To read or not to read
April 19, 2012 9:47 AM   Subscribe

How to read a paper is a series by Trisha Greenhalgh in BMJ, the British Medical Journal, that explains how to critically read and apply the biomedical literature. Deciding what the paper is about. Assessing methodological quality. Statistics for the non-statistician: parts I and II. Drug trials, diagnostic and screening tests, economic analyses, systematic reviews and meta-analyses (PDF), and qualitative research (PDF).
posted by grouse (14 comments total) 100 users marked this as a favorite
how to read a paper:
1. read the abstract.
2. skip to the discussion.
3. look at any graphs.
posted by rebent at 10:17 AM on April 19, 2012 [8 favorites]

Many of these require subscription to the BMJ. Are there public links for those non-pdf links?
posted by bonehead at 10:19 AM on April 19, 2012

ARGH. I checked all the links before posting this. It looks like if you follow the links via the main page, you will be able to read the links without payment. My mistake for stripping out the magic referral codes.
posted by grouse at 10:22 AM on April 19, 2012

I've just looked at a bit of this, but it's superb; kudos to BMJ for doing it.
posted by escabeche at 10:24 AM on April 19, 2012

how to read a paper:
1. read the abstract.
2. skip to the discussion.
3. look at any graphs.

Not enough. You have to read the methods section if you want to know what's actually being plotted on the graphs.
posted by escabeche at 10:24 AM on April 19, 2012 [5 favorites]

This is a wonderful resource. Accessibility of scientific research is something that I've been thinking about a lot, lately. The sciences are ostensibly mostly transparent (or at least, most scientists agree that most research should be mostly transparent... there are a lot of caveats) but in practice if one wants to be able to access primary sources in the sciences there are many barriers that must be surmounted. (Closed journals, I'm looking at you here. Also, researchers who never took a composition course, you too.)

One of those barriers is the rather obtuse formatting and style guidelines that scientific publications seem to have settled on as a standard, and anything that helps people who are interested in reading research papers cut through the stuffy language and rigid formatting and be able to understand and assess the underlying research is a Good Thing as far as I'm concerned.

If I had my druthers I'd oversee significant reform of scientific publication. Here are some things that I might try:
  • Protocols should be pulled out into their own section of the paper and referenced in the Methods section rather than spelled out in detail every time they come up. If I'm using the paper to guide my own research then I shouldn't have to comb through it to find out exactly what the PCR conditions for my next reaction should be, and if I'm not trying to replicate or build off of the paper then I shouldn't have to read about how many microliters of magnesium chloride went into the PCR master mix.
  • Technical terms that are specific to the subspecialty should be briefly defined the first time they appear, or defined in a glossary, so that at least people whose understanding of the subject is one or two levels broader than that subspecialty can understand them. Ideally anybody should be able to read any paper, and while this ideal may be hard to achieve in practice researchers should at least make an effort.
  • Graphics should not just explain what their components are and what they are illustrating, but should also come with a one-sentence takeaway of what the viewer is supposed to see and why the graph supports that. Graphics can be just as obtuse as text and on close examination sometimes don't even support the research they purport to support. It should be easy to see what a graphic is trying to show so that readers can decide if it actually does show that.
  • Relevant graduate and undergraduate programs should make it standard practice for students to take a science writing course. A lot of papers fall down in clarity simply because they are not very well written in a mechanical sense, with ambiguous phrasing or tortured clauses that make it very difficult (sometimes impossible) to understand what the researcher is trying to say. Explicit training in writing scientific papers should be a significant component of education for people who are going to be publishing such papers down the line.
That's just a start. If I ever get a chance, I might try some of these things with my own research someday. Of course, journals can be very rigid about what kinds of things they do and do not want to see in a paper, and those standards don't always serve the interest of scientific clarity. But still, I feel like there's a lot of room for improvement and if I ever get a chance to experiment with some of this stuff, I will. In the meantime, these sorts of guides are a valuable resource for anyone who wants to be an active consumer of scientific research. We need more stuff like this.
posted by Scientist at 10:35 AM on April 19, 2012 [5 favorites]

I'm an academic reference librarian and I work with a lot of nursing students. Some of this is similar to the advice I give them. I'm sharing with all of them on our library's nursing resources website. Thanks for posting.
posted by mareli at 10:44 AM on April 19, 2012

Also, does anybody know of any resources or guides that are similar to this one but which have a broader scope than this? This particular resource seems to be focused mainly on how to read clinical research papers, which are definitely their own breed. Much of the advice, while excellent, doesn't really translate to the field as a whole. Has anyone run across a resource that tries to do this sort of thing but for a more general audience?
posted by Scientist at 10:51 AM on April 19, 2012

I've worked for some colorful researchers in the past. By that I mean world-class assholes. But they did occasionally drop a word of wisdom. One told me, the most important section is the methods. The methods get the results. The limitations in the methods limit the scope of the results. Everything else is decoration.
This is what science is about and it is what scientific paper reading should be about: experiments, not claims, not discussions, not summaries in the abstract.
posted by dances_with_sneetches at 2:55 PM on April 19, 2012

Read the abstract? I've yet to meet an abstract that didn't feel like the linguistic equivalent of RPN. I mean, yeah, technically there's data in there, but it's sometimes quicker to just read everything else than try to figure out what they're saying in too few worlds in the abstract.
posted by Kid Charlemagne at 3:44 PM on April 19, 2012 [1 favorite]

The CMAJ did this years ago and they are still online:

Part I

Part II

Part III

Part IV

Part V

There are more that came in later years, but these were extremely helpful...
posted by Alexandra Kitty at 4:04 PM on April 19, 2012 [3 favorites]

I've taught how to write an abstract. Below is an example I used, an abstract from Science that was brief, elegant and clear.

Global Trends in Wind Speed and Wave Height


Studies of climate change typically consider measurements or predictions of temperature over extended periods of time. Climate, however, is much more than temperature. Over the oceans, changes in wind speed and the surface gravity waves generated by such winds play an important role. We used a 23-year database of calibrated and validated satellite altimeter measurements to investigate global changes in oceanic wind speed and wave height over this period. We find a general global trend of increasing values of wind speed and, to a lesser degree, wave height, over this period. The rate of increase is greater for extreme events as compared to the mean condition.
posted by dances_with_sneetches at 4:58 PM on April 19, 2012 [5 favorites]

That's a really good abstract. I'm going to add it to my collection of writing samples.
posted by wintermind at 5:52 PM on April 19, 2012

Ah! Funny coincidence, I just looked at the screening and diagnostic testing article earlier this morning. Uh, comparing the performance of simple binary diagnostic tests is more complicated for me than I want it to be.
posted by mean square error at 6:28 PM on April 19, 2012

« Older Pain cake   |   Go long on hot saurce Newer »

This thread has been archived and is closed to new comments