Inferences

What an inference question asks

A friend tells you: “The last time Marcus canceled plans this last-minute, it turned out he’d lost his job the week before and didn’t want to say so yet.” You don’t need to be told what to think the next time Marcus cancels. That is an inference: moving from a stated pattern to an unstated but supported conclusion. You also know the difference between that reasonable bridge and an unsupported one. Concluding that Marcus has lost his job because he canceled once is a much bigger leap than concluding that something might be wrong.

The SAT calls this skill Inferences. The question gives a passage that stops short of saying something outright, often breaking off mid-sentence at a blank, and asks what follows most logically from what the passage actually gives you. The correct answer is the choice the text supports most strongly while adding the fewest assumptions of the four.

Choose the conclusion the evidence supports

The answer may paraphrase language already in the passage, but it must express a conclusion the evidence supports, not simply repeat a detail. What it cannot do is rest on a fact the passage never supplied. Read the passage as a small argument with its final step missing; that step is often visible before you reach the choices.

When the direction of the argument is clear, predict the conclusion first. When it is not, treat each choice as a hypothesis and test it against the passage. Several choices may be true in the real world and still be unsupported by this passage. Reading the choices cold invites you to build a case from assumptions you supplied yourself.

Use connectives as logical clues

Many passages place a connective directly before the blank, and that word narrows what can follow.

• Therefore, thus, so: a consequence of what was just established, not something that adds nothing.
• However, but, yet: something running against the expectation just raised, not a simple continuation.
• Because, since: the reason for the preceding statement, not a result (which would run the causal arrow backward).
• This suggests that: a cautious conclusion, hedged as the phrase is; it cannot claim proof.
• In other words: the same content restated, sometimes drawing out what is implied; it cannot add a new fact.
• It follows that: a conclusion strongly supported by what precedes, not a guess needing outside facts.

When no connective appears, take the relationship from the passage itself. Either way, read the words before the blank and finish the sentence in your own words. The completion has to satisfy both the logic of the passage and the grammar of the sentence, so check the two separately.

Recognize common inference patterns

Many passages hand you two facts and leave the connection unstated. These shapes recur often, each valid only under the condition named.

• A necessary condition, then a case where it is absent supports the conclusion that the outcome cannot occur, but only when the passage actually states the requirement (“only when,” “is required for”).
• An expectation, then a measurement running against it supports the conclusion that the expectation needs revising in the respect measured.
• A requirement, then a supply falling short supports the conclusion that the requirement goes unmet.
• Two groups alike in one way, then a difference in outcome supports the conclusion that the shared feature cannot by itself explain the difference.
• A timing, then a second timing that no longer matches supports the conclusion that two processes that must overlap may now fail to.

Worked example: timing

An orchard’s apples are pollinated by a bee that emerges in the second week of April and flies for about fourteen days. The old variety blossomed in that same week. Growers replaced it with a variety that blossoms three weeks earlier, and the bees still emerge on their usual date. The change therefore ___. The best completion says it risks leaving the blossom open before the pollinators are flying. The passage supplies the flight period, the old blossom date, and the new, earlier date. A choice claiming it “proves” lower fruit yield invents a yield comparison; a choice saying the bees adjusted states the opposite of the records.

Worked example: apply the rule to the case

A fungus fruits only after a week in which night temperatures stay above twelve degrees. Weather records show no such week in 2023. It follows that in 2023 the fungus ___. It produced no fruiting bodies. “Only after” makes the warm week a requirement, and the record shows it never occurred. Saying the fungus was killed by cold adds a fate the passage never reports, and saying it will not fruit in future seasons reaches past the one season on the page.

Match strength to the evidence

A common distractor keeps the right topic and the right direction but overstates certainty or scope without announcing it. Match the strength of your language to the strength of the passage’s.

• If a factor may contribute to an outcome, you can say it is one factor associated with the outcome, not that it causes it.
• If something held in the plots studied, you can say it held in these plots, not in all soils.
• If researchers suggest, you can say they propose, not that they have shown.
• If no difference was found, you can say the study did not detect one, not that the factor has no effect.
• If an effect held for most of a sample, you can say a majority, not every case.

Supported versus merely possible

A mill closed in 1971, the town lost four hundred jobs within a year, and census records show the population falling from 1972 onward. What is supported: the decline began after the closure, around the time the jobs were lost. What is merely possible: families left because they could not find other work. The second feels natural, but it needs a premise about motive that the passage never supplies, since census records count departures, not reasons.

To separate the two, ask what unstated premise each choice requires. The best answer needs no unsupported factual premise. A merely possible choice needs one you supplied without noticing. Rank the survivors by how much each assumes, and prefer the one supported most directly. A handy test: finish the sentence “for this answer to be true, I also have to assume ___” and prefer the choice that needs the least filled in.

Comparison groups and multiple changes

When two cases share a feature, that feature cannot by itself explain an outcome found in only one of them. A feature that differs becomes relevant, though difference alone does not prove cause. When several things changed at once (temperature rose, habitat changed, predators declined), an answer that picks one as the cause goes beyond the passage unless its evidence separates them.

