64

I am starting my PhD in statistics program next year, and I have some concerns. In my institution, there is a mandatory course that every PhD student has to take in lieu of taking one qualifying examination. The course requires students to read five academic papers and write a report, which will be graded by faculty members.

I was browsing through the reading list for this course, and I often find myself pondering over one twenty-page paper for days and weeks because I want to make sure that I understand all the math, concepts, and other details of the paper. I am supposed to make a report for one paper/month, and I am now quite worried about what I am expected to do as a PhD student because it feels like it takes me forever to comprehend a single paper. I feel discouraged because I have done very well in my bachelor’s and master’s programs in statistics, but I am having such a hard time comprehending these academic papers.

Is it normal for beginning PhD students to have a hard time comprehending academic papers?

  • 68
    One thing that's important to learn on the meta level is how deep your understanding of a single unit of knowledge (e.g. a paper, a concept etc.) needs to be a at a given time for a given purpose. Many beginners get lost in various rabbit holes because they overestimate how deep they need to go. – henning -- reinstate Monica Dec 28 '18 at 0:45
  • 1
    Different problem, but partly applicable: academia.stackexchange.com/questions/89032/… – henning -- reinstate Monica Dec 28 '18 at 0:51
  • 3
    Who selects the 5 papers? If you choose them yourself, either randomly or based on the fact that the title looks interesting, you will eventually discover what proportion of papers in your field aren't worth understanding anyway. In my experience, that is at least 90% of what gets published in high quality journals (ignoring the rest completely) but YMMV. – alephzero Dec 28 '18 at 11:16
  • 1
    I am disappointed that academia hasn't fully embraced the technology of hyperlinks. It would be great if papers which used high-level jargon would link to a source which could describe what that term actually means. Wikipedia has managed it. Why not academia? – joeytwiddle Dec 29 '18 at 4:21
  • 6
    I know it is tempting to think that authors should put a little effort into making the paper more accessible to non-experts, but the simple fact is that in many fields the group of people who would practically benefit from that effort is of the same scale as the group that doesn't need it (rather than being much larger). Abstruse Goose explains: abstrusegoose.com/272 . And writing papers is hard as it is: that technical language exists for a reason and it makes it easier for the in-group to process the papers. – dmckee Dec 29 '18 at 23:10
108

Yes, yes it is. But it gets easier. What most people don't realize when starting out reading academic papers, is that not every paper is an island (to paraphrase...). Terminology and 'lingo' is something you learn over time, and suddenly you realize that you no longer have to look up every second concept you stumble upon in a paper - it simply references stuff that you already know.

This is of course also the reason that forcing new students to read papers, and even write a report about them to make sure that the papers have been properly digested, is a very good idea. The fact that members of faculty even take time out to grade these reports, tells me that you are probably in capable hands.

  • 35
    "It gets easier. Every day, it gets a little easier. But you gotta do it every day. That's the hard part. But it does get easier." – Pedro Tamaroff Dec 28 '18 at 16:04
  • 1
    Indeed, this is the very process of research. – Lightness Races with Monica Dec 30 '18 at 1:15
46

I am in my second year of a statistics PhD. I have by now examined many hundreds of papers. Some of them are, I now know, of only tangential relevance to my research. Others are relevant but when I first saw them I did not understand them enough to know even that. And some are so relevant that I have sought to reproduce their findings and in doing so I have re-read them many times, often finding something new in them that I had not previously noticed. In parallel with reading these papers I have been learning about branches of statistics that I knew nothing of before.

The most important thing to realise, as I now have, is that academic papers are not generally written with the aim of explaining something to a novice, but rather are there to tell someone who is already expert how wonderful the author's research in that field of expertise should be seen to be.The day will come when you too can write papers that only a few people will understand, and to get there you will have struggled through countless really difficult papers.

Now, given that academic papers in statistics are bound to be hard to understand and that you have been asked to summarise as many as five of them in a short time, you have to accept that your summary will not be based on a complete understanding of all the material in all the papers. Imagine that you are a journalist rather than a researcher. You need to be able to write down:

  1. what question does this paper seek to answer?
  2. what is the answer?
  3. what reasons does the author give for that answer?

If you can do that you already have a good summary of the paper. To do it you do not need to understand all the author's reasons, still less agree with them all. Later in your research, maybe, you will recall one of these papers and realise that it is relevant to your own work: then you really do have to roll up your sleeves and understand in detail, but not now.

  • 5
    w.r.t. paragraph two I am reminded of the following: "Oppenheimer once said that most people gave talks to show others how to do the calculation, while Schwinger gave talks to show that only he could do it." (The paper I found this in, however, continues: "Although a commonly shared view, this witticism is unkind and untrue.") – davidbak Dec 28 '18 at 0:45
  • 5
    Re: "The day will come when you too can write papers that only a few people will understand, and to get there you will have struggled through countless really difficult papers". I'm hoping that that day never comes for you. When you write, write for the reader, not as @davidbak points out "to show that only [you[ can do it". It's possible to write an advanced paper that is readable. Learn (and strive) to do that instead. – Flydog57 Dec 28 '18 at 19:58
  • "The most important thing to realise, as I now have, is that academic papers are not generally written with the aim of explaining something to a novice, but rather are there to tell someone who is already expert how wonderful the author's research in that field of expertise should be seen to be" this is just the naked truth about Academia, sadly. – gented Jan 1 at 4:22
  • Another aspect is that we are used to reading things written by people practiced in writing. In Academia, you often have people who are not necessarily good writers trying to communicate very difficult content. – xdhmoore Jan 2 at 3:41
16

Yes. But it's a question of practice. More reading= More understanding. More understanding easier to understand a new paper.

