The first task is to choose a topic that is both interesting and not recently reviewed. This requires some judicious use of Google Scholar. I think it’s also useful to have some demonstrable expertise in the area, although for most senior scientists, they are broad enough in their interests and knowledge to get by with a pretty shallow knowledge base. This first task is arguably the most important since it will drive everything else.
Once the topic is chosen, the next step is to conduct a first cursory review of the most important literature. In other words, you actually have to actively read some papers. This first review of the literature is not meant to be exhaustive. That comes later. Rather, its purpose is to prepare the ground for the next step: the review paper equivalent of a news ‘hook’. This is the notion of an insight that comes from a synthesis of the extant relevant literature. So if we were reviewing explainable AI, the hook might be that biological brains might provide both existence proof and likely explanation. If we were reviewing macrosystems approaches to ecology, the hook might be that robust standardization of measurements at continental-scale is required. The key is that the hook should be non-obvious, but emerges from a rigorous synthesis of the literature.
At this stage, the bibliography construction should begin using appropriate software. I use Zotero, but there are many good choices. The key function we are looking for is the ability to seamlessly grab a citation (and the full pdf) from Google Scholar and store it for later insertion into the manuscript document. This is the most important research stage of writing the review paper and it feeds back onto the prior ‘hook’ step since a full reading of the literature may change the results of the synthesis. That’s fine and is perfectly normal for the proper evolution of an excellent review. When this step is complete manuscript writing can commence.
I tend to write from an outline. This is one method for proceeding. It’s not the only one. If one does use the outline approach, the key is to outline the review so that the hook is never standing alone and obvious as the author’s opinion, but rather naturally emerges from the evidence presented in the paper in the mind of the reader. This is tricky. The reason for this subtlety is that review papers are not hypothesis or position papers.Their proper function is to lay out the evidence in such a way that synthesis emerges organically from the evidence. So the hook has to stay in your ‘mind’s eye’ and hidden from the printed page.
My parents, who were scientists themselves, taught me to use the scientifically formal passive voice for writing. This style was echoed during my training. Both my thesis advisor and post-doc mentor employed such a style. Today, we use a more active voice in our writing–not quite as informal as a blog entry, but certainly using more direct and simple language. Whatever the style, the key is to have citations accompany each claim. In review papers, we usually summarize a set of results in a sentence that contains the relevant paper’s first author as the subject and then the summary of what their scientific RESULTS revealed. So an example would be: “Olds et al (2020) found that soil exposed to fixed nitrogen had reduced bacterial diversity compared to controls.”
For a review paper, I complete the figures after I complete the text (I do the opposite for a regular research publication). Figures for a review paper can be adapted from other figures in the literature with written permission and explicit acknowledgement in the figure legend. The key is to adapt the figure so that it is not identical to the original. I choose figures to assist the reader in their synthesis laid out in the review. Once again, the idea is not to overtly lay out your hook, but rather to schematize the results from the literature in such a way that the reader gleans the hidden gem.
Finally, the bibliography can be built automatically by your software and inserted at the end of the document in the proper format. Congrats, you are ready to submit!