Most summaries fail in the same way: they are the original text with words removed. Every paragraph gets a shorter paragraph, the order is preserved, the emphasis is preserved, and nothing has been understood. It reads fine and it is worthless, because the work of deciding what mattered was never done.
A real summary is a different document with the same content. It reorganises, it takes a position on what is central, and it drops things the original spent pages on. If your summary has the same shape as the source, you have abridged rather than summarised.
The one-sentence test
Before writing anything, force the whole thing into a single sentence: what is this actually saying?
This is uncomfortable and it is the entire exercise in miniature. If you cannot do it, you do not yet understand the material well enough to summarise it, and more notes will not fix that. Go back and reread with that one question in mind.
Once you have the sentence, the rest of the summary is support for it. Anything that does not support, qualify or contradict that sentence is a candidate for deletion, no matter how interesting it was.
The four moves of compression
Brown and Day's classic work on summarising describes a small set of operations that skilled summarisers perform and novices do not. They are worth knowing explicitly, because they are learnable:
- Delete the trivial and the redundant. Most texts repeat their main point several times in different clothes. Keep one.
- Replace lists with the category. "Apples, pears, plums and cherries" becomes "fruit". Almost every long paragraph contains a list that wants collapsing.
- Replace a sequence of actions with the action. "She opened the drawer, took out the file, and put it on the desk" becomes "she fetched the file".
- Find or invent the topic sentence. If the paragraph states its own point, use it. If it does not, write the sentence the author should have written. Novices only ever do the first; the fourth is where understanding shows.
The difference between a good and bad summariser is almost entirely how much of the work happens at steps three and four.
Structure first, prose second
Write the skeleton before the sentences. For most material that means:
- The claim. The one sentence, refined.
- The support. Two to five things the argument rests on. Not every point made, only the load-bearing ones.
- The limits. What it does not claim, where it is weak, what the author admits.
- The consequence. What changes if this is true.
That last section is the one people skip and the one that makes a summary worth rereading. A summary with no "so what" is a receipt.
For an argumentative text, add a line for the strongest objection. For a study, add sample and effect size — see how to read a research paper, where the numbers are the summary.
Write it without looking
The single highest-leverage rule: draft the summary from memory, then check the source.
Rereading while you write produces prose that shadows the original, because the original's phrasing is right there and it is easier than thinking. Writing from memory forces reconstruction, which is active recall — the technique with the most consistent evidence behind it in the learning literature. The gaps you discover while drafting are exactly the things you had not actually learned.
Then check, and fix what you got wrong. Do not fix the style to match the source; only fix the errors.
Length is a decision, not a result
Decide the length before you write. "As short as it can be" is not a constraint and produces sprawl.
Useful defaults: one sentence to decide whether to revisit, a paragraph to remember a book in five years, a page to use it in your own writing. A ten-page summary of a two-hundred-page book is not a summary; it is a transcript with ambition.
The constraint is the tool. Halving your word limit is the fastest way to discover what you actually think is important, and it does more for the quality of a summary than any amount of rereading.
Summarising as study: the honest caveat
Summarising is often recommended as a study technique, and the evidence is more mixed than its popularity suggests. The large review by Dunlosky and colleagues rated summarisation as low utility overall — not because compression is useless, but because most students do it badly, producing the shortened-transcript version that involves no retrieval and no judgement.
Done as described here — from memory, restructured, with the limits and the "so what" — it becomes a form of the generation effect and earns its keep. Done as highlighting-then-rewriting, it is the illusion of competence with extra steps.
Where the summary should live
A summary is only useful if you find it later, which means it needs a title stating the claim rather than the topic. "Deliberate practice requires immediate feedback" beats "Notes on Ericsson", because the first tells you what is inside without opening it. That is the core discipline of atomic notes, and it applies to summaries more than anything else.
Keep the summary and your reaction separate. Six months on, the line between what the author claimed and what you concluded is invisible unless you drew it at the time. See literature notes.
The minimum viable version
One sentence for the claim. Three bullets for the support. One line for what you disagree with. Written from memory, checked afterwards.
Four lines, ten minutes, and it will outperform three pages of shortened original every time.
More in progressive summarization, how to remember what you read, and how to take notes.