SAT — command of evidence

Command of evidence is the question type that asks not what a writer said but what would make a claim more or less believable, and it rewards a habit rather than a fact: before looking at the choices, ask what would be different if the claim were false. Ten questions here, in two halves. Five give a researcher's or a writer's hypothesis in a sentence or two and ask which finding would most strongly support it, or most directly undermine it. Five describe the results of an invented study in ordinary sentences — percentages for two or three groups, sometimes across three years — and ask which statement those numbers actually support.

The textual half is built around a distinction that decides most of these questions: being consistent with a hypothesis is not the same as being evidence for it. One question puts the point directly. A researcher thinks birds nest in ornamental trees because the foliage is denser; one choice reports that the birds also feed in the native woods nearby, which fits her hypothesis perfectly well and fits its denial just as well, and is therefore worth nothing to her. The keyed choice is the one where the proposed cause is taken away — the trees pruned to equal density — and the preference disappears with it. The same shape governs the two questions that ask what weakens a claim: a rival cause unseats a claim about the main cause, while a consequence of the thing being explained cannot explain it, which is the commonest way this type is failed.

The quantitative half is about reading three or four numbers carefully once. Each wrong answer misreads them in a particular way, and the ways repeat across the section: comparing the wrong pair of figures, stating a correct comparison in the wrong direction, treating a difference between two groups as a change over time, and drawing a causal conclusion from figures that show only an association. The three-year soil trial carries all four at once — the gap between treated and untreated plots narrows from twenty-four points to eight, which a candidate in a hurry reads as widening, and the untreated plots rise substantially on their own, which is exactly why a trial keeps a group that receives nothing.

Every explanation names what is wrong with each choice by describing it rather than by pointing at its position, so the reasoning holds however the options are ordered. Several also note something the figures cannot show at all: that reported output is not measured output, and that students who chose their own textbook format may differ from each other in ways no percentage records.

Every study, researcher, firm and place here is invented, and no invented finding is attributed to any real organization, journal or university. Nothing is reproduced from any College Board publication, released test, practice test, question bank or scoring guideline. The real section presents some of this data as graphs and tables; here it is all in sentences, so the chart-reading itself is not practiced.

Zestly is an independent study tool. It is not affiliated with the College Board, which owns the SAT, and it is not an exam center.

• Ask what would differ if a claim were false, and use that to pick the evidence that bears on it
• Tell a finding consistent with a hypothesis from a finding that counts as evidence for it
• Recognize that removing a proposed cause and losing the effect is the strongest support available
• Weaken a claim about a main cause by producing a rival cause, not a restatement
• Refuse a consequence of the thing being explained as an explanation of it
• Spot a confound, where the people who do one thing already tend to do others
• Compare the right pair of figures, in the right direction, before choosing
• Refuse a causal conclusion drawn from figures that record only an association
• Notice when a stated measure has been quietly swapped for a different one
• Read a control group's own movement as a reason the control group exists

In an invented three-year trial of a soil treatment, 34 percent of the untreated plots reached the target yield in the first year, 41 percent in the second and 52 percent in the third. Among the treated plots the figures were 58 percent, 61 percent and 60 percent. Which statement do these results support?