01Papers are built to be read out of order.
Academic writing is unusually formulaic, and the formula is not an accident. The same sections appear in the same sequence across enormous stretches of the literature precisely so that a reader can enter wherever their question is and leave when it has been answered.
Reading front to back defeats that design. It spends your attention on the literature review, which is written for people who already know the field, before you have established whether the paper is relevant to you at all. Then it arrives at the finding when the attention is gone.
The alternative is not skimming. It is entering at the places where the paper tells you what it is, deciding, and then reading properly if the decision is yes.
02The order that works.
Four minutes for the first four steps, and most papers are settled by the end of them.
- The abstract, for what was asked and what was found. Not for the method, which abstracts describe badly.
- The last paragraph of the introduction, where the specific question or hypothesis is stated. This is more precise than the abstract and it is where the paper says what it is actually about.
- The first paragraph or two of the discussion, where the authors say what they think it means. Read this before the results, because it tells you what to look for in them.
- Then the method, but only the parts governing how far the finding travels: who the participants were, how many, and whether anything was manipulated or only observed.
- And only then the results, and only the parts bearing on your question. Tables reward a reader who already knows what they are looking for and punish one who does not.
03The decision at four minutes.
By this point you can usually answer three things. Whether the paper is about your question or a neighbouring one. Whether the evidence is the kind that supports the claim you want to make. And whether it is recent enough, or foundational enough, to be worth citing.
If any of those is no, stop, and record why in one line so you do not open it again in three weeks having forgotten. That single line is worth more than most of what people write while reading, and it belongs in whatever record you keep, which is its own subject.
If they are yes, then read it properly, and reading it properly now means something different: you already know what the paper claims and roughly how good the evidence is, so the careful read is about detail rather than orientation. That is a much faster hour than an undirected one.
04What to read when the numbers are unfamiliar.
This is where most students stop, and the reasonable-sounding conclusion they draw is that the paper is beyond them. Usually it is not. Quantitative results can be read at a useful level without the technical apparatus.
Look for direction and size rather than significance. Did the thing go up or down, and by how much in the units of the actual world rather than in standardised terms. A great many papers answer that in a sentence in the discussion, written for exactly this reader.
Then look at how many people it happened to. A finding in forty participants and one in four thousand carry very different weight regardless of what the statistics say about either. Those two habits get somebody most of the way through a quantitative literature without ever running a test, and where an assignment genuinely requires the technical reading, that is what the statistics desk is for.
05Reading the reference list on the way out.
Two minutes at the end of any paper you found useful, and it is the most efficient searching available. The reference list is a curated bibliography of the thing you just decided was relevant, assembled by somebody who knows the field.
Note anything cited repeatedly across several papers you have read, because a work everybody cites is usually the one you should have started from. That technique has a name and a method, and it is citation chaining.
It also gives you something a database cannot: a sense of which works the field treats as load-bearing, which is knowledge about the shape of a literature rather than about its contents, and it is what makes a review read as though written by somebody who understands the area.
Questions at this desk.
Is this the same as skimming?
No, and the difference matters. Skimming means reading everything shallowly and finishing with a vague impression. This means reading a few specific sections properly, deciding, then either stopping or reading the whole thing carefully. The reading that happens is deep; only the selection is fast.
How long should a paper take?
Four minutes to decide, and anything from twenty minutes to two hours to read properly once you have decided yes. The variance is enormous and depends on the paper and your purpose, which is exactly why the decision step is worth having: it stops the two-hour treatment being applied indiscriminately.
What if I do not understand the theory it assumes?
Very common, and usually not fatal. Read the introduction's last paragraph and the discussion's first, and see whether the paper's own claim is intelligible without the framework. Frequently it is. Where it genuinely is not, that is a signal to find something more introductory first rather than to persist.
Should I take notes while doing this?
One line at the decision point, always: what the paper claims and whether it is useful to you. Fuller notes belong to the careful read, if there is one. Taking detailed notes during the fast pass defeats the purpose, because it commits your attention before the decision has been made.
Notes in the margin, from the people at the desks.
READ THE LETTER