Reading a paper backwards
A sample article, included so the writing section is not empty at launch. Replace or delete it once real pieces are published.
This is a sample entry included with the site so the index has something in it. Edit it, or delete the file content/articles/reading-a-paper-backwards.md, once real articles are up.
The instinct with a new paper is to start at the abstract and read forward. For a student that order is close to useless, because the abstract is written to persuade and the introduction is written to justify. Both arrive before you have any way to judge them.
A better order
Start with the figures. A well-conducted study is legible from its figures alone: the axes tell you what was measured, the error bars tell you how confidently, and the sample sizes in the captions tell you whether to care. If the figures do not make sense on their own, that is information about the paper, not about you.
Then read the methods, slowly, and ask one question throughout: what would this design fail to detect? Every method has a blind spot, and the discussion section will rarely name it.
Only then read the results text, which is mostly narration of the figures you have already understood. Read the discussion last, and read it sceptically — it is the only part of a paper where the authors are permitted to speculate.
Why it works
Reading in this order forces you to form your own view before you are handed one. By the time you reach the discussion you have a position, and you can notice where the authors' reading of their own data is more generous than yours.
That gap — between what a study showed and what its authors say it showed — is where most of the useful learning sits.