How to Write a Scientific Review (And The Upstream Work That Makes It Easier)

Researchers who have written review articles will tell you that the hardest part isn’t writing per se, but finding the inspiration to begin, deciding which papers to include, and building a coherent story.

This article is about where to focus your effort when creating a narrative or conceptual scientific review so that you don’t get stuck. I’ll explore how to overcome the two biggest bottlenecks in this workflow: article selection and hypothesis framing.

Critical Reframe: A Review is a Story, Not a List of Literature

A narrative scientific review paper is a structured critical synthesis of existing research. It explains what is known, what is unknown, and where the field is headed. Contrary to popular belief, it is not a summary of papers or a list of literature.

A useful way to think about a scientific review is like a story with a:

• Protagonist: a biological process, a protein of interest, a core concept
• Conflict: contradictory data, unresolved questions, competing schools of thought
• Resolution: where the field should go next, even if that resolution is naming the open question clearly

This means that the papers you select and the hypotheses you propose are narrative decisions, and preparing that narrative is critical.

Bottleneck One: Choosing The Papers

The temptation when told to write a review is to find as many relevant papers as possible and start reading. The problem is that there is too much literature, and prioritizing everything can create confusion. To inform a cohesive narrative for your review, you should first decide which papers carry the most weight.

1. Gather the relevant literature

Start by reading a few recent reviews on your topic. This will give you an idea of which questions the authors left open and what has changed in the last two or three years. If no recent review exists, two or three recent research articles can serve the same purpose.

Once you have read those key reviews, you can draft an outline for your own review, which will point you toward the next papers worth reading. To pick the best core papers, apply the following three filters:

1. Recency: The literature moves fast in many fields, and a review built primarily on older papers may already be out of date. How much this matters depends on your discipline, as some areas will move more slowly than others.
2. Diversity: Papers from different research groups and using varied approaches offer a broader view of the field than three papers authored by the same lab.
3. Influence: As an exception to the first point, consider work that has shaped how the field thinks, regardless of when it was published. An older paper could still be included if it describes seminal research. In some fields, seminal mechanism papers outweigh recent incremental work.

2. Rank the literature

In my experience, an essential step that review writers often skip is ranking the papers before reading them.

To do this, number the articles 1, 2, 3, and so on, weighting by a combination of relevance to your scope, methodological quality, mechanistic centrality, and how much the field has built on them.

Then start reading at the top of the list and build the structure of your review from a small set of high-weight papers before looking at the rest.

If you are struggling to find your creative spark, remember that inspiration often comes from the reading process itself. You don’t need a grand idea to begin. If you aren’t sure what to write in your notes, start by answering a few simple questions for your top papers: Why is this paper important? What is novel here? What do the authors suggest in the discussion? Note down the key results, techniques, and ideas. This structured note-taking will help you find your angle.

Your paper list will almost certainly grow during this phase, as gaps in the narrative force targeted reading. That is totally normal, so don’t panic!

It’s important to note that this can be a chaotic process that looks worse before it gets better. But the messiness is a necessary part of synthesizing your ideas. Researchers who panic at this point sometimes restart from scratch, but don’t do this! Let the outline evolve and work through the initial chaos.

A helpful note on organized reading and notetaking

The notes you take while reading are what you’ll be working from when you write. If they’re disorganized, the writing will stall because you can’t find what you need.

Use these four rules to ensure your notes are legible and comprehensive:

1. Write notes in simple language. You are not writing the manuscript at this stage. Instead, your job is to capture what each paper says in a form you can recombine later. Beautiful sentences are wasted effort here.
2. Write the citation next to every note as you make it. After twenty papers, you are likely to lose track of which note came from which source. It happens faster than most people expect. Including a citation with every note is non-negotiable for this reason.
3. Mark any verbatim text visibly, in red quotes or brackets. If you copy a sentence directly while taking notes, you need to remember weeks later that it is not yours. Otherwise, it risks migrating into the manuscript and becoming accidental plagiarism!
4. It may go without saying, but only cite papers you have actually read. Scientific integrity depends on this, and so does your ability to defend your review.

For tactical advice on locating papers (e.g., Google Scholar tips, reference managers, requesting full texts), see Sergey Pustylnikov’s article on gathering information for a scientific review.

Bottleneck Two: The Hypothesis*

*Note that a hypothesis in a review article is very different from a testable hypothesis in a primary research paper. Here, your hypothesis is your opportunity to add your own voice, interpretations, and insights to the field, which is exactly what journals look for when publishing – analytical, rather than purely descriptive reviews.

