Over at Tor.com, Jo Walton is surprised that people skim over boring bits of novels. While she explicitly excludes non-fiction from her discussion, this immediately made me think of Timothy Burke’s How to Read in College, which offers tips to prospective humanities and social science majors on how to most effectively skim through huge reading assignments for the information that’s really important.
I’ve mentioned this before, but I don’t think I’ve done a science version. I’ve been doing more reading of journal articles lately than I have in a while, though, and it occurs to me that similar skills come into play in the sciences, so I thought I would provide a quick guide for the most effective way to skim a scientific paper for the information you really need. This is, as most of my posts are, primarily applicable to experimental papers in physics, but I’ll try to make it as general as possible.
The first and most important point is to Know What You’re Looking For. Different bits of information are found in different places and in different forms, so what you’re looking for will determine where you look, and how you find it.
For example, if you’re just trying to get a general sense of what a given paper is about, it’s often enough to read only the introduction and conclusion. If you’re just after a specific numerical result, it’s probably in the abstract, or toward the end of the paper.
You should also be aware that what you’re looking for may not be in the paper you’re reading. If you want a sense of the context of a field, you’re often looking for a reference to earlier work, possibly a review article. If you want the gory details of a measurement technique, you may very well be looking for some reference to an earlier or longer paper by the same group (a sentence of the form “using the method of [citation of earlier paper]”), or, even more annoyingly, some online supplement to the article you have.
Next, you should Know the Structure. While real scientific papers don’t follow the strict “Abstract-Introduction-Procedure-Results-Conclusion” format we force undergraduates to learn, there are some similarities. Most papers have an introductory section that contains nearly all of the references to past work by other people, a middle section where they talk about their results and how they got them, and a concluding section with a discussion of possible implications of the results.
If the paper you’re reading is in a field you know well, you can probably skip the first page or so, and get right to the part where they talk about their actual measurements. If it’s from another field, you may not be able to read much past the first page, at least in detail. If you just want the result and what it means, you want to look near the end.
There are some subfield-specific sections to know about, as well. For example, a paper about a precision measurement will usually include a section devoted to the analysis of systematic errors. This can be a great source of colorful anecdotes (elevators needing to be disabled to ensure good data, and that sort of thing), but unless you’re a hard-core precision measurement junkie, you can probably skip most of it to get to the final numbers.
The fastest way to get an idea of what’s going on is to Look at the Pictures. Most experimental papers will have figures showing the arrangement of their apparatus, and often include things like timing diagrams showing how they did something. You can figure out a lot from these figures, and their captions. this is especially true in journals with tight page limits, as the captions are usually printed in a smaller font, so some experimental details are often crammed in there where they take up less space. In Science and Nature papers, figure captions don’t seem to count against your page limit, so you’ll often see really massive figure captions in those journals, that are like mini articles unto themselves.
The final data are generally presented in graphical form as well, so if you want a quick idea of the quality of a paper and its results, you’re generally looking for something that will be in a figure, so that’s the first place to look. If you find a figure with data that look interesting, but aren’t immediately clear from the picture and caption, scan the nearby text for the figure label (“Our results are plotted in Fig. 2”), and read backwards and forwards from there.
Speaking of that, Learn to Read in Both Directions. If you want to skip around in a paper, you’ll need to know how to read backwards from a figure reference as well as forwards. If the paragraph about the results contains an acronym whose meaning isn’t obvious, scan backwards through the text for the first appearance of it (or at least a point before which is doesn’t appear– some authors aren’t good about putting the acronym right with its definition, but expect you to figure it out when it appears in the next line). If one of the axes of a graph is labeled by an unfamiliar Greek letter, it’s probably a combination of some measured quantities and physical constants, so you need to scan backwards until you find the first appearance of that symbol, which ideally will be in a simple equation defining it.
I’m sure I’m leaving out some critical tips, but these are the ones that strike me as the most critical tips for quickly finding what you need in a scientific article. If you know how to do these things, you can get a good idea of what’s going on in a paper much more quickly than you can by reading the entire thing. Sometimes, this is enough to get you the critical fact you need, and let you move on to something else; other times, it just shows you that a paper is good/important enough that you need to sit down and read the whole thing.