Causation, design, and sampling

A correlation can arise because a third factor influences both variables. Random assignment is the design feature that helps close that gap: it makes group assignment independent of participants’ preexisting characteristics in expectation, so a later difference between groups can typically be attributed to the treatment. It reduces systematic confounding and tends to balance known and unknown factors, especially in large samples, but does not guarantee exact balance. An observational study that finds the same 18 percent difference, with farmers choosing whether to use a fertilizer, supports only an association, since those farmers may also have better soil or more experience.

Watch how a sample was gathered. Selection bias arises when cases reached the study through a process tied to the trait being studied, as when a survey of gym members says something about people who join gyms, not people generally. Survivorship bias is a narrower form: only cases that remain observable are included, so whatever caused others to drop out never appears. A study of successful founders who took big early risks cannot show that risk leads to success if failed risk-takers were never included.

Possibility, probability, certainty

Could, may, and is consistent with leave real alternatives open. Likely needs evidence that favors one possibility over the others. Likely does not become must. A fish survey comparing a protected zone with unprotected areas matched on depth and habitat can support “protection may be a contributing factor,” since the matching reduces two alternative explanations. It cannot support “proves protection causes growth.”

Keep population, time, and setting fixed

A result in one species, age group, city, soil, or period cannot automatically be applied to a larger category, the distant future, or another setting. A twelve-week trial on wheat in sandy soil speaks to wheat in sandy soil under those conditions. A failure to detect a difference under the conditions tested does not prove that no difference exists under any condition.

Ground motive and attitude in the passage

A person delaying, returning, hiding, repeating, or refusing something supports a limited inference about the behavior. A specific feeling behind it (fear, jealousy, guilt, strategy) needs separate evidence. A council that debated a fee for three sessions and gave no public reasoning supports “deliberated for an extended period,” not “feared backlash” or “intended to minimize opposition.” Likewise, an author spending several sentences on a theory has not revealed approval or skepticism.

Direction, boundaries, and complete comparisons

• A boundary is not an exact value. Paper made after 1812 means the manuscript cannot predate 1812. It gives a floor, not a window like “soon after.” Two bounds together, such as no earlier than 1812 and existing by 1840, support a span, not a single year.
• A comparison needs both sides. A method that is 90 percent accurate in the laboratory cannot be ranked against methods the passage never describes. A greater than B and B greater than C supports A greater than C; A greater than B alone says nothing about C.
• Conditionals run one way. “If X, then Y” lets you conclude Y from X, and not-X from not-Y. It does not let you conclude X from Y, or not-Y from not-X. “X only if Y” makes Y necessary, not sufficient. If X requires both Y and Z, neither alone suffices.
• One exception defeats “all,” not “generally.” A single edible bright beetle weakens “bright insects always signal toxicity” while leaving “generally” intact.
• The final sentence may set the limit. A coating that halves corrosion in humid conditions but gives no protection in freezing conditions has an effect that depends on the conditions. Do not complete with only the headline result.

Rival explanations and partial explanations

When a passage offers two explanations without choosing, the question sometimes asks what remains true whichever is correct. If bacteria of the same strain appear on leaves and in roots, and one account says soil bacteria splashed up while the other says airborne bacteria settled down, the supported statement is that the bacteria in both places may share a common source. Do not pick a winner the passage never picks.

When two components each cover part of a task and the combination does better, the supported inference is that the combination covers more of what the task requires. When a finding is one that an expected process alone would not predict, the supported inference is that the process is incomplete, without naming the missing cause.

Reading the data behind the claim

Base rates matter. A test that correctly flags 99 percent of people with a condition, but also flags 1 percent of those without it, will produce mostly false positives when the condition affects 1 in 10,000 people: about 1 true case against about 100 false alarms. Group-level patterns do not describe each individual either; a link between district spending and average scores does not show that spending raises any one student’s score by the same amount.

The method

1. Read to the blank and note the word standing directly in front of it.
2. Name the two or three facts the passage has actually put in play.
3. Finish the sentence in your own words before comparing the choices.
4. Match your ending against the four choices, looking for the closest fit rather than an exact word match.
5. Test the survivors by asking how much each assumes beyond what the passage gives.

The strongest answer is not automatically the narrowest. It is the most complete conclusion the supplied facts justify. A broader answer is right when it preserves every condition the passage states, and a narrow one is wrong when it leaves out a conclusion the facts clearly support.

Compare close conclusions

One word can alter scope, probability, or logical force. Keep these pairs apart: can and will; may and likely; likely and certain; some and most; most and all; in this sample and generally; associated with and caused by; could explain and does explain; weakens and disproves; limits and reverses; not before X and at X; at least X and exactly X; no detected effect and no effect exists; not observed and impossible; A rose more than B and A finished higher; happened before and caused; if X then Y and if Y then X; the behavior suggests hesitation and the person is afraid; fits the theory and the theory is correct; more precise and more extreme. Precision fits all the supplied facts. Extremity adds reach, certainty, cause, or motive that the facts do not earn.

Finally, a passage full of names or technical vocabulary does not change the task. Reduce names to A, B, group 1, group 2. Translate only the relationship the conclusion needs: earlier or later, more or less, same or different, one explanation against another.