To make a short video explaining a scientific paper's method, rewrite the methods section as a sequence of five scenes: what was studied and who got in, how they were allocated, what each group received or what each stage did, what was measured and when, and how the result was analysed. Add one scene for the design choice that makes the method clever or controversial. That is 300 to 450 words, two to three minutes of narration. Draw it as a flow that moves left to right, the way a participant or a data point moves through the study, then check every number, dose and time point against the paper before anyone sees it. A methods video with one wrong number is worse than none.

Why a methods video is a different job

A paper summary covers the question, the finding and the limitations, and gives the method a sentence or two. That is the right balance for most audiences, and it is covered in how to turn a research paper into a video summary.

Sometimes the method is the thing you need to explain. A methods module asks you to critique a design. A journal club is arguing about whether the result can be trusted, which is always a question about the method. Your supervisor wants you to present a protocol you plan to copy. Or the paper's contribution is a pipeline, and the finding only makes sense once you see how data moved through it.

In those cases the method gets the whole video, and the result is one scene at the end, there only so the viewer knows what the machinery produced.

The five questions every method answers

Methods sections are written for replication, so they contain far more than a viewer needs: reagent suppliers, software versions, exclusion criteria running to a page. Strip each one down to five questions.

Then add the sixth scene, which is what turns a description into an explanation: why this design? Every method is a set of choices. Name the one that matters most and say what it protects against.

Worked example: the RECOVERY dexamethasone trial

The RECOVERY trial's dexamethasone result, published in the New England Journal of Medicine (online as a preliminary report in July 2020, in print in February 2021), is a good teaching case because the method is simple, large and has one design choice worth arguing about.

Scene 1, who got in. Patients admitted to hospital in the UK with Covid-19. RECOVERY was a platform trial, testing several treatments at once, and this comparison involved 6,425 patients.

Scene 2, how they were split. Patients were randomly assigned in a 2 to 1 ratio: 4,321 to usual care alone, 2,104 to usual care plus dexamethasone. Randomisation means that, on average, the two groups differ only in the treatment.

Scene 3, what was done. Dexamethasone 6 mg once a day, by mouth or intravenously, for up to 10 days, on top of whatever care the hospital would otherwise give.

Scene 4, what was measured. Death within 28 days of randomisation. One outcome, chosen in advance, that cannot be argued about.

Scene 5, how it was analysed. The proportion who died in each group, compared as an age-adjusted rate ratio. 22.9 percent died in the dexamethasone group and 25.7 percent with usual care, a rate ratio of 0.83 (95 percent confidence interval 0.75 to 0.93).

Scene 6, why this design. The trial was open-label: patients and doctors knew who got dexamethasone. Normally that invites bias. The designers accepted it because the outcome was death, which a doctor's expectations cannot easily change, and because removing blinding made the trial simple enough to run in a very large number of hospitals in the middle of a pandemic. That trade is the thing to discuss.

A seventh scene is optional but useful in a methods class: the benefit was seen in patients receiving oxygen or ventilation at randomisation, not in those with no respiratory support, and the paper reports that split. Pre-specified subgroups like this are a method question in their own right.

Here is the narration for Scene 2 in full:

"Each patient was randomly assigned to one of two groups, two to usual care for every one to dexamethasone. Four thousand three hundred and twenty-one in the first group, two thousand one hundred and four in the second. Because chance decided, not the doctor, any difference in deaths between the groups can be put down to the drug."

That is about 55 words, a little over 20 seconds, and it is the most important scene in the video.

Drawing a method

A method is a process, and processes are drawn as flows. Put the population on the left, the split in the middle and the outcome on the right.

Use the CONSORT shape for trials. Clinical trials report a flow diagram (enrolled, allocated, followed up, analysed) and examiners recognise it. Your drawing can be a simplified version: one box splitting into two arms, each arm ending in a number.

Draw the numbers at the point they apply. 6,425 at the start, 2,104 and 4,321 on the two arms, 22.9 and 25.7 percent at the end. A viewer who loses the narration can still follow the numbers left to right.

Make the design choice visible. For open-label, draw both arms with labelled bottles. For a blinded trial, draw identical unlabelled bottles. The contrast teaches the term better than a definition.

More on turning steps into scenes is in how to explain a process in a video.

When the method is a pipeline

Computer science, data science and many engineering papers have a method that is a pipeline rather than an experiment: data comes in, passes through stages, and something is produced and then evaluated. Say a paper builds a pipeline that takes incident reports from a security team, extracts events, links related ones and drafts a timeline for a responder.

The five questions shift slightly:

1. Input. What data, from where, how much. "2,000 incident tickets from one organisation over three years" (use the paper's real figures).
2. Stages. Each step in order, with what goes in and what comes out. One scene per stage, never more than five stages; group the rest.
3. Output. What a single run produces, shown as one real example.
4. Evaluation. Compared with what baseline, measured how, judged by whom.
5. Design choice. Why this stage order, or why a rule-based step where a learned one would be expected.

The trap with pipelines is drawing boxes labelled with component names nobody knows. Follow one real item through every stage instead: one ticket going in, the events pulled from it, the link to another ticket, the line it becomes in the timeline. A viewer understands the pipeline by watching something move through it.

Make it

1. 1

   Answer the five questions from the methods section

   Who or what, how split, what was done, what was measured and when, how it was analysed. One or two sentences each, with the paper's exact numbers.

2. 2

   Write the design-choice scene

   Name the single choice that matters most, such as open-label, a 2 to 1 ratio, or a fixed stage order, and say what it gains and what it risks.

3. 3

   Paste it into openCanviz and keep your wording

   Choose Keep my wording so every number is narrated exactly as you wrote it. Set two to three minutes and pick whiteboard, so the flow builds left to right.

4. 4

   Check numbers, doses and time points against the paper

   Pause on every scene. Sample sizes, ratios, doses, durations, outcome time points and confidence intervals must match the paper exactly. Fix any label in the editor.

5. 5

   Add the citation to the last scene

   Authors or group name, year, title and journal, so the viewer can find the original.

6. 6

   Test it on someone outside the field

   Ask them to explain back why the groups were split by chance. If they cannot, Scene 2 needs rewriting.

Common questions

How long should a methods video be? Two to three minutes for a single study, which is 300 to 450 spoken words. A methods video longer than four minutes usually means you kept detail that belongs in the paper.

Should I include the results at all? Yes, one scene. A method is easier to follow when the viewer knows what it produced. Keep it to the primary outcome.

What if the method section is unclear or missing details? Say so in the video. "The paper does not report how participants were allocated" is a legitimate and useful methods point, and in a critique it is often the main one.

Can I use the paper's own flow diagram? Redraw it rather than screenshot it. Journal figures are usually copyrighted, and a simplified version drawn in step with the narration is clearer anyway. If you present live, show the original on a slide afterwards.

Is this useful for my own research? Very. Explaining your own planned method in two minutes to your supervisor exposes gaps fast. For a public audience, take the different approach in how to explain your research to the public in a video.