We recently started describing some of the aspects of moving from the world of problem sets to the world of original research, and here we're going to continue expounding on that.
In this article we're going to describe the single most important thing we have ever learned when it comes to doing research. It is an incredibly simple idea, and one that can be applied to virtually all forms of research (and probably most areas of life, too). That said, it is (for us at least) incredibly difficult to be disciplined about. Simple, but not easy. It all comes down to the following simple statement:
"Have something to do when you get out of bed in the morning."
Huh?
Yes, that is the single most important piece of advice I've ever received. And after over 10 years of doing research, I still have to remind myself of it almost daily. So what exactly does it mean?
The word that we really have to define in that sentence is the word "something". And I think the reason this advice is sometimes so hard to follow is that we can often confuse ourselves about the meaning of this word, in this context. In particular, the "something" that we have to do needs to be concrete, precise, actionable, measurable, and trackable. Moreover, a lot of things might seem like they have these qualities until we look carefully at them.
Let's learn by example.
Example
Suppose you're a fresh grad student ready to take on the world (of research) and so you start reading some papers. Let's even suppose you've been disciplined enough to choose one paper to read — that's a good start. You read the abstract and a lot of words are new, but you power through.
Now you're in the introduction and the authors cite some review articles to set the stage. Nice. Maybe those review articles will be a bit more pedagogical, so you print one or two of them out and start reading those.
Now you're on page 2 of one of the review articles and it cites a few of the papers that it's summarizing, and one of them sounds really cool. You print that puppy out and start reading (hopefully you have a university footing your printing costs).
This paper was written in 1985 on a literal typewriter, and the notation is all wonky and out-dated. But that's awesome. Shoot, it's noon already, time for some lunch.
During lunch you realize that you've lost your focus and you go back to the paper that you started the day with, hoping to at least finish the introduction. But then it becomes total gibberish so you decide to instead focus on one of those review articles.
Before you know it, if you don't watch out, it'll be 5pm and you'll have read the first 1.5 pages of 6 different papers, as well as the first paragraph of 8 different Wikipedia pages. You're familiar with a lot more words now, but you don't really know what any of them mean — just kind of what they "feel" like.
That's All Well And Good
The picture I just painted is surely all too familiar to anyone who has done research in any capacity, and it can be fine to do on occasion, with one huge caveat: you must have a plan.
And the "plan" cannot be "eventually read all those papers I just printed out." The reason that's a terrible plan is that — I promise — each of those papers now on your desk will require at least 10 more papers to be printed out, and so your "plan" will forever be drifting further and further away from you.
If you're like me, you'll spend weeks thinking that you just need to "read a few more papers" before you'll start making sense of things and can finally get down to doing research. Then one day you'll wake up and realize that, despite having "read" dozens of papers, you can't really sink your teeth into any of it. Hopefully you're not like me and can save yourself some precious time.
Waking Up In The Morning
What does all this have to do with getting out of bed in the morning? Well, you might think that "reading all those papers" is the "something" that you need to do when you get out of bed. The problem is that this "something" is not very well-defined. What do you mean by "read"? Do you mean "read and deeply understand every detail", or do you mean "get the general idea of," or do you mean "simply expose myself to the new vocabulary"?
Without a very clear understanding of what your goals are, it'll be difficult to motivate yourself to start, and it'll be almost impossible to assess whether or not you've made any progress towards them. So, the "something" that you need to have to do when you get out of the morning needs to be concrete.
And I mean, very concrete.
An Example Of A "Something"
No two people's journeys through research are the same so it's hard to generalize too much. Therefore, let me give an example of a concrete "something" — the type of thing that it took me way too long to learn.
A fantastic "something" to have to do is: reproduce a certain paper. Now, before I go much further, this example is assuming two things. One, it's assuming that you're not already in the exciting part of your own research problem(s), because in that case you probably already have plenty of "somethings". And two, it's assuming that your field of choice doesn't require too much expensive lab materials (or other such things). Namely, the assumption here is that it's physically possible to reproduce a paper, either with pen and paper or with your laptop.