It also helps to be familiar with the reference style that journal uses. There are minor house variants, but the ones that I have seen fall into three basic categories:
1. Citation by name and year: “We build on previous results that X (Smith and Jones, 2008) and Y (Fulano et al., 2009) are true.” Common in some subfields of physics, including the one I work in. Helpful if you are already somewhat familiar with the literature, because you may have already read one or both of those papers. But if you found the article via a citation search, you may have to skim the whole thing to find the citation.
2. Sequential numbering: “We build on previous results that X [1] and Y [2] are true.” You then look at the reference list to find that [1] refers to Smith and Jones (2008), etc. APS and AIP journals, as well as Science and Nature, use versions of this scheme. More compact than the first option, and may be faster if you found the article via a citation search (look up the reference number, after which it’s easy to find the first citation. But you have to look at the entry in the reference list before you can say, “Oh, yeah, that paper.”
3. Alphabetical numbering: “We build on previous results that X [42] and Y [17] are true.” When you look at the alphabetically arranged reference list you find that the 42nd entry on the list is Smith and Jones (2008), etc. Not common in physics, but I have seen it in some math and/or computer science papers. Shares compactness with sequential numbering, but otherwise combines the disadvantages of the previous two methods.
Some tips I used
Reading up on a new technique/method? Look mostly at the intro section and try to find review or at least more general articles, especially if there is a canonical review for this area (and if there is, you want to find it). Articles listed as refs in papers from different groups were the next thing I tried to look up.
If you are trying to understand/duplicate equipment, don’t discount papers that are using it differently than you will. Does everyone use the same prep, manufacturer, analysis, etc?
You have to understand what you are trying to get out of the paper. I have pulled up papers and looked only at refrences. Sometimes its a method or technique and I don’t really care what the results were. Sometimes I was simply looking to see if there was analysis or plots that I should be doing but wasn’t.
It is becoming more common for journals to offer some “supporting materials” on a website. I wouldn’t do this for every paper, but it is worthwhile to check and see if there is anything there. I have found structure/data files (fabulous if you want to use a similar structure in a simulation, or need it to construct a visualization), fitting parameters, and sometimes very specific details used in the experiment or analysis.
Good advice (although I admit I skimmed it!).
I think the most important ingredient for being able to read scientific papers quickly and well is practice. If I compare myself today to how I was in the beginning of grad school, I suspect that my ability to concentrate is worse, and I often FEEL dumber, but man am I better at reading papers. Unfortunately it’s the only thing I do professionally that I’m dramatically better at now than I was a decade ago.
I do most of the things Chad mentions, but I think the biggest reasons for my improvement are practice and familiarity with the field (which comes partially from reading papers).
One more piece of advice: shop around for what papers to read. Often there are multiple groups doing similar work, and some may be much better writers than others. Once you figure out one paper, parsing the other will be much easier.
You sometimes need to be careful when you just pick out the results from a paper. For instance, if the graph is labelled in “Saturation Intensities”, you won’t know what definition they’re using (I’ve seen three so far, differing by factors of 2 or 3) unless you go and look.
More importantly, there may be a vital caveat (e.g., 100 samples were tested and the data shown are from the best sample only) buried somewhere in the meat of the paper, and if you only look at the graph you’ll be misled.
Another thing to remember about references: some journals (e.g. Physical Review Letters) mix notes with the references. You’d skip these notes if you were skimming anyway; but more thorough readers may want to check that they’re not missing interesting asides.
I often skip the actual equations in a paper, unless I need to actually use them myself. It saves the effort of decoding whatever non-standard symbols the authors have decided on, and I’d never understand an abbreviated derivation without working it through myself anyway.
The three citation styles listed by Eric Lund are also common in the biology, geology, and environmental sciences journals I tend to read. You were successful in making this advice general to science, in my opinion.
Thanks for this, I’ve got a small pile that’s becoming more urgent (along with the 5 other really really really important things I need to do but am instead procrastinating on blogs).
I guess for the sort of papers I read you should replace “look at the pictures” with “look at the equations.” If I’m searching for something specific, that’s indeed the fastest and easiest way to figure out if the paper is useful. (Forget about the abstract. People write all sorts of fantasies in their abstracts.) It doesn’t work well with papers older than, say, 30 years or so, because notations and textbook terms change.
Reading is not like watching TV or playing computer games, wasters. Nature Materials
So what is your point “Nature Materials”? And breathing is not like eating, so far so clear, but what can I “gain” out of this?
Are you philosopher?