How to Read a Scientific Paper

1. What they did

This is a short guide, written by two University of Arizona scientists for a biochemistry and molecular biology course, on how to read and judge a scientific paper. It is not a research paper. It answers four questions: how papers are organised, how to prepare to read one (especially in an unfamiliar field), what difficulties to expect, and how to understand and evaluate what a paper contains.

2. What they found

How papers are organised

• Most papers follow a standard order. The Abstract gives a short background, the main findings and how they fit the field. The Introduction sets out what was already known and the question being asked. Materials and Methods should, in principle, let others repeat the work, but are often compressed. Results describe the experiments and why they were done. The Discussion interprets the data, should admit its limits, should separate fact from speculation, and relates the findings to other work.
• Journals with tight space limits, such as Science and Nature, have no separate sections, so authors squeeze details into endnotes, figure legends and online supplements, and often leave out the logical links that make a paper easy to follow.

How to prepare

• Read the title and Abstract first, then pause and recall what you already know about the topic. This shows whether you need more background, refreshes your memory, and helps you fit the new information into what you know.
• Readers who know the field can skim the Introduction, read straight through to the Results, and turn to the Methods only when needed.
• Watch for “codewords”. “Data not shown” is usually for reasons of space. “Unpublished data” can mean the data are not good enough to publish or are part of a larger story. “Preliminary data” can mean the experiment was done only once.

Difficulties

• Some problems lie with the reader, but many are the writer’s fault. Many papers are poorly written: the reasons for an experiment are left out, there is a lot of jargon, and side issues get as much space as the main thread.
• Methods are often hard to pin down, because papers point back to earlier papers in a long chain.
• Authors may be uncritical about their favourite explanation, may not separate fact from speculation, and, wanting to publish in fashionable journals, may overstate their findings or put a speculation in the title as if it were settled.

Six questions for evaluating a paper

• What questions does it address? Research can be descriptive (what is there?), comparative (is this general?) or analytical (how does it work?), and many papers combine all three.
• What are its main conclusions? The Abstract is a starting point, but not enough.
• What evidence supports them? Look for how the experiments in the Results fit together.
• Do the data really support the conclusions? The logic may be unsound, or other explanations may fit the data. Several independent lines of evidence are more persuasive than one. Look for hidden assumptions.
• What is the quality of the evidence? This is the hardest question. Understand the methods and their limits, separate what the data show from what the authors say they show, look at the original figures rather than photocopies, and ask where the controls are.
• Why are the conclusions important? Do they advance knowledge or open new directions?

3. What they concluded

• Reading a paper well means reading it actively and critically rather than accepting it because it appeared in a top journal. Learning to judge evidence is one of the hardest and most important skills for a young scientist, and it takes years.
• Understanding the methods, and their limits, is the key to judging whether the data show what the authors say they show.