As a first year PhD student, it is not an uncommon thing to be asked to write a review paper on your subject area. It is both a great way to get acquainted with your research field and to get the background portion of your thesis completed early. However, it can seem like a daunting task to go from knowing almost nothing about your research field to producing something of interest for experts who have spent years studying your subject matter.
In my first year, I was exactly in this position and I found very little online to help guide this process. Thus, here is my reflective look at writing a review paper that will hopefully help someone else in the future.
Collecting references
The first stage to this is collecting material to go in the review. Most likely, when you started your PhD your supervisor sent you some papers to get ideas about the project- start here. Go through these papers, get a gist of what they were doing, look at what other works they cited and start collecting a list of references in this way. Use tools and search engines like google scholar and PubMed to search for keywords relating to your topic area and add the relevant references to your list. If your field is moving quickly and the frequency of publications is high, consider setting up research alerts using things like Web of Science and BioRxiv (or whatever the relevant pre-print server is in your field). These alerts will help you collect new papers as they come out and help keep you on top of the literature.
At this stage, I wouldn’t worry about reading the papers in too much detail, you will get bogged down in the specifics and feel a bit overwhelmed. Think of this stage as creating a chores list that will be refined later.
Use a reference manager like Zotero, Mendeley, or EndNote (or whatever software you prefer). These tools will help you generate citations lists, collect notes, organize topics, and overall be really useful throughout the process.
Reading
A common pitfall of this process is trying to read every paper in full detail. Your final review will most likely contain around a hundred references so there is no way to possibly grasp every detail of all of those. For that reason, I like to have different levels to reading a paper.
- Level 1: Title – Abstract. I read most things at this level to see if they are relevant to the review and if I should add them to the collection.
- Level 2: Last paragraph of introduction – first paragraph of discussion – titles of results. Here is where I see what the author thinks their paper is about. It is useful to get a feel for what the paper looks at and what categories it fits in.
- Level 3: discussion – results – conclusion. Here I get a look at the paper actually showed and the conclusions taken from the work. I will read the key papers of the review in this detail once they have been identified.
- Level 4: methods – rest of the introduction. The final tier is reserved for a select few papers where it is very relevant to the review and will most likely shape the themes discussed. Often the introductions are very similar across a field and there is little information to be extracted. The methods are often very detailed and useful only if you are considering taking similar approaches in your own work.
Sizing up what level to read a paper at is quite an art. Maybe counterintuitively, I would recommend trying to do the minimum level of reading as much as possible. You can always go back and read a paper in more detail when flushing out your arguments in your review, but you can’t recoup the lost time spent going through the methods of a paper in perfect detail.
At whatever level you are reading, it is important to understand what you are reading. If there are terms or techniques you don’t know, look them up! Often these will be universal across a field (think ROC-AUC metrics in machine learning) and will help speed up reading in the future when you come across the same data comparisons.
Organizing
Now that the reference list is growing, it is time to start shaping it into an interesting story. I like the analogy of a sculpter chiseling away at a chunk of wood. The collection stage is growing the tree and ensuring that the piece of wood is large enough, now it comes to draw the outline of what we want the work to look like. To do this, I like to come up with themes from what I have seen in the reading list. In my field it might look like, papers relating to data collection and curation and models that use this data to make predictions, and applications of these models on a particular use case. Then maybe these broad themes can be sub-dived by other things and so on.
At this stage, it is worth putting these references into a document and start coming up with titles for each theme. To me, the hardest part of this whole process is trying to create a coherent story from many different papers that have not been written by the same group or person. Each one will have their own perspective and methods so it is not always an easy task to connect them with a common thread. Some creativity can be required. Trying to do “pseudo-benchmarks” of many works might not be possible if the methods are evaluated in different ways so think more in terms of high level ideas and use tables and figures to illustrate these connections.
Writing, reviewing and editing
Everyone likes to write differently and their are many other resources online to get advice on this part of the process so I won’t discuss it too much here. My word of advice is to have some way of versioning your work so that you aren’t afraid to make mistakes, change things, and restructure. When writing my own review, after coming up with a first draft, I completely restructured the flow of the paper to make it more interesting. Don’t be afraid to do similar things in your work.
Other tips
This process is not perfectly linear, you will still be collecting new references as you are writing, reading papers again and again etc. The pipeline described here is just to help move the process along and not make any one-step feel too overwhelming. Happy writing!
