Ideas 1 — evidence, data and what follows

Thirteen questions covering all four skill points of the domain, and every one of them turns on the same distinction: what the text says, what the text implies, and what merely happens to be true. Only the first two can be the answer. The third is where most wrong options live.

Four on central ideas and details. What is the main idea, or what does the writer actually state about a named thing? The wrong choices here are of two kinds, and both are worth recognising on sight: something true of the world but never stated in these words, and something true of one sentence but not of the whole. A passage about the printing press notes that early books were expensive; a choice that says printing had no effect on literacy is false, and a choice that says the press caused a sudden rise in literacy is exactly the claim the passage opens by rejecting.

Three on command of evidence in prose. You are given a hypothesis — that fifteenth-century trade was driven by luxury goods rather than conquest, that arctic bird migration is governed by temperature — and asked which finding would most strongly support or most strongly weaken it. The question is never "which of these is true". It is "which of these does work on this claim", and the wrong options are usually findings that are perfectly plausible and simply irrelevant to the specific mechanism proposed.

Note the ones that say weaken. A question can ask you to break a hypothesis rather than prop it up, and reading the stem too fast is the commonest way to lose a question you fully understood.

Three on data. Since these materials carry no images, each table is described in the question in full — every variable, every unit, every value. Read the units before you read any option; that habit alone accounts for a good share of the marks here. The wrong answers are built the way the real ones are: they read the figures perfectly correctly and then do something the figures do not license — compare the wrong pair, assert a trend from two points, or conclude something about a cause when the data only shows an association.

Three inferences. A passage stops and you finish it. "Most logically completes" is a logical constraint, not a stylistic one: the right ending is the one the passage's own reasoning forces, and the wrong ones will read perfectly well. A passage describing a feedback loop in which warming causes thawing which causes more warming has already told you the shape of the answer — the process accelerates. What you happen to know about the topic is not evidence, and if you know a lot it is a liability, because there will be a choice that agrees with what you know and is not supported by anything on the page.

The habit that ties all four together: put your finger on it. For a detail or a main-idea question, the answer can be pointed at in the passage. If you cannot point at it, you have chosen the plausible one.

• Tell what a text states from what it implies, and both from what is merely true
• Identify a main idea that holds for a whole passage rather than for one of its sentences
• Judge whether a finding does work on a specific claim, instead of whether it is plausible
• Notice when a question asks you to WEAKEN rather than support a hypothesis
• Read units and variable names off a described data set before evaluating any option
• Refuse a conclusion the data does not license, even when the data has been read correctly
• Complete a passage with the ending its own reasoning forces, not the one that reads best
• Set aside what you already know about a topic when only the words on the page count

A table shows the average test scores of students in four study groups: peer-led 85, teacher-led 82, independent 78, online 75. A researcher claims that collaborative learning environments are more effective than individual ones. Which choice most effectively uses the data to support this claim? — The data has to be grouped the way the claim is framed, collaborative against individual, before it supports anything; a choice that compares the wrong pair, or that reads a ranking as a cause, has used correct figures to say something they do not license.