Two tests for a summary worth making
Was it written from memory? If the source was open while you wrote, your hand moved and your understanding didn't have to. This is the same failure as copying lecture slides — it produces a document and no encoding.
Did compression force decisions? A summary that's 60% of the original length required no ranking. One that's 10% required you to decide, for every paragraph, whether it earns its place — and deciding that requires knowing what the chapter is actually claiming.
The method
- 1
Read the whole chapter once, with a pen, marking nothing but structure
Section boundaries and the one sentence per section that carries the claim. No highlighting of interesting facts — you're mapping, not extracting.
- 2
Close the book
This is the step. Everything after this happens without the source open. If you can't proceed, that's information — you didn't understand it, and re-reading is the fix for that specific gap, not for the summary.
- 3
Write the chapter's argument in three sentences
What is it claiming, on what basis, and why does it matter? Three sentences for a forty-page chapter. This is brutal and it's where the understanding shows up.
- 4
Expand to one line per section, still closed
The claim of each section, not the topic. "Enzymes lower activation energy without changing equilibrium" not "enzyme kinetics".
- 5
Now open the book and check — marking gaps, not adding detail
The temptation is to enrich the summary with everything you missed. Don't. Circle the gaps; they become questions, not paragraphs.
- 6
Turn each gap into a question
The gaps are what you didn't retain, which makes them precisely the items worth reviewing. A summary plus a question list is a far better artefact than a longer summary.
Formats, and when each earns its place
| Format | Best for | Fails when |
|---|---|---|
| Three-sentence argument | Anything with a thesis: papers, humanities chapters | Reference material with no argument. |
| One line per section | Textbook chapters with clear structure | The chapter is one long undivided narrative. |
| Question list | Examinable content of any kind | You need the connective narrative, not the facts. |
| Diagram or map | Processes, hierarchies, causal chains | The material is a list rather than a structure — mind maps get overused here. |
| Comparison table | Anything with confusable categories | There's only one thing being described. |
| Worked example you produce | Methods and procedures | Conceptual material with nothing to execute. |
Choosing the format is itself a comprehension act. If you can't tell whether a chapter is an argument, a structure or a procedure, you haven't understood it yet — and that realisation is worth more than the summary.
How short is short enough
Aim for about 5–10% of the original. A forty-page chapter becomes roughly two to four hundred words. That will feel insufficient and the insufficiency is the point: it forces the ranking that makes the exercise worth doing.
- If your summary is over a quarter of the original, you transcribed. No decisions were made.
- Cut examples, keep what they exemplify. The example is how you understood it; the principle is what you need.
- Cut anything you're confident you'll never forget. Summaries are for the forgettable, same logic as a cheat sheet.
- Keep exceptions and conditions. These are what summaries always drop and what exams always ask.
- Keep the causal words. Because, therefore, unless, only when. Bullet lists that strip them lose the content.
The summary is not the deliverable
A summary's value is 80% in the writing and 20% in the document. Which means: don't optimise the document. Don't type it up neatly, don't format it, don't make it shareable. Handwritten, ugly, and generated from memory beats typed, beautiful, and copied every time.
It also means the summary should be re-generated rather than re-read. A week later, write it again from memory and compare. Twenty minutes, and it's a full retrieval of the chapter — considerably more valuable than reading your own summary, which is just rereading with extra steps.
When not to summarise at all
Summarising is expensive and it is not the universal solvent students treat it as. Skip it when the chapter is a reference you'll consult rather than know, when the material is procedural and needs practice rather than compression, and when you're already short on time — in a one-week run-up, questions beat summaries decisively because they can be tested and a summary can only be read.
The default should probably be: summarise the two or three chapters that carry the module's argument, and convert everything else straight into questions. How to remember what you read covers the lighter-weight alternative for the rest.