Guide to Lit Searching for the Disinclined

By Chad Topaz, Williams College and May Mei, Denison University

Summer is upon us, which means it’s time for sunglasses, sandcastles, and student research. Every summer, we find ourselves explaining to students how to do a literature review. Prompted by interest at a recent applied mathematics faculty workshop sponsored by the Alliance to Advance Liberal Arts Colleges (AALAC), we finally put into writing some recommended practices. We hope they will be helpful to you and your students.

Step 1: Recognize the value of the task. Even though lit searches seem annoying — oh wait, they ARE annoying — it’s crucial to do them. We have seen students enthusiastic about publishing skip doing a thorough lit search and spend a summer or a year working on a project only to learn at the last minute that it has all been done before. Even more substantively, when you do a piece of scientific work, it’s crucial to contextualize it. How does it fit in with other work that has been done? Think of your future work as being in dialogue with all past work. Which pieces are you talking to? What are you saying? Are you answering a question that was posed by another paper? Why should anyone care? You can’t answer these questions without knowing the literature.

Step 2: Get organized. Get access to bibliographic management software. We favor cheap or free software. In math, many of us write papers in LaTeX, so apps like BibDesk interface well. JabRef is also another popular option. These apps provide a convenient WYSIWYG interface for bibliographic entries. Even better, many of them have the capability to process data for you. For instance, you can open a web page with a paper from within the app and the app will figure out how to extract the bibliographic reference. Or you can drag a web clipping onto the app and it will, again, extract the reference. Whatever app you choose, if it is decent, it will have slick features that exempt you from having to type in the bibliographic info yourself. Figure out how to use these features. In addition to your software, dedicate a folder on your hard drive to saving PDFs of papers you will read.

Step 3: Search for the first time and save files. While there many tools for conducting literature searches, we recommend Google Scholar because it is free and easily accessible. Use Google Scholar to search for keywords relevant to your project. You may wish to create a text file to keep track your search phrases. For each paper you find, if the title makes the paper seem obviously irrelevant, ignore it. If the title seems relevant, find a way to read the paper. Google Scholar might give you a link to a .pdf of it. If it doesn’t, find another way to get the paper. You could try just Googling the title, or you could use your institution’s online library capabilities, or request the paper via Interlibrary Loan, or ask a colleague for the paper, or ask one of the authors for the paper… or one of many other options. If you decide to get a paper, you should immediately add a bibliographic entry for it, as described above. Most of the bibliographic software will allow you to add a citation key for the paper, that is, the key you would use to cite the paper in LaTeX using a \cite{} command. Use the following convention for the cite key: the first three letters of the last names of (up to) the first three authors, followed by the year of publication. For instance, if the paper is by Cortez, Smith, and Johnson and was published in 2005, you’d give it the cite key CorSmiJoh2005. If the paper is by Lee and was published in 1994, you’d give it the cite key Lee1994. Then, save the paper to your designated hard drive folder and give it a name that is the citation key, with a .pdf extension. For the first example above, you’d call the file CorSmiJoh2005.pdf. The reason for using this convention is that we tend to know authors and years of papers in our heads (if we’ve read them enough). With our cite key convention, it becomes easy to search for a paper on one’s hard drive, and easy to cite the paper — both off the top of one’s head. In contrast, titles of related math papers tend to involve many of the same words, which can make it difficult to distinguish papers by title.

Step 4: Read for the first time. Once you have a collection of .pdfs on your hard drive with corresponding bibliographic entries, take a look at each one. Read the abstract and introduction carefully. If it still seems relevant, take a few notes into a document — perhaps a GoogleDoc or a LaTeX document. You might want to focus on the type of work done (experimental, analytical, numerical, model creation, etc.), what the main results are (at a very high level), and why the work is relevant to your project.

