To Read One Research Paper, One Must Read Many Research Papers
Compared to a lot of my classmates, I came into my PhD lacking research experience. I only did research for two summers: one project the summer after my freshman year of undergrad, and another the summer between the first and second year of my masters. But most of my “research experience” was writing mediocre code while I read textbooks on the side to learn the basics of the subfield I was working in.
When I got to graduate school, the first new skill I had to acquire was how to read research papers. And at first, I found reading papers to be incredibly challenging.
For my first reading assignment, my research advisor gave me a classic paper in the field of bacterial chemotaxis called Physics of Chemoreception, co-authored by Harvard physicists Howard Burg and Edward Purcell. It’s a brilliant paper that perfectly encapsulates the very best that the field of biophysics has to offer.
The paper is so brilliant, in fact, that my advisor was more than willing to let me take a month to read through it in full. Yes, that’s how long it took me. To be fair, I had other things on my plate: new student orientation, the start of classes, and other textbooks that I was juggling on the side. But still. A month?
Why did it take me so long? Well, it was a huge adjustment going from reading textbooks to reading a research paper. While Burg & Purcell is well-written, textbooks are explicitely meant to be pedagogical. Textbooks usually have a lot of exposition. They slowly introduce and motivate new ideas at a comfortable pace. A well-written textbook feels like going on a brisk jog up an evenly-sloped hill. In comparison, one’s first research paper feels like sprinting head-first into a brick wall.
Research papers are just so dense—especially when compared to the undergraduate-level monographs that I was used to reading. Concepts are introduced in rapid-fire fashion. Sometimes, especially when they are meant to be novel, new ideas are carefully explained in the main body of the paper. But other times, whole ideas are casually introduced with the explanation delegated to a quick citation. And then you are presented with a dilemma: do I delve into the citation, hoping to better understand what the author is referencing? Or do I charge on ahead, hoping that I can grok the concept just from the context of the paper alone? A typical research paper has dozens of citations. And for each of them, you have to implicitly make this decision. This is only further complicated by the fact that not all citations are made equal. Some citations are functional: the author derived an important idea, concept, or fact from pre-existing work, and they are citing that previous work both to properly attribute credit and to help the reader track the genealogy of ideas. But other citations are more political. These are either landmark papers in the field that you just have to cite, or they can be the product of the ubiquitous academic politicking. Maybe it’s the paper of a frequent collaborator who read a draft of the paper. Or maybe an obstinate reviewer kept hinting that surely, surely you must be aware of this PNAS paper that showed how the Tar/Tsr fraction changes as a function of the viscosity of the media (a paper that the reviewer just so happened to be the first-author of).
Clearly, a paper a month was not going to cut it. I was going to need to get faster. But how?
This would be where you are supposed to ask the professor, postdocs, or senior graduate students in your lab for advice. Here is a stylized version of the advice they might give you:
-
You read the abstract and make sure you understand all of the key terms, the context of the work, and importantly, what the main (new) results are.
-
Then you skim through the paper, paying some attention to the text, but mostly focusing in on the key figures—especially those that directly pertain to the main result of the paper.
-
If something in one of the figures doesn’t make sense, you take a closer look at the methods to clarify their approach. If their method surprises you, you look for their rationale for why they chose that specific method as opposed to the “obvious” method given the subfield.
-
You skim through the discussion section to confirm that the author’s interpretation of the implications of their work matches your own interpretation.
This checklist (that I made up) probably isn’t a bad description of what it would look like to peer over the shoulder of a top-level scientist who is reading a paper fresh off the Arxiv.
There is only one problem. It’s terrible advice for someone starting out their graduate career.
There is a difference between the process needed to acquire a skill and the internal experience of having that skill. Often, people with expertise have trouble getting into the heads of people who lack that expertise. In their minds, they approach a paper with a butcher-like “Identify the key information. Get in and get out.” But most of their expertise is encoded in a subconscious evaluation process that allows them to quickly discriminate between what’s new and relevant information, what’s standard background knowledge, and what’s speculative and should be taken with a grain of salt.
If you are an aspiring researcher and you try to read a paper like your professor does, you will likely end up frustrated. Rather than accruing knowledge over time, you will end up with a shallow engagement with the literature. You will stagnate as you aren’t able to retain the information in any given paper due to never developing the mental schema needed to contextualize the knowledge.
But then how do you learn how to read research papers?
There is a classic Scott Alexander essay called “Should You Reverse Any Advice You Hear?” The main idea is that different people need to hear different advice. The stoner-slacker-party-animal needs to hear that their studies are important and they should take school more seriously. The type-A-straightlaced nerd needs to hear that life is meant to be lived and they need to take risks sometimes and have some fun.
But importantly, because who we are is correlated with who we interact with, we often get exactly the wrong advice. If you are a party animal, then you’re likely to be friends with other degenerates who will be egging you on in their degenerate ways. “One more hit, bruh.” And if you are an overachieving nerd obsessed with the next professional and academic hoop, then your friends are likely to be the same, reinforcing the importance of winning the competition at all costs.
