Reading a paper

Why this matters

Research papers are how physical scientists share results. In a research group you might be asked to read several a week; in industry you read them to choose a method or check a competitor's claim. Papers are dense and written for experts, so reading one front to back on the first try is slow and discouraging. A strategy makes it manageable.

How a paper is organised

Most experimental and computational papers follow a structure often called IMRaDThe common paper structure of Introduction, Methods, Results and Discussion.:

Physics papers sometimes merge sections, and letters are shorter, but the same questions are always answered somewhere.

The three-pass method

A widely used approach, described by computer scientist S. Keshav in "How to Read a Paper", reads in three passes, each deeper than the last.

Pass 1: the survey (5 to 10 minutes). Read the title, abstract, section headings, figures and captions, and the conclusion. Skip the equations. Afterwards you should be able to say what kind of paper it is, what question it answers and what it claims. Often this is enough to decide whether to go further.

Pass 2: the content (about an hour). Read the whole paper, but skip proofs and fine details. Study each figure carefully: what are the axes and units, what are the error bars, does the figure actually support the claim in the text? Mark terms you do not know and references you might need.

Pass 3: the deep read (several hours). Rarely needed as a beginner. Here you try to reproduce the reasoning, check assumptions and think about how you would do the work yourself. This is what reviewers and researchers who build directly on a paper do.

Worked example: summarising a materials paper

Suppose you read a paper titled "Doping increases the ionic conductivity of a solid electrolyte for lithium batteries". A good summary looks like this:

• Question: Does adding a small amount of element X to the solid electrolyte increase how fast lithium ions move through it?
• Gap: Solid electrolytes are safer than liquids but conduct ions too slowly; previous doping attempts gave small improvements.
• Method: Samples with 0, 1, 2 and 5% doping were made; conductivity was measured by impedance spectroscopy from 25 to 100 °C; crystal structure was checked by X-ray diffraction; three samples per composition.
• Key figure: conductivity (S/cm, logarithmic axis) against temperature for each composition.
• Main result: 2% doping gave about ten times the conductivity of the undoped sample at room temperature, with sample-to-sample variation of about 20%.
• Limitation: long-term stability in a working battery was not tested.

That summary fits on an index card and is what your supervisor will usually want.

Reading figures critically

Figures carry most of a paper's evidence. For each one, ask:

1. What is plotted, in what units, and on what scale (linear or logarithmic)?
2. How many samples or runs does each point represent?
3. Are there error bars, and what do they represent?
4. Does the text claim more than the figure shows?

Peer review and preprints

Journal papers have been through peer reviewChecking of a paper by experts before publication, which reduces errors but does not guarantee correctness.: experts checked them before publication. That reduces errors but does not guarantee correctness. PreprintsPapers posted publicly, for example on arXiv, that have not necessarily been peer reviewed., such as those on arXiv, have not necessarily been peer reviewed. They are valuable for speed, but read them with extra care. The next lesson covers finding both.

Common mistakes

• Reading every word in order on the first pass. Start with the abstract, figures and conclusion.
• Getting stuck on one equation. Mark it and move on; come back if you need it.
• Trusting the abstract alone. Abstracts present the best version of the result; check the figures.
• Not writing anything down. A short structured summary is what makes the reading useful later.

Quick recap

• Papers follow a predictable structure: abstract, introduction, methods, results, discussion.
• Read in passes: survey, content, then a deep read only when needed.
• Summarise as question, gap, method, key figure, result with a number, and limitation.

Try it

Open the Do tab and summarise an open-access paper using the three-pass methodReading a paper in three increasingly deep passes: survey, content, then a deep read when needed..