Do you remember when you read your first scientific paper? For me it was hard. Some parts I did not understand. Other parts were interesting. The structure seemed odd, with a narrative that did not flow. I read it from beginning to end a couple of times. I felt like the places I did not understand were my fault, not the writer’s. I then went to a class discussion of the paper and was amazed at all the flaws others could point out in the paper, flaws I did not see at all until they were pointed out. How could I become the person that could find the flaws and also see the strengths?
Now I know how to read a scientific paper and will share some tips that should help everyone. One thing about these tips is that they can also be very useful to keep in mind when you are writing a paper. A scientific paper need not be about science. It is simply a paper that backs up its claims with evidence, in the form of citations, either as inline references to other articles or as footnotes. My perspective, however, comes from my own field, biology.
The first thing to decide is why you are reading the paper. Did its title catch your eye? Are you working on something similar? Did you get it to review for a journal? Or are you reading it because it was assigned for a class? Why you are reading the paper will determine how you read the paper. In fact, you won’t read most papers. You will scan them for what you are looking for and then move on. It might be in the discussion where you find other references. It might simply be the data in the figures. You might be looking for a method to try. Don’t feel you have to read every paper through. I often don’t.
Most papers you come across, you will just read the title. This is true if you look at the bibliography of a paper, or get a table of contents emailed to you, or have set up a Google Scholar reminder on a certain topic. I see hundreds of papers a week in this way and only read past the title for a handful of them. So a strong title should tell what a paper is about.
The second thing you will read is the abstract. You will read hundreds more abstracts than full papers. The abstract should tell the whole story, not just what the paper is about. I have written about how to write a perfect abstract here. Such an abstract should identify the topic, state what has gone before, add what will be new here, give the results, then indicate how the field has changed because of the study. Sound impossible? Even complex work can do this in a few sentences. In fact, the paper I’m reading right now does it marvelously, here.
Where you go next might vary according to your specific interests, but I tend to look at the figures next. The figures and their captions alone should tell the story. All important findings should be in the figures. Look at the figures carefully to see what the research has discovered. Which figures are key and which are secondary?
OK, let’s say this is a paper you really want to understand and you are actually going to read the whole thing. There are lots of ways of doing this. I tend to read title, abstract, figures, results, methods. Then I make up my own mind what the paper is about, and what it actually shows. Then I read the introduction and discussion. This should tell how these results are put in context. I like to read them second because authors all too often make conclusions from their results that are grander than they actually warrant. If I read their own framing first, I might be taken in.
But if it is not a familiar field, I might actually read the paper in the order it is presented because the results won’t make much sense to me without the framing. Then I am less able to judge whether the work delivers on its promises. Assessing this is an important part of critical reading and is always something to address in a discussion of a paper.
What if you are totally new to reading papers and the paper you are supposed to be critical about seems totally fine and you can’t imagine what you might say? What to do? I have two tips. One is to find another paper close in subject to the first one and compare them. This will usually give you something interesting to say. You can find that other paper with a Google Scholar search of the topic, limiting to the more recent couple of years. Or it might be one actually cited in your paper. This can help a lot in a discussion.
The other thing to do if you can’t think of what to say, or if you really want to understand the paper is to follow the sample sizes and degrees of freedom on the experiments. Degrees of freedom should be independent and free to vary. All too often they are not, which means the statistics are incorrect. Tracing through these numbers can help you understand the exact experiments or observations of the paper and will often give you something to say.
Each paper you read is part of the great web of knowledge. The better it tells you where it fits in that network, the easier it will be to read. The more you know of that corner of the web, the more easily you will understand and critique the paper. Just remember to read critically, not to assume the study does what the authors claim it does, and to try hard to see what is new. Have fun!