Successful academics are people who have completed every task in front of them their entire lives—no questions asked. They were stellar students in high school, allowing them to get into great undergrad programs. They were excellent students in undergrad, juggling their demanding coursework with obtaining research experience in a lab. And they were excellent grad students, putting in late nights to get their projects over the finish line and doing all of the little things to truly obtain mastery in their subfield of choice.
Let me be blunt: Academics are deeply, irrevocably broken people who are looking to fill the gaping holes inside of themselves with achievement and external validation. They do everything 100% to the best of their abilities because that’s the only way they know how to.
But as you go further in your career, time management starts to increasingly be a factor. You aren’t a cute undergraduate anymore without responsibilities or expectations. If you are a postdoc, then you are put on an immediate timer to produce work that will make you competitive for a tenure-track position. And when you are a professor, you go from being a full-time researcher to more like being the CEO of a small company with all of the implied administrative overhead. You don’t have time anymore to go down that rabbit hole to make sure you really understand why the theorem only requires T1 separation and not Hausdorff. Get the information you need to move the current project forward and then get out.
It’s been a little over a year since I’ve started graduate school, and I’ve gone from being able to read a paper in a month to being able to read multiple papers a day. I’m currently in the brief liminal phase where I’ve acquired enough expertise to feel like I know what I’m talking about, but I’m not so far removed from my novice days that I can no longer empathize.
My advice would be that if you want to read one research paper quickly, you should read many research papers slowly. Read them in detail and without cutting corners.
Okay, so “work hard” isn’t the most profound advice in the world. Here’s an actual trick: research papers don’t have to be a slog. Most of them—even the well-written ones—will be grueling. But there will be some small fraction of papers—probably as few as 1% of papers—that due to a combination of their content and their writing style will be genuinely enjoyable to read. Find those papers.
I would say the difficulty is that most people are first forced to read research papers in the context of a specific project where they don’t have much freedom to choose the papers they have to read. This is a shame. For me, I got better at reading papers when I started doing it in my free time, focusing on random topics that I was interested in—with no bearing on the current project I was working on.
This isn’t meant to be radical advice. When children are first learning how to read, we teach them the basics like how to decipher letters. But otherwise, we let them read what they find enjoyable. This will inevitably be popular books like Percy Jackson and Harry Potter. And you know what? This does build reading practice such that when they get to high school and have to read “serious” books like the Iliad, they are better prepared for it. But imagine a world where people weren’t taught to read until high school, and they were only allowed to learn by reading Faulkner?
Here’s some practice advice for finding papers that you might like to read:
-
Think about the niche topic that you are hyperfixated on.
-
Identify papers in that subfield. Focus your efforts on papers by famous researchers, as there is a correlation between the researcher’s fame and how well-written their papers tend to be. But this correlation is not so strong that you should dismiss papers by less famous researchers, either.
-
Skim through the abstract and introduction of a couple of papers whose titles grab your attention. The vast majority of the papers will be boring and hard to read.
-
But there will be a tiny fraction of papers that just grab you. Sure, you don’t understand every derivation or piece of jargon. But there will be an energy that pulls you through the paper. If you’re not sure whether a paper has grabbed you, then it hasn’t. Keep looking.
-
Figure out who is most responsible for the writing style of this paper. If the first author (in fields that have authors listed by importance) is a full professor, then it’s likely them. If the first author is a graduate student, then you probably want to look at the last author of the paper. I find that a paper being outright fun to read is mostly just a function of the author. If I find one paper by an author fun to read, then that’s a strong indication that the author and I share similar intellectual and stylistic taste, and that I will find a significant fraction—if not an outright majority—of their papers fun to read.
-
Start reading a bunch of papers by this author. You will not only gain knowledge about something you are interested in, but you’ll also start to subconsciously and consciously pick up the genre conventions of scientific writing.
(Why scientific writing is so genre-specific in the first place is an interesting question. The most commonly given reason is that it’s the style that allows experts to communicate clearly and unambiguously with each other while remaining concise, though I am somewhat suspicious of that explanation.)
When you read a bunch of papers by the same author on the same topic, certain things become clear. For one, you will see that most information in a paper is background information. This explains why experts can read very little of the written content of a paper and still extract most of the relevant information. Second, the introductions will be written in incredibly similar ways between papers. This will not only help you learn the subfield in question by having key details repeated over and over again, but it will also teach you what sort of speech markers writers adopt when they are relaying background information. This will be important when you have to do the work of reading a paper in a field in which you don’t have expertise, as it’s more difficult to disentangle the background information from the novel contributions.
After you have read a couple of papers, all on the same topic, you will naturally feel a strong compulsion to skim when you read through parts that you are familiar with—even if your original intention was to read the paper in full. You will also start to feel in your body the emotional contour of a scientific paper and how the different sections flow from one to the next and build on each other. Just like how a pop song has verses, choruses, bridges—all with their own unique feel and purpose in the overall song structure—scientific papers have emotional highs and lows.
My overall advice is that reading papers should become a hobby in the same way that listening to music or reading fantasy novels is a hobby to most people. So it helps to dive in and just starting reading the stuff that you like!