Skim Then Answer, or Answer Then Skim?

When text first wins

Text first wins wherever the answers are implied rather than stated, because you cannot search for an implication. A search needs something to search for, and the thing you would search for is not on the page.

Part 1 is the clearest case. A letter writer does not state that they are anxious about a decision. They mention it once, describe it as the part they have been putting off, and move on. Several of the eleven questions in Part 1 turn on that kind of implication. A candidate who goes question first reads a stem like the writer suggests that her decision was, scans the letter for the word decision, finds three of them, and picks the wrong one.

Part 4 is the second case, for a different reason. The article is not hard to understand, but every fact in it belongs to a named person, and attribution is not something you can scan for. You have to have read the article with the owners attached. Going question first in Part 4 means searching for a position you will certainly find, in the mouth of somebody who did not say it.

In both parts the text is short enough to justify the investment: about 90 seconds for a Part 1 letter, about three minutes for a Part 4 article and its ledger of who thinks what.

When question first wins

Question first wins wherever the task is location rather than comprehension, and Part 3 is the purest example in the section. Nobody asks what the passage means. They ask, nine times, which paragraph holds this fact.

Reading four paragraphs closely before you know what the nine statements want is the most expensive mistake available in Part 3. You cannot retain four paragraphs of unlabelled detail, so you read them, learn nothing usable, and then read them again once the statements tell you what mattered.

The correct opening is a hybrid that leans heavily toward question first: spend about a minute getting four topic labels rather than four summaries, then work statement by statement, translating each statement into the plain fact it is asking for before you search. The map is not reading the text, it is indexing it.

Part 2 also leans this way, though with its own shape. Read the chart for structure, not content, then take the constraints from the email, then go into the chart for the specific bullet that settles the blank in front of you. Reading twelve feature bullets closely before you know the person's constraints is the Part 3 mistake in a different costume.

The rule that decides it

The deciding question is not how long the text is or how fast you read. It is this: can the answer be found by looking, or does it have to be understood first?

If the answer is a stated fact sitting in one place, the text is a database and you should query it. Go question first, and do not read anything you were not sent to read. That is Part 3, and it is most of Part 2.

If the answer depends on tone, implication, or who holds a position, the text is an argument and you have to have listened to it. Go text first, read properly once, and accept that some of what you read will never be asked about. That is Part 1 and Part 4.

There is a useful proxy if you are unsure on the day. Look at whether the questions repeat the text's vocabulary. Part 3 statements are deliberately written without the paragraph's words, which tells you that matching words will fail and that you should translate first. Part 1 stems half-quote the letter, which tells you the order rule is live and that reading the letter whole will pay.

The routine that loses in every part

There is a third routine, and almost everybody does it at least once. Read the text quickly, not properly. Then read a question, feel unsure, and go back and read the text again. Then the next question, and back again.

This is the worst of both approaches. You have paid the cost of text first, which is time spent on things nobody asks about, without getting its benefit, which is understanding. And you have paid the cost of question first, which is repeated searching, without getting its benefit, which is that you only read what you need.

It is also self-reinforcing under pressure. Each reread makes you later, each moment of lateness makes the next skim shallower, and the shallower skim guarantees the next reread. Candidates who finish a part with three questions unanswered have usually read the passage four times.

The way out is to commit to one opening move and stop renegotiating it mid-part. Either the text gets a proper read at the start, or it gets no read at all and every visit is a targeted search. Half a read is worse than neither.

Building the habit before test day

Decide your opening move for each of the four parts now, in writing, and rehearse it until it is automatic. Text first for Part 1 and Part 4. Structure-then-question for Part 2 and Part 3. Four decisions, made once, so that none of them has to be made under a timer.

Then practise the openings on their own. Take a Part 3 passage and do nothing but the one minute map: four labels, then stop. Take a Part 4 article and do nothing but the ledger: four names, four positions, four details. You are not answering questions, you are training the first ninety seconds, which is where the part is decided.

One warning about transfer from other exams. If you have prepared for a test that rewards pure scanning, question first will feel obviously correct and Part 1 will punish you for it. If you have come from academic reading, text first will feel obviously correct and Part 3 will punish you for that. The discomfort you feel switching routines is a sign you are switching, not a sign you have it wrong.

Practise it

Five questions on choosing an opening move, and on the routine that loses everywhere.