Step 5: Do a backwards and forwards search. For any paper you read that seems relevant, take a look at the references in the paper. You will need to evaluate each of these. If it seems relevant to your literature search based on the title and its context in the paper, then get the .pdf, put it in your bibliography, and apply Step 4. This is called a backwards reference search. Do we really mean that for all of your papers in Step 3, you need to read their bibliographies and consider each and every cited paper? Yes. Yes we do mean that. Then, again, for each of your papers in Step 3, look up that paper in Google Scholar and click on the “cited by” link to see papers that cite that paper. This is called a forwards reference search. In some cases, there may be too many papers appearing in the forwards search for you to process them all. Focus on highly cited papers and on very recent papers in the forwards reference search. For any of these results that seem relevant, apply Step 4 again.

Step 6: Do it all again… we are not even joking. From Step 5, you added a bunch of new papers to your database. You need to do a forward and backwards search on each of these. For realzies. Keep repeating this step until you cease getting relevant new papers. You might start finding the same papers over and over again, and if this is the case (and if everything else you are finding seems not relevant) then you are done.

This seems like a lot of work. It is. But it’s worth it. It will save you a lot of pain at the end of your project if you invest the time up front. Most importantly, you will learn a huge amount from doing it.

This entry was posted in career advancement, General, Graduate School, Journals and Publications, Uncategorized. Bookmark the permalink.

3 Responses to Guide to Lit Searching for the Disinclined

  1. Edward Dunne says:

    Most of us were taught the importance of literature searches in high school and as undergraduates, but it is particularly important when doing research that you plan to publish.

    Chad and May mention Google Scholar. I would recommend using the AMS’s powerful research tool, MathSciNet, which has the advantage of being specialized for the research literature in mathematics and the mathematical sciences. BibDesk has the ability to use MathSciNet built right into it, or you can use it directly. MathSciNet has the benefit of being curated by the expert editors at Mathematical Reviews. (I’m the Executive Editor, so I might be biased.) Moreover, MathSciNet includes reviews of many of the papers, which can help you decide the relevance to your own work. It has two types of citations, which help you with Step 5 above. The first type is citations from reference lists, which are the sort discussed in the blog post. The second type is citations from reviews, where a third party – the reviewer – points out a link between this paper and another. With the size of the literature growing exponentially (measured by number of articles published per year), you need all the help you can get when trying to navigate it.

    MathSciNet is a subscription service, so your college or university has to subscribe. Most libraries subscribe through consortia, which can bring the cost down dramatically, particularly for four-year colleges.

    I have an AMS blog about MathSciNet called “Beyond Reviews”, where you can find out more about the tool. Somewhere on the right-hand side of the page you are reading, there is a link to it.

    PS: For an old-school take on literature searches, I recommend Umberto Eco’s “How to Write a Thesis”. Most of the specifics are woefully out of date, but the basis process is the same and the philosophy of why you do it remains absolutely true today. Finally, Eco was a brilliant writer – it is enjoyable to read just about anything he writes.

  2. Chad Topaz says:

    Edward, thanks for your comment. I agree that MathSciNet is a wonderful tool with rich features. I do think having tools that are absolutely free is important for the community — hence my mention of GoogleScholar. I am also curious about to what degree MathSciNet covers applied math. Applied mathematicians publish in many venues, some of which might be classified as math journals, but many of which are not. In my own case, prompted by your comment, I just did an experiment and looked myself up. I found that some of my publications are there, and some not (for instance, publications in the Phys. Rev. journals and in PLOS One appear not to be indexed there). In any case, please don’t misunderstand me. MathSciNet is great!

  3. Edward Dunne says:

    MathSciNet has good coverage of applied mathematics. Looking at the data on what we added to the database in 2016, and separating by primary MSCs, more than half the items were in applied mathematics or applications: 42% of the items were in pure math, 19% in physical sciences and engineering, 18% were in computer science, numerical analysis, and information science, 10% were in social sciences, biology, and economics, 9% were in statistics and probability, and 2% were in other areas.

    Phys. Rev. A, B, C, D, E, and Phys. Rev. Lett. are all covered in MathSciNet. PLoS Computational Bio is covered, but not PLoS One.

Leave a Reply