Skip to content

How to Read a Research Paper

I do not read every paper from the first word of the introduction to the last reference -- and you probably should not either.

Some papers only need five minutes. Some are worth an hour. A few will become key papers that you return to throughout the project. The trick is deciding which is which.

Before you read: know why you opened it

Write down what you want from the paper. For example:

  • I need a baseline method for remote manipulation.
  • I want to know how workload is measured in teleoperation studies.
  • I need to understand why this paper keeps being cited.
  • I want to check whether someone has already tested our idea.

Without a question, it is very easy to highlight half the PDF and retain nothing.

Pass 1: is this paper useful?

Spend about 5–10 minutes on:

  • the title and abstract;
  • the figures and tables;
  • the first and last parts of the introduction;
  • the stated contributions; and
  • the conclusion and limitations.

Try to answer three things:

  1. What question are the authors asking?
  2. What did they actually do?
  3. What do they claim they found?

Then make a decision: stop, save for later, or keep reading.

Stopping is allowed. The first pass has done its job if it prevents you from spending an hour on a paper that is not relevant. I know it's hard, I have severe FOMO too, but you will save time in the long run.

Pass 2: how did they get that result?

If the paper matters, read the method, experimental setup and results more carefully.

Ask:

  • What system, dataset, task or participants did they use?
  • What was compared with what?
  • What assumptions does the method rely on?
  • Which variables and metrics were measured?
  • Are the baselines fair and useful?
  • Does the evidence match the claims?
  • Under which conditions might the result not hold?

Do not panic if one equation or technical detail is unfamiliar. Mark it, work out whether it is central to the paper, and follow it up separately if needed.

Pass 3: take apart the papers that really matter

For a paper central to your project, go deeper:

  • trace the main claims back to figures, tables and analyses;
  • follow important references for methods that were not explained;
  • look for code, datasets and experiment details;
  • sketch the system or experiment in your own way; and
  • compare it directly with another paper or with your planned approach.

A good test is whether you can close the PDF and explain the question, method, evidence and limitation without repeating the abstract.

Keep these three things separate

This distinction is one of the most useful habits you can build:

  1. What the authors claim. Their interpretation of the work.
  2. What the evidence shows. What is supported under the actual task, data, participants and conditions tested.
  3. What you think. Your interpretation and what the result might mean for your project.

For example, a paper may claim that an interface "improves teleoperation". The experiment may only show faster completion on one simulated task with twelve students. Your note should preserve that difference.

Warning

Be careful with sentences beginning with "the paper proves" or "the paper shows". Most studies support a more specific claim under a particular set of conditions.

Put the result into your notes

Use this as a starting template in Obsidian:

---
citekey:
year:
status: read
topics: []
---

# <Paper title>

## Why I opened this

## Research question

## What they did

## What they found

## Important evidence

## Assumptions and limitations

## What the authors claim

## What I think the evidence supports

## Relevance to my project

## Related papers and concepts

## Questions to follow up

Write in your own words. Copying the abstract into Obsidian has mostly moved the abstract from one window to another.

Link the note to relevant concept and project notes, then update the paper's Zotero tag from to-read to read.

You are done when the paper has somewhere to go

Reading is finished when the useful part of the paper has entered your system:

  • the reference is clean in Zotero;
  • the useful evidence is recorded in your notes;
  • the paper is connected to other ideas or papers; and
  • you know whether it changes a project question, decision or next step.

Once several papers are in that state, you can start turning the notes into a literature review.