How to Read a Scientific Paper: A Beginner's Walk-Through

Do not read a scientific paper from top to bottom. Read it in an order that matches what you are trying to find out: start with the question the authors asked, skim the abstract, then go to the methods to find out who or what was actually studied, then the figures and tables, and only then the authors' own interpretation. A paper is a tool for checking a claim, not a story to be read in sequence.

This article is a beginner's walk-through for someone who has followed a link from an article to the original study and found a dense page of unfamiliar terms. It explains the parts of a paper, a sensible reading order, and the few places where the most important facts hide. It describes where things are, and it does not discuss what any particular study found about any particular substance.

The parts of a paper and what each is for

Think of a house viewing. The abstract is the estate agent's listing: a short, polished summary. The methods are the building plans, showing exactly how the thing was put together. The results are the rooms themselves, what you can actually see and walk around. And the discussion is the owner telling you what they think of the place. All four are useful, but they are written by different parts of the same person, and the listing and the owner's opinion are the ones most inclined to flatter the property.

Most biomedical papers follow this layout, sometimes described by its initials: introduction, methods, results and discussion. Knowing the layout means you always know where to look for a given fact, which is half the battle.

Start with the question, not the abstract

The most helpful single habit is to work out what the paper is about before reading it closely. Experienced readers do this by looking at the title and the last paragraph of the introduction, where the authors usually state their question or aim in a sentence 12.

Try to put the question in plain words, in the form: 'they wanted to find out whether X changes Y in Z.' X is what they did, Y is what they measured, and Z is the thing it was done to, such as cells in a dish, mice, or a group of people. If you can fill in all three blanks, you already understand the paper's skeleton. If you cannot, the paper may be badly written, or it may be doing something other than you assumed.

Then read the abstract, but treat it as the listing it is. It is a summary that the authors chose to write, and it often leads with the most striking finding. It is a good map, and a poor substitute for the territory. Many wrong impressions of research begin with someone who read only the abstract.

Methods: who or what was studied, and how many

The methods section is the dullest and the most important. It is where the answers to the questions that decide how much weight a study deserves are kept. Go through it with a short list in mind.

• Species or population: cells, a named animal strain, or people. This one fact tells you what the findings can and cannot speak to.
• Sample size: how many cells, animals or participants were in each group. Small groups are easily misled by chance.
• Groups and comparison: was there a control group, and how were individuals assigned? Random assignment is stronger than self-selection.
• Blinding: did the people measuring the outcome know which group was which? If they did, expectation can leak in.
• What was measured: the outcome the study actually tracked, and whether it is the real thing of interest or a stand-in for it.
• What was given and how: the substance, its source and its form as described, so a reader could see what was really tested.

For animal studies there is a useful yardstick. A set of reporting guidelines called ARRIVE lists the details that a good animal paper should state, such as how many animals were used, how they were assigned to groups, whether the assessors were blinded, and how the results were analysed 34. You do not need to memorise it. It is enough to know that a paper which says nothing about these things is telling you something.

Results: figures, tables and the numbers that matter

Resist the urge to read the results as running text. Go straight to the figures and tables, and try to see what they say before reading the authors' description of them. A graph usually has a vertical axis, showing what was measured, and a horizontal axis, showing the groups or time. Check what the vertical axis starts at and what units it uses. A dramatic-looking difference can be a small one on a stretched scale.

Look at the spread, not only the average. Many graphs show a bar or a point for the group's average, with thin lines called error bars to show how much the individual values varied. If the bars for two groups overlap heavily, the groups may not truly differ. Look also for the number of individuals behind each bar. An impressive average from five animals is a different thing from the same average from five hundred people.

You will often meet the phrase 'statistically significant', and a figure called a p-value next to it. In brief, this is a measure of how surprising a result would be if nothing were really going on. It is not a measure of how large or important the result is, and a very small effect can be statistically significant in a big enough study. It is worth reading the dedicated explanation of that phrase if you meet it often.

Discussion and limitations: the authors' own caveats

By the time you reach the discussion you should already have your own impression of what the results show. Now see how the authors compare. A good discussion states what was found, relates it to earlier work, and, importantly, admits its limits. Look for the paragraph usually headed 'limitations'. It is often brief and sometimes buried, and it frequently contains the most candid sentences in the paper: that the sample was small, that the study could not show cause, that the findings need confirming.

Watch for the gap between the results and the language used to describe them. A study in cells or animals often ends with phrases like 'may be useful for' or 'suggests a potential role in'. These are hypotheses, not findings. If the language in the discussion, or in a news story about the paper, is stronger than the evidence in the results, the stretch is in the telling, not in the data.

Funding, conflicts and the fine print

Near the end of most papers, often in small print, are statements about who paid for the work and what interests the authors have. Funding from a public research body, a charity or a company is all normal. The point is not to dismiss any of them, but to know about them. Where the funder or an author stands to gain from a particular answer, it is reasonable to look a little harder at how the study was designed and reported.

Also check where the paper was published and whether it has been through peer review, the process in which independent experts check a manuscript before publication. A preprint, an early version posted before that checking, has not yet been through it. Neither fact makes a paper right or wrong, but both belong in your sense of how much weight to give it.

A worked walk-through of a typical animal study

Here is how the structure of a typical rodent study reads in practice, with an invented, generic layout so that it describes no real substance or result. The point is where each fact lives.

Notice that nothing in that table requires you to understand the chemistry. You can judge a great deal about a paper by finding these six facts. That is what careful readers do: they interrogate the structure first, and only then wrestle with the specialist detail, often skipping what they cannot follow 2.

In one sentence

Read a paper like a house viewing: find the question, treat the abstract as a listing, inspect the plans and the rooms in the methods and figures, and take the owner's opinion in the discussion with a pinch of salt, remembering that what you are checking is how well the evidence supports the claim, not whether the prose is persuasive.