2

Here is the situation:

I am currently writing a thesis which uses a behavioral experiment to answer the main research question. Its main aim is to test how a certain welfare scheme would affect taxpayer behaviour, and thus relates to a very broad strain of literature on taxpayer behaviour. I have found a paper which uses (in my opinion) a very clever experimental procedure to test how behaviour is affected by the way taxes are used. Therefore I decided to use this experimental procedure and adapt it to the specific welfare scheme I am interested in so I can use it to answer my research question.

Currently I am in the progress of writing my thesis and was wondering where the line ends from building on previous work and plagiarism. Because the experimental procedure is similar in it's structure, the model used for the formulation of my hypothesis is also an adapted version of the model in the original paper. Consequently, the data I have also takes a similar form, which will result in taking similar steps for a correct analysis of this data.

So, would it be considered plagiarism if I have adapted an already existing experimental procedure, adapted the (mathematical) model previously used to formulate hypothesis, and use a similar structure of hypothesis testing (that is: first using statistical test A, then B, then C, etc on my own data)?

I have obviously referenced a great deal to this original paper, and even explained why I think the experimental procedure is so good and fitting for answering my question. Obviously, I am using my own wording and data, but was wondering if something like this can be seen as 'structure plagiarism' or is something desirable as it can be used for direct comparison of the papers.

  • "Plagiarism" means you do not cite the source. To find out if something you are doing is plagiarism, check that definition. – GEdgar Aug 24 '17 at 13:35
2

You are only building on the shoulders of giants, there is nothing wrong with that, as long as you cite the sources. Your hypothesis and your data are original, you are only applying a method that has the advantage of having been validated in the literature. This might make your work less original, but what is the point of developing a new statistical analysis if a perfectly fitting one already exists?

  • Is this also the case when I adapt an mathematical model by adding one or two variables? By doing this the most general implications of the model still stand (and I also formulate them as testable hypothesis) but also makes me able to test other hypotheses. To what extent should I describe the full model in my thesis? – ELC Aug 22 '17 at 15:52
  • You are improving an existing model, it's not plagiarism. The originality of your work does not lie in model-development, but in the modelling of some phenomenon. Concerning your second question, you should ask your supervisor. The answer entirely depends on your field, your doctoral school, etc. Personally, I would give a thorough description in the thesis, highlighting your additions. – Zep Aug 23 '17 at 7:14
2

You are confounding two issues here:

  • Is it plagiarism? Not if you cite this source for everything you take out of it.
  • Will it be enough to do only that? It depends.

For a master thesis and even more so for a bachelor thesis, it would probably be enough to replicate a known experiment in another context and see if it holds up.

For a doctoral thesis, this is not always the case, but it still might. If the original work made a strong conclusion that has had large implications and you can rigorously show that it doesn't replicate, this can be very interesting. If all you find out is "It worked in America. There were no reasons to think it wouldn't work in Europe and indeed it does work in Europe", then this is probably not a very substantial contribution.

That brings me your extension of the model. Be very specific about the reasons behind this extension. If you can motivate them well, the extensions will be the original contribution of your work. If you can't, the reviewer might assume you just wanted to add on something so that it looks like you were busy.

1

There are only so many ways to design behavioral experiments and so many ways to analyze them. If you consider experiments at any level of abstraction you will find that most of experiments in cognitive and social psychology can be reduced to a limited number of patterns. Consequently, the statistical analysis that follows is also basically the same from paper to paper. What novel research introduces is the creation of experimental materials, and more rarely, insightful experimental manipulation and control of important variables.

Successful applications of these sort become paradigmatic, and other papers re-use the procedure by adding more variables to rule out spurious effects or extend the scope of the explanatory adequacy of the approach. (For example Solomon Asch's and Henry Tajfel's experimental procedures have been reused and modified in a plethora of other studies.) Even direct replication of a study can be published, especially if the result is surprising and counterintuitive. (For example this finding https://www.newscientist.com/article/dn13596-male-monkeys-prefer-boys-toys/ about human-like gender differences in toy choice of chimps, was replicated by another lab.)

Plagiarism per se refers strictly to the reproduction of the material expression of a concept (e.g. written prose, or the experimental stimuli) which results from the intellectual labor of an individual. It cannot refer to abstract ideas. If you construct your own material, following the methodology of the original paper, it is your original intellectual labor. You can ask the permission of the original author to use the exact material, which, in case it is granted, will result in the studies being comparable.

Finally, with respect to updating the mathematical model by adding more variables, it is common place to do so in analysis of behavioral experiments with general linear models and their variations (including all type of analyses of variance for example), and it is also desirable because adding variables to these models can control their effect to the main variables of interest and provide a better overall fit to the data. I guess the same holds for other type of mathematical models, if the new model provides a better account of the observations and a deeper understanding of the effect you are studying.

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.