But you do not need to understand every single phrase. For instance, if it is not exactly your field, I would jump over the methods section.

I got a recommendation during my Ph.D. first read Abstract, second Intro, third Conclusion. Some cases 3rd Results, 4th Conclusion. The more you read the more you'll understand.

Read, read, read. That's the key

Reviews on the field of study are excellent starting points

  • 1
    Reviews are a very good starting point. – pink.slash Dec 28 '18 at 0:29
  • @pinkslash The recommendation you had was good. Do you wonder why it is that authors do not write their papers in that order? I do not think that there is a good reason. – JeremyC Dec 28 '18 at 1:26
  • Completely agree with your advice. I'd also add "look at the figures" between Abstract and Intro. – user71659 Dec 28 '18 at 7:46
  • I like this advice a lot. In some sense it's like peeling an onion: (1) Abstract gives you executive summary; if promising then: (2) Intro & Conclusion bookends together give more detail to overall story; if still promising then: (3) read the whole Methods/Results heart of the paper and get all the gory details. – Daniel R. Collins Dec 31 '18 at 6:25
3

Unfortunately, a paper is limited in length by money (lots of publications have a per page cost after their standard length) or guidelines. As such, authors have to sum up lots of dense information in a small space. They may take for granted lots of "known" papers. In a book, the authors would have more space to develop their ideas, add explanations that are only a citation in a paper.

It is hard indeed to start reading papers (it's even harder when it's not your mother tongue). I remember when I started that my understanding was very shallow. With time and experience, I'm now capable of understanding also the context, not just the paper itself. At the beginning, I had to look at the used terms and look them up online or read the cited papers/books to understand some of the mathematical basis for the paper I needed.

But it gets better, the curve in a specific domain may be similar close to 1-e^-t. Going from one domain to another gets easier, but there is always a learning curve to understand what everyone in a (new) domain takes for granted.

  • Papers are limited in length more by venue length restrictions (conferences, journals etc.) than by money - as far as I can tell. – einpoklum - reinstate Monica Dec 29 '18 at 21:05
  • Yes, but you can (for papers) pay for more pages. All this limits the amount of explanations you can add. – Matthieu Brucher Dec 29 '18 at 21:12
  • That is only true in some fields and some publications. In my (limited) experience in Computer Science - that's not an option even for journal papers; and I am 100% certain it is not an option for conference papers. The page limit is very strict and money has nothing to do with it. – einpoklum - reinstate Monica Dec 29 '18 at 21:14
  • I had a different experience in Computer Science for journal papers. But I do agree that I should have stated that some have guidelines. This is now fixed. – Matthieu Brucher Dec 29 '18 at 21:17
  • Indeed, papers are limited in length, but some authors write parallel "technical reports" which do not have the same length constraints. – emory Dec 31 '18 at 2:17
2

I agree with @nabla's answer that as you go further and read more and more papers, it gets easier for you and you don't have the difficulties in comprehending the terminologies. It is also very good that you think a lot on each individual paper and want to understand each bit of it. However, remember that you should know how to read a paper. Reading a paper is different from studying a mathematical textbook, which you do in order to prepare yourself for the exam. I suggest you the following steps:

  1. First, consider yourself as a reviewer, who is employed to judge the paper and its novelties and then gives his/her opinion.

  2. Try to scan the paper first. Read the abstract and conclusion and try to understand the framework of the paper.

  3. After reading the abstract you will understand how far you are from the topic. Thus, try to search a little about the terminologies and methods that you are unfamiliar with (by simply googling them and checking Wikipedia).

  4. Read the introduction carefully to see the state of art and the previous studies on the topic. People usually explain things in detail in the introduction, if no they refer to some general and review papers. Try to get those papers and read them.

  5. Now read the whole paper. But remember nobody expects you to understand the mathematical model that the author is presenting in details. Try to have an idea about it. You are not going to write about the detailed mathematics in your report. Don't waste your time on that. If you are personally interested in details then that is another story.

  6. After finishing the paper, ask the following questions from yourself: 6.1. What was the main idea of the paper? 6.2 What was the method that the author implemented? 6.3 how satisfactory were the results? and so on...

  7. Try to write your report according to the above information.

Good Luck!

1

Every person had that time of not comprehending some papers/ academic papers especially if it does have a lot of information. More pages to read we sometimes forget the ideas contained in some pages while reading the concurrent ones. So I think it is very normal if sometimes you did not understand a certain concept because everybody does. What I did if that time happens to me, I read it succeedingly 5 or more times to grasp the information contained in the article and then I can now write my own reflection or output.

protected by Alexandros Dec 31 '18 at 15:02

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

Not the answer you're looking for? Browse other questions tagged or ask your own question.