unpublishable work is "we have no idea what we're doing".

Answer 4

Well there was that whole Michelson-Morley failure, a few interesting things followed it. Give some thought and look for the insight your results suggest.

Answer 5

In physics, lots and lots of papers are published discussing models which turn out (or are even known at the time of publishing) to be failures.

One possibility is that the model fails but is published on the chance that a similar approach taken in the paper might lead to a model which ultimately does work (see, for example).

Answer 6

In some cases where experimental tests are involved there's a high chance of results being published despite the predicted results were not shown up. it's because they mostly commit the results of experiments which may be expensive for others to do. By publishing them even with a failed hypothesis create a bases of truth(tests) for other researched to work (evaluate) their models. it's like registering test results.

As an example in some Engineering fields like Aerodynamics (or maybe chemistry) this is the case. Testing models in wind tunnels require high amount of energy consumed and wind tunnel itself is expensive, then it's common that you would see some papers from well established wind tunnel facilities in research institutes (which has trustable test results) that despite having trouble showing their theory, or even with clear failure will see their results being published.