The structure of a scientific paper

This note is intended to capture typical structures for scientific papers and specifically to track how conventions differ between research fields (e.g. humanities vs. computer science).

A paper that presents experimental work is often structured as follows:

  • Title, abstract, keywords
  • Introduction: What is the problem / research question?
  • Materials and methods: How is the question addressed or solved?
  • Results: What did I find out?
  • Discussion: What does it mean?
  • Acknowledgements
  • Literature cited
  • Appendices

Ok, some questions and remarks:

  • I often see that a discussion of literature goes in its own "background" or "literature review" section
    • This is very common in theses, because here showing a good understanding of the field is one of the learning goals
  • I see a possible meaningful difference between a "methods" section and an "experiment" section
    • E.g. "methods" section could contain an explanation of used algorithms
    • Whereas an "experiments" section could outline the details of experiments (and their hypotheses?)

TODO:

  • Thesis structure
  • Paper structure in non-experimental papers, e.g. philosophy

Tips on writing order

  1. Start with having your figures and tables ready as soon as possible. This is the stuff you write about, so it's most efficient if they are already in good order
    • Personally, for course papers I have often had to rewrite quite a lot because I didn't have my results finalized yet; sometimes results changed, hence figures changed, hence the accompanying text needed to be rewritten.
  2. Write methods and results next
  3. Write the discussion before the introduction. This way you avoid making claims in the introduction that are not sufficiently supported by the discussion of your results
  4. Then write conclusion and introduction
  5. Then abstract and a catchy title
  6. etc.

The discussion is key

The discussion section is arguably the most important section of an article.

Tips and thoughts:

  • The discussion needs to correspond to the results, but without just repeating the results. Don't make claims not backed by the results.
  • Compare your results with previous results (stated as relevant literature e.g. in the literature or a background section)
  • Avoid unspecific expressions, try to be precise and quantitative
  • Do not introduce new ideas; the proper place to do this is the introduction
  • If multiple possible interpretations of the results are possible, you still need grounding of these interpretations (i.e. you should show why they are good and plausible interpretations; this is not just a work of imagination)
  • If you cannot clarify an idea, this may mean that you need more results and that more work is needed there.

Tips for a good introduction

  • It is good to state the purpose and the general research strategy, but make sure to not overlap with methods and results etc.
  • State your contribution but without overpromising. It is probably not "paradigm-shifting" work and even if it is, this is probably for others to determine. (I suspect no paradigm shifting article has called itself such; it's shitty writing and moreover, whether it actually achieves something also largely depends on its reception).
  • Sketch the bigger picture first and guide towards your research problem; zoom in gradually
  • State hypotheses and objectives at the end of the introduction!

Sources

  • https://www.elsevier.com/connect/11-steps-to-structuring-a-science-paper-editors-will-take-seriously