Tuesday, January 26, 2016

An important but basic skill: how to quickly "read" a scientific paper

Basically, look at the figures.

The amount of literature we might read is increasing exponentially. It is overwhelming. We all need some strong filters to focus on a few papers. This can save a lot of time.

The question is really, "Should I read this particular paper?"
This means answering two questions.
1. Does the paper contain some results that are of interest to me personally?
2. Are the results valid and important?

These days I tend to only look at papers that someone else recommends to me or are cited or linked to in papers I have decided to "read" using the procedure below.

The quickest and most efficient way to answer these questions is.

a. Read the title and abstract.
Is there potentially something of concrete interest to me?
If not, ditch the paper.

b. Look at the figures.
Are they comprehensible? If not, ditch the paper.
Do they contain new results, I did not know about? Are they interesting and important?
Are they valid? Do they make sense in certain limits I already know about? Are they consistent with other work?

If I get positive answers, only then will I actually look at details in the paper, such as the methods used. Then I may actually read the paper properly, perhaps even trying to work through some details.

Many senior scientists follow similar procedures.

Corollary. There are important implications of this for you when you write a paper.
Pick your title carefully.
Polish your abstract.
Work particularly hard on your figures and their captions.
Indeed, my mentor John Wilkins taught me to "write" a PRL by first constructing polished figures and captions, and then writing the text. The abstract is written last.
If you don't the paper may never get read, even though it does contain important and interesting results.


  1. Also read the conclusion, if one exists. Some journals insist on consise, well-written conclusions. Others -- notably Science and Nature -- are very inconsistent and often have papers with poor or non-existent conclusions.

    1. Thanks for the comment.
      Although, I agree the conclusion should be useful, often it is not.

      I fear the reason that Science and Nature papers often have "poor or non-existent conclusions" is because they (too often) don't report conclusive results but rather hype and speculation.