A strong review should make a clear, evidence-constrained argument that individual papers cannot. There are two common errors made when proposing a hypothesis: either you freeze and write a voiceless summary because you don’t want to risk being wrong, or you overreach and make claims that the data do not support.

Here are two reliable hypothesis structures you can use to avoid both:

1. The synthetic hypothesis: You have study A and study B, which appear to be about separate things. You hypothesize a connection between them; for example, a shared drug mechanism, a common intracellular pathway, or a cause that links the two observations. This hypothesis is well-grounded because each piece of supporting evidence is already in the literature, and you are proposing the connection between them, not the underlying findings.
2. The gap hypothesis: You identify a gap that the literature has not yet filled, and you predict what would be found if it were filled. This hypothesis is well-grounded because you are explicit that the gap is unfilled: you aren’t claiming the prediction is true, you are claiming it is what the field should test next.

Note: I use the labels “synthetic” and “gap” hypothesis to describe two patterns that recur in narrative reviews; they aren’t formal categories in the publishing literature, but they capture recognizable structures.

Three additional traps to avoid when forming your hypothesis

1. Cherry-picking: Only including studies that support your model and ignoring studies that contradict it. Easy to slip into when you have a hypothesis you like. Counter this by making sure you include contradicting studies and explain what you think the contradiction means.
2. Overgeneralization: Taking results from a specific niche (like in vitro work, a single mouse model, or one cell line) and extending them to broader claims about populations or systems where they have not been tested. Reviewers spot this immediately. The discipline is to state what kind of evidence supports the claim and which system it has been tested in.
3. Overconfident language: Use qualified words like suggest, indicate, may, and appear to. Avoid using “definitely” and “always” unless the data genuinely support them.

Two more tips to strengthen your hypothesis

1. Acknowledge the limitations of your own hypothesis: Many authors think that doing this weakens their review. However, it demonstrates scientific rigor and pre-empts the reviewer who would otherwise raise the limitation as a flaw.
2. Include outliers: Studies that don’t fit your theory should be mentioned. Explain why they don’t fit, e.g., different cell types, different methodology, different conditions. A review that handles contradictory data explicitly is far more defensible than one that ignores it.

And Now Writing Becomes More Tractable

By now, you will have decided what you think, and the writing process will be significantly more manageable. If paper selection is solid and the hypothesis is set, four rules are enough to shape this stage:

A Note on Figures

Many skimming readers will look at your figures before they decide whether to read the text. Note that figures are not decorative, but part of how the review builds its argument. Once you have a rough first draft and understand your logical flow, start designing your figures. You can then refine your text so that the words and visuals tell a cohesive story.

Here are three principles to handle the figure work in a review:

A Note on Titles, Abstracts, Reviewers, and AI

These final steps are practical rather than structural. Here are a few brief notes on each to add the finishing touches to your review:

• Title and abstract: Simple, precise titles with the keywords readers will search for tend to perform better than clever ones. For the abstract, four elements earn their place: what the review is about, why the topic matters, what the review explores, and why the knowledge matters to the field.
• Reviewer comments: Reviewers are not your enemies (not even the infamous reviewer 2). If a reviewer asks about something outside your expertise, treat it as an opportunity. Read enough to answer the question and let the response strengthen your narrative.
• AI tools: AI can assist with brainstorming and grammar editing, particularly for non-native English speakers. But use them with caution for literature search: AI can produce hallucinated citations to papers that do not exist and may return incomplete or biased results. Verify every citation against the source before it enters your manuscript.

Where to put the effort

For most researchers, the review-writing process fails before the pen meets paper. Paper selection is where most of your effort should go, followed by hypothesis framing. Once those are done properly, the prose becomes more manageable.

If you are stalled on a review, the problem is most likely in one of three places: paper weighting, hypothesis framing, or structure. Start by ranking your papers, but don’t spend too much time worrying about the perfect list. Just start reading the strongest few. If you are struggling with how to begin, remember that the ideas and inspiration will follow once you are reading, letting the narrative emerge from there.

To explore more of my medical writing projects, please visit my online portfolio.

Gorana Jendrisek

Gorana completed her PhD in Molecular Oncology in 2022. Currently, she is a postdoctoral researcher at the European Institute of Oncology (IEO) in Italy. Gorana specializes in synthesizing complex oncology research into compelling narratives, end-to-end manuscript management, and science communication.

Share this article: