10

For a paper I am currently writing, I am doing an extensive literature review on a subject which is much more vast than I had anticipated. Are there any specific tricks and tips for organizing ones reading when the body of literature is so large? I tend to get carried away with interesting references and lose the overview a little. If I read everything linearly from A-Z, I would not have the time to finish my review. My goal is thus instead to have a non-exhaustive, but ideally representative impression of the structure of the literature (that is, different strands of research, who 'responded' to whom, etc.).

I would very much appreciate any hints from an experienced academic.

  • I assume you are already using a reference manager (Mendeley, EndNote etc.) This seems to be a problem for a lot of my fellow grad students as well as faculty. I would love to see some good discussion / answers in this thread. Unfortunately I don't have it solved yet.. :) – Matt S Mar 11 '15 at 15:53
  • I'm doing my bibliography management using BiBTeX, which I am very happy with. – Constantin Mar 11 '15 at 15:56
  • I seem to understand that the focus of your question is not tools (which programs to use to organize citations and references) but methods (which papers to read, how to understand which references are important and which are not). Am I correct? – Federico Poloni Mar 15 '15 at 13:02
  • While both types of advice are welcome, you understood correctly that I am looking for methods, rather than just software. – Constantin Mar 15 '15 at 13:55
4
+50

If my opinion, a good survey is not a snapshot of what is out there in the literature, but rather is a presentation of the current state of knowledge on a subject (which will necessarily involve some snapshots of what is out there). This model of knowledge is the novel contribution of the survey. Your goal, then, is to develop enough expertise in the "lay of the land" to be able to see a pattern that may have never been explicitly enunciated in the literature before.

As such, I recommend the following iterative approach, which has been useful for me in preparing surveys:

  1. Begin with a hypothesis of the structure of knowledge on your topic.
  2. With keyword searching, find lots of articles related to the topic.
  3. Do not read the articles: read just enough to make sure they are actually relevant, and download and organize them into categories based on your hypothesis.
  4. When you've got a big enough pile, skim them to see how your hypothesis is holding up, and adjust it to fit the reality of what you've found.
  5. Repeat from 2, with your adjusted hypothesis, gathering more references, until you find your hypothesis is beginning to stabilize. Once you've got a stable hypothesis, you are ready to write, and your final organization will reflect the discovery enunciated in your final hypothesis.

A good survey is probably going to be in the 50 - 200 reference range. If you are finding that you are gathering many more than that, consider whether your subject is too broad and you need to narrow its focus. It may also be that you have an appropriate focus, but that you are thinking about too many details, and you need to go into less depth in your investigation.

2

First: Use a modern reference manager. Mendeley, Zotero or Endnote are all fine. BibTex is wonderful for formatting the citations, and all the previous mentioned exports to BibTex. What BibTex lacks is search functions, storage of the original publications, networking, annotations, easy import and so on. The reference manager can save you significant amounts of work.

Second: Start with the existing literature reviews in the field. If you are publishing the review you will need to position yourself against them in the related work anyway.

Third: The trick to a good literature review is the focus. You want to include all the papers that are important considering the scope of your review, and nothing else. A review which covers "the 200 most cited papers in this broad field" is pretty much useless. If you feel you are getting overwhelmed it's time to start pruning ruthlessly.

  • 2
    What bibtex lacks is search functions: Bibtex databases are plain-text files; there is plenty of tools for searching in them. – Federico Poloni Mar 15 '15 at 12:59
  • 3
    @FedericoPoloni: Which is fine, until you want to find the papers published in Network Communications on MIMO somewhere between 2001 and 2009... In a bibtex database containing several thousands of papers. Sure, you can solve it with grep, regexps and so on, but in my experience features such as full-text search in the original publications and the ability to easily filter search result on different fields makes your life a lot easier. But, as always, YMMV. – pehrs Mar 15 '15 at 13:13
2

Define useful subcategories within the field you are examining. When you are not familiar enough with subject to define applicable categories, then first read a few papers that have been cited a lot.

Now assign each paper to one or more categories, just by reading the abstract and conclusions, and glancing over the rest. Also keep track of how many citations each paper received (you can use Google scholar or Scopus for this).

Within each category you can build a tree by assigning the highly cited papers to knots, and the less cited papers to leaves. Connect them according to how they cite each other. Preferably, connect each leaf to only one knot. Keep in mind that since older papers can never cite the newer ones, it is most efficient to start with the older papers.

Now you have a pretty structured overview, and you can spend some time to read the most important knots in more detail. Finally, write down for each category the most important differences with the other categories, and how the knots and branches are structured.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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