Question: What to write in a hypothesis that can't claim statistical results of any kind -- only those of engineering nature.

Explanation: In thesis writing, hypothesis is claimed to be one of the most important aspects. I'm not arguing with that but in some cases, it seems a little over-priced. I am currently writing a thesis that deals with enhancing an existing security model. I can't give out the exact idea obviously but please let me explain with the help of an example.

Let's say, I have a security problem that the existing "role-based access control" of MS-Windows does not solve some problems -- so, I come up with time-based constraints for the model used and incorporate them in a new "enhanced" model. In the thesis explaining this contribution, what do I write in the thesis? I can't claim an "N%" increase in security or any other statistic that I can think of.

So, what's the hypothesis in this case?

  • +1 for a bit over-priced. I sometimes see people dealing with the cases of this kind by saying "de facto" Commented Aug 25, 2013 at 10:42

2 Answers 2


The hypothesis-model is good for research where you want to understand how something works, but I think it is ill-suited for capturing the full scope of engineering research. After all, in engineering, you're primary goal is not to learn something about how the world works, but rather to change how the world works! So, instead of a hypothesis on how something works, I'd put up existing gaps in the ability to do something as a working basis. That will then put a focus on your research result as an extension of technical capabilities. In order to evaluate your research results, you would then have to show that your results actually close the existing gap.

Of course, also engineering research needs to understand something about how existing things work in order to be able to create something new. Hypotheses are suitable in engineering to clarify these preliminary things. In your case, you state that "the existing 'role-based access control' of MS-Windows does not solve some problems" - that sounds like a perfect hypothesis to test for. But verifying this hypothesis is certainly not the key step in your research, and maybe it has already been done previously. That's why I'd recommend not to focus on a hypothesis as the basis for engineering research (though one might use them to clarify preliminaries), but focus on identified gaps in current technical abilities.


Effectively, what you are doing is development of existing research, rather than designing something de novo. The notion of a research hypothesis is therefore somewhat inappropriate to such work, and you wouldn't write a paper describing this work specifying a definitive "hypothesis."

Instead, you'd write the paper emphasizing that your model does something "better," "faster," "more securely," or specifying whatever other accomplishments advance your work from the previous state of affairs. Your thesis should then show how that is accomplished, and give some evidence thereof.

  • "better," "faster," "more securely," sounds like a testable hypothesis to me...
    – StrongBad
    Commented Aug 25, 2013 at 15:35
  • It's a testable hypothesis if the method came before the hypothesis. I think this is more of an "implicit" hypothesis: we can do something better, and we went out and found something better that did it.
    – aeismail
    Commented Aug 25, 2013 at 17:20
  • @DanielE.Shub very similar to my hypotheses that I have used in my PhD and published work.
    – user7130
    Commented Aug 25, 2013 at 20:06
  • @aeismail exactly we hypothesised that we could do something better and we went out and tested that hypothesis. Even if the "better" method doesn't exist yet, you can still hypothesise that there is a better solution.
    – StrongBad
    Commented Aug 25, 2013 at 20:16
  • @DanielE.Shub: Perhaps that is a hypothesis. However, it's still a weak structure to hang a thesis on.
    – aeismail
    Commented Aug 25, 2013 at 21:43

You must log in to answer this question.

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