Evidence Has to Hit the Exact Claim
If a friend tells you a new restaurant is better than the old one, and then mentions it has valet parking, you would not consider the question settled. Valet parking might be true, and it might matter to some other question, but it does nothing to prove one restaurant is better. Three separate diners returning the same dish bears more directly on food quality, though it still would not settle the broad claim that the new place is better overall.
The College Board calls this skill Command of Evidence. This chapter covers the textual form: a passage gives you a claim, an interpretation, or a conclusion, and asks which piece of evidence bears on it most directly. The target usually sits in the question's stem and sometimes inside the passage itself. The correct choice must do more than discuss the same subject. It has to support, weaken, illustrate, or otherwise address the exact point named.
Identify the Exact Claim
Summarizing the passage is not the task. Choosing the evidence that bears most directly on one target is. Before comparing choices, fix four things: the claim's owner, its assertion, its scope, and the direction the question wants (strengthen, weaken, or simply bear on it).
A useful test is to ask whether a candidate finding would be more expected if the claim were correct than if a relevant alternative were correct. The larger the difference, the more the finding discriminates. Not every question has a rival explanation in play, but when one is, compare the predictions the two accounts actually make.
"If true" means grant it. When a stem asks which finding, if true, would support or weaken a claim, accept every proposed finding as given. You are not being asked whether it is plausible or whether such a study could be run. Ask what it does to the claim once granted. A student who starts doubting the finding has begun answering a question the exam never asked.
Match the Evidence to the Claim Type
Support raises the plausibility of a claim; it need not prove it. The shape of the claim tells you what will satisfy it.
| Claim shape | What the evidence must supply |
|---|---|
| A comparison | Evidence that lets the sides be compared, directly or against a shared reference point. A single figure from one side cannot serve. |
| A cause | Evidence that makes competing explanations less plausible, such as a well-chosen comparison, manipulation, or timing. A bare association leaves more alternatives open. |
| A change over time | A baseline and a later value, or equivalent evidence of change. One undated snapshot cannot establish it. |
| Something universal (all, always, never) | For weakening, one clear counterexample. A counterexample does not defeat a claim worded as most, generally, or usually. |
| Limited to a stated condition | Evidence on both sides of the boundary: presence under the condition, absence or reduction outside it. |
| A purpose or intention | A direct statement of the aim, or behavior or context that fits the aim better than the alternatives. A result alone cannot serve, since a result may be unintended. |
Repeating the claim is not evidence for it. One choice on many questions restates the claim in fresh words. It reads like agreement, and agreement is not evidence. Ask what the choice adds that the claim did not already assert: an observation, a comparison, a measurement, or a quotation.
Two choices can support the claim, and one can support it more directly. When several choices help, compare directness, relevance, scope, and (where rivals are in play) how well each separates the rivals.
Association and causation need different levels of trust. Two things occurring together supports a causal claim only weakly, because a third factor or reverse direction can produce the same pattern. A comparison in which the cause was deliberately assigned supports it far more strongly. Neither is disqualified; say what the evidence does: it supports, it is consistent with, it weakens, or it does not establish.
The right outcome can arrive by the wrong route. If a claim says X works because of mechanism A, evidence that X works through mechanism B supports the result but leaves A untested. On mechanism questions, ask not only whether the effect happened but whether it happened for the reason the claim names.
Match Weakening Evidence to the Claim
What counts as weakening depends on what the claim asserts.
- If the claim says X is necessary for Y, show a clear case of Y without X.
- If the claim says X is sufficient for Y, show a clear case of X without Y.
- If the claim says X contributes to or increases Y, show that the association vanishes under a better comparison, that a rival explanation fits as well or better, or that the mechanism fails where it should operate. An ordinary causal claim permits Y without X sometimes, so that fact alone does not touch it.
- If the claim is universal, one valid counterexample defeats it.
Relevance decides everything. A finding about cost or unpopularity leaves an empirical claim about what happened exactly where it stood, though it bears directly on whether a program can be adopted or sustained. Necessity and sufficiency only apply when the claim itself asserts one of them; most claims assert neither. On a weaken question, also reverse your evaluation first: a choice that would be a fine supporting finding is wrong here for exactly that reason.
Match Quotations to the Whole Claim
The literary form of this question describes a character or work and asks which quotation illustrates it. Two checks resolve many of them.
Match the form of the quotation to the form of the claim. If the claim concerns behavior, choose a quotation showing the behavior in action. If it concerns what the narrator explicitly states, a direct statement may be exactly right. Direct narration is not automatically weaker than a scene, and neither form wins outside the context of what a specific claim needs.
Match the source and the moment. A claim about how a character behaves in company is not carried by a line about her private thoughts alone in a room.
Take the claim that Adeline covers her nervousness with an excess of formality. A quotation that only states it, "Adeline had never been comfortable in large rooms," carries one half at best. A quotation that shows it, "'I trust the journey was not disagreeable,' she said, and set the cup down twice before it stayed," carries both halves at once: the stiff, ceremonious sentence is the formality, and the hand that cannot place a cup is the nervousness underneath.
Worked Examples by Claim Type
Support a proposed mechanism
Researchers propose that tiny rider ants guard leaf-carrying ants against parasitic flies. The finding that most directly supports the proposal is that carriers without riders are struck by flies several times more often than carriers with them, because it shows a lower attack rate specifically when riders are present. A finding that carriers travel more slowly with a rider reports a cost and says nothing about flies. A finding that riders are smaller than carriers merely restates a fact already in the passage.