So, you've picked a paper (or, you've asked your supervisor to recommend one, or they have given you one, etc.). Now reproduce it. In its entirety, and in its full detail. The paper is already written so you don't need to reproduce the research (that'd be much harder), you just need to reproduce the results. To be clear, when I say you "just" need to reproduce the results, please understand that that might legitimately take 6 months.
It's important to understand that this is very different from "reading a paper." Lots of people read lots of papers. If you're an experienced researcher in your field, you can probably read a paper in an hour or two and gain a deep understanding of what the authors did in that paper and how impactful it is.
If you're a junior researcher, though, simply "reading a paper" will likely only give you the slightest whisper of a shadow of an approximation of the ideas that are in the paper. While that's something, it's certainly not preparing you very well to be able to do that kind of research yourself one day. For that, you need to reproduce the paper, and that's much harder.
Why Do I Have To Do This?
It's hard to describe all the ways in which reproducing a paper is beneficial, there are just too many. One such reason, though, is that it forces you to understand the material in that paper just as well as you would material from a textbook. Would you read a math textbook without fully understanding each of the proofs? Probably not, and if you did, those gaps in your knowledge will eventually be a roadblock. Reproducing a paper makes you reconstruct every detail of the argument(s) that you're reading.
If the authors wrote their own python script to do some calculations, then you should write one as well. (If they downloaded some huge library to assist them, please don't reproduce the whole library. Try to find your happy medium.) If you run your script and get a different answer, figure out why. This might take a week. You might have to email one of the authors with a question (but please, make sure you've given it a really solid go before doing this). You might even find a typo or a mistake in the original paper, in which case, great!
Now You Have Something To Do
The best thing about reproducing a paper, in my opinion, is that it gives you a real "something" to do when you get out of bed in the morning. Everything you do and/or learn needs to be geared towards understanding that paper.
Need to read some chapters in another textbook to understand something in the paper? Fine. But now you're not just "reading the textbook" — you have a clear vision of what you need to get from that textbook before coming back to your paper. Need to print out a paper that your paper references? Cool. But you know what you're coming back to when you're done with this other paper.
Every single day you work like this, you will be getting closer to having reproduced your paper. You now have a "north star" guiding your work, which is incredibly important when you no longer have the crutch of "lectures" and "problem sets" to make it clear that you're making progress.
As mentioned above, this paper that you're reproducing might legitimately take 6 months to "complete." But here's the best part...
It Won't Always Take 6 Months
The first paper you reproduce will take a long time, the second might take half that time, and the third will probably be another factor of 2 shorter. This is because for each paper, you're building your scar tissue (and the number of useful python scripts that you have on your laptop that you might be able to re-use). You're learning how your chosen field works, where the bottlenecks and devilish details are, what assumptions you can and can't work with.
You'll be pleasantly surprised how fast this process snowballs. There'll be a growing amount of literature that you're intimately familiar with — that you're an expert on (because you've reproduced them). You'll be able to go to talks from people in that field and ask them meaningful questions. You'll start to identify what types of questions haven't been answered yet. Even better, you'll start to identify what types of questions haven't been asked yet. In short, you'll be doing research!
Plenty Of Other "Somethings"
There are plenty of other "somethings" out there — we've just focused on the "reproducing a paper" type of "something". While we do think that this is one of the very best "somethings" around, you and/or your supervisor may come up with other types of things. But you must have a thing. A really concrete thing to do, and be able to measure your progress on, when you get out of bed in the morning.
Once you're no longer taking classes — or once classes become a distant second priority to your research goals — it's very easy to lose track of the progress your making. And honestly, it's very easy to stop making progress altogether. Lectures and classes have built in to them a linear framework that make "progress" easy to obtain and track.
Once you're in the world of research, you need to build this framework for yourself. In our opinion, it's much, much, easier to do this when you have something to do when you get out of bed in the morning.