Challenge a conclusion
A city built protected bike lanes on six corridors and injuries fell 40 percent, so planners concluded the lanes made the corridors safer. The finding that weakens this most directly is that injuries fell by a similar amount on comparable corridors with no lanes, which supplies the comparison the conclusion lacked. Cost overruns bear on whether the lanes were worth building, not on their safety effect.
Support both sides of a comparison
To support the claim that sourdough loaves stale more slowly than yeast loaves, firmness measured for both kinds from a shared starting point, with the yeast loaves gaining twice as much firmness by day three, is exactly what a claim about a slower rate requires. Sourdough staying soft for five days is one-sided, and bakers' impressions are not a measurement.
Support every part of a two-part claim
A bird species moved north because winters in its old range grew milder, which reduced its cold-adapted prey. The strongest support shows both predicted components: winters measurably milder and the prey declining there over the same years. Evidence that the species has been recorded farther north supports only the movement, which the claim already treats as a premise. Even the two-part finding shows two trends co-occurring, not each causal link.
Use evidence that distinguishes rival explanations
If birds leave a roost before a storm, a pressure theory and a temperature theory both predict it, so the departure itself cannot choose between them. A finding that birds left during a sharp pressure drop while temperatures stayed constant isolates the variable that separates the theories.
Separate weakening a rival from supporting the main explanation
If climate records show no significant change near an abandoned settlement, that weakens the climate explanation and may raise the relative plausibility of a trade-route explanation, but it is not direct evidence that trade routes were the cause.
Check for confounding
Farms with a new fertilizer grew bigger tomatoes but also received much more irrigation. The finding that most directly weakens the fertilizer claim is that farms given the same extra irrigation without the fertilizer grew equally large tomatoes. That isolates the confound.
Keep direction and timing straight
A claim that rising income leads to more small-business formation is supported by income rising first and formation following, not by formation rising first. Even then, order alone does not prove cause, since a third factor could drive both. And a decline in crime that began in 2014 cannot have been triggered by a lighting program installed in 2018, whatever the program did to the trend afterward.
Match the named group and the claim's reach
A claim about students who enter below grade level needs evidence about those students, such as larger gains among them than among peers, not a schoolwide rise that may hide a different pattern. A claim that a method works across a wide range of climates needs trial sites in several genuinely different climate zones, not one favorable test.
Interpret a null finding carefully
A twelve-week trial that finds no significant difference in soil moisture between mulches supports "this trial did not detect an effect," not "the new mulch has no effect." A trial can fail to detect a real effect because of its size, duration, or measurement limits.
Keep certainty at the right level
One observational field study showing more bird species after a wetland restoration shows that the increase happened. It does not by itself show that the restoration caused it, and answers using "proves," "definitively," or "always" outrun the design. Likewise a stated aim made before the outcome existed is strong evidence of purpose, while an outcome alone is not.
Identify the Hidden Comparison
A claim can carry a comparison without the word than. Words like preserved, distinctive, unusually, disproportionately, more stable, and exceptional measure against a reference point the sentence never names. A claim that a dialect preserves an old vowel is a claim that the sound survives there and has vanished in the comparison dialects. Evidence that the sound is present locally answers half the claim. Translate the claim into the comparison it is making before you test any choice.
When Absence Is Informative, and When It Is Not
Absence of an observation counts as evidence only when a hypothesis specifically predicted that the observation should appear and the method used was actually capable of detecting it. Meet both conditions and a missing result becomes a genuine finding. Miss either and the absence is uninformative. If a theory predicts a chemical signature and instruments able to detect it at far lower concentrations than required find none, that weakens the theory. A small survey that merely failed to find something proves little.
Check Whether the Measure Matches the Construct
If a study claims a program improved employee engagement and offers a rise in optional meeting attendance as evidence, attendance is a plausible proxy, not engagement itself. It supports the broader claim only if the passage establishes that attendance validly represents engagement in this setting. Likewise, a study that defines civic engagement as voter turnout has real evidence that civic engagement, as defined, rose, but it says nothing about volunteering or political knowledge. Ask what the measure captures and what it leaves out.
Use the Right Rate, Denominator, and Dose
A count and a rate can move in opposite directions. When a claim uses disproportionately, prevalence, likelihood, rate, share, or per capita, find the denominator. A factory with 200 injuries among 20,000 workers has a 1 percent rate; one with 150 injuries among 5,000 has a 3 percent rate, three times higher despite the smaller raw count.
A graded relationship can also strengthen a causal reading. A treatment that reduces symptoms slightly at a low dose and substantially at a high dose, tracking what the theory predicted, is supported by more than one data point. Its value still depends on design and possible confounding.
Independent Evidence
A different research team using a different method and reaching the same result adds real support. A second paper that re-analyzes the very same dataset with a different statistical model does not provide an independent check, because a new model cannot correct a flaw or bias already built into the data collection.
Use This Method
- Read the stem and find the claim. Note whose it is and whether you are supporting or weakening it.
- Restate the claim in your own words, keeping its group, condition, and mechanism intact.
- Name the demand. A comparison needs both sides measured. A causal claim needs evidence bearing on the proposed relationship and, where a rival is in play, evidence that separates them. A universal claim falls to one counterexample.
- Grant every choice as true and ask what each actually does to the claim.
- Choose the choice that acts on the claim itself, not on its surrounding subject matter.
When two choices still look useful, do not ask which sounds more impressive. Ask what exact part of the claim each one supports, and check what the other changes without announcing it, leaves out, or assumes. When language is dense, strip the claim to its plainest form first, then recheck it for limiting words such as more, only, because, all, intended, or changed.