The instinct when you miss an inference question is to read the passage harder next time. It almost never works, because the passage is four sentences long and you understood it fine. The problem is on the other side of the screen — and the wrong answers fail in four repeating ways that you can learn to name on sight.
The question, and the one that hides beside it
The main form: a short passage ending in a blank, asking which choice most logically completes the text.
The variant: "Which finding, if true, would most strongly support / most directly weaken the researchers' hypothesis?" Different wrapper, identical skill — you're being asked what follows from what's stated, not what's true in the world.
Both live in Information and Ideas, and together they're one of the two biggest sources of loss above a 650. The full breakdown of where 700-band points go is here.
The rule underneath everything
You're checking whether an answer follows, not whether it's true.
Almost every wrong answer on this question type is a reasonable statement. Some are certainly true. The test builds them that way on purpose, which means the check most students run — "does this sound right?" — eliminates nothing at all.
One consequence worth sitting with: knowing the subject makes you more vulnerable here, not less. A student who's studied ecology will nod along at a plausible claim about predator populations that the passage never established. Your knowledge is inadmissible evidence.
The four failure modes
| Trap | What it looks like | The tell |
|---|---|---|
| 1. Too strong | The passage hedges; the answer declares. | always, never, only, entirely, must, proves, all |
| 2. Half right | First clause matches perfectly. Second clause invents something. | You stopped reading at the comma. |
| 3. True but unsupported | Perfectly sensible. The passage gives you no basis for it. | You're nodding because you already knew it. |
| 4. Reversed | Right variables, wrong direction. Cause and effect swapped. | All the right words are present. |
Trap 2 is the one that catches strong readers, and it's worth a moment. When an answer's opening clause matches your prediction exactly, you feel the click of recognition and select it. The invention is in the back half. Read every option to its full stop — this single habit is worth more on this question type than any amount of additional reading practice.
The method
- Read the passage, then the last sentence again. The blank almost always completes a logical move set up in the final one or two sentences.
- Predict, with the choices covered. Finish the sentence in your own words. Rough is fine — you're building a target, not a final answer.
- Match, don't browse. Find the option closest to your prediction.
- Eliminate by naming. For each remaining choice, say which of the four traps it is. If you can't name one, that choice stays alive.
Step 4 is what stops the loop where you re-read the same two tempting answers three times. Naming the failure is a decision; re-reading is a stall.
Worked example 1
Archaeologists long assumed that the pottery found at the site was imported, since its clay composition differs from local surface deposits. A recent survey, however, identified a clay source twelve kilometres away whose composition matches the pottery precisely. This evidence suggests that the pottery was ______
Prediction: made nearby, not brought in from far away.
| Choice | Verdict |
|---|---|
| "manufactured using techniques unknown elsewhere in the region." | Trap 3 — true but unsupported. The passage says nothing whatsoever about techniques. Entirely plausible, entirely unestablished. |
| "the only ceramic type produced at the site." | Trap 1 — too strong. "Only" is doing work the passage never authorised. |
| "imported from a settlement twelve kilometres away." | Trap 4 — reversed. Every word is drawn from the text and the direction is backwards: a matching clay source nearby argues for local production, not import. |
| "produced locally rather than acquired through long-distance trade." | Correct. Matches the prediction and claims nothing beyond what the survey establishes. |
Note that the reversed answer is the most textually faithful of the four — it reuses the passage's own numbers. Recycled vocabulary is not evidence of correctness, and on this question type it's often the opposite.
Worked example 2
The naked mole-rat is among the few mammals that live in eusocial colonies, with a single breeding female and many non-reproductive workers. Unlike eusocial insects, however, naked mole-rat workers are not physically specialized for their roles: a worker can begin reproducing if the queen dies. This suggests that in naked mole-rats, social role is ______
Prediction: flexible — set by circumstances, not by the body.
| Choice | Verdict |
|---|---|
| "determined by physiology, much as it is in eusocial insects." | Trap 4 — reversed. The passage draws exactly the opposite contrast. |
| "flexible, and the colony therefore contains several breeding females at once." | Trap 2 — half right. "Flexible" is right. The second clause contradicts "a single breeding female" in the first sentence. This is the answer most students pick. |
| "unrelated to the presence of a queen." | Trap 4 again. The queen's death is precisely what triggers the change. |
| "shaped by circumstance rather than fixed by physical specialization." | Correct. |
The half-right answer here is instructive. Its first word is the one you predicted. If you select on the strength of that word you'll get this wrong every time — and the questions that punish this are specifically the ones separating a 700 from a 750.
On any answer with two clauses, cover the first half and read only the second. Does the passage support that on its own? Trap 2 dissolves instantly under this check, and it costs about three seconds.
The support / weaken variant
Same skill, one extra step: identify the claim before you look at the findings.
Say a passage reports that city birds sing at a higher pitch than forest birds, and researchers hypothesise that higher pitch carries better over traffic noise. A question asks which finding would most strongly support that.
The hypothesis has a specific mechanism in it — pitch helps against low-frequency noise. So the supporting finding has to engage that mechanism, not merely the topic. A finding that city birds also sing more loudly is about the topic. A finding that in quiet suburbs the pitch difference disappears engages the mechanism, because it links the pitch shift to the noise rather than to the city in general.
Say the mechanism out loud before reading the options. Wrong answers on this variant are nearly always on-topic and off-mechanism.
Practising it
Filter the official Question Bank to Information and Ideas, set difficulty to hard, and do fifteen a day untimed for a week. Untimed is deliberate — you're building recognition, and speed arrives after recognition rather than producing it.
For each question, write down which trap each wrong answer was. Including the ones you didn't pick. That second part feels like busywork for about three days and is the reason the pattern becomes portable to passages you've never seen. It's the same habit that makes a proper error log work.
Budget 80 to 100 seconds each on test day. These are legitimately long questions, and the time comes from the Standard English Conventions block where 30 to 40 seconds is plenty.
Frequently asked questions
What are inference questions on the digital SAT?
Short passages ending in a blank asking which choice most logically completes the text, plus the support-or-weaken variant. Both sit in Information and Ideas.
How do you answer SAT inference questions?
Predict with the choices covered, match to your prediction, then eliminate by naming which of the four traps each remaining option is.
Why do I keep getting SAT inference questions wrong?
Because you're checking whether answers are true rather than whether they follow. Most wrong choices are reasonable statements the passage never established.
Should I use outside knowledge on SAT inference questions?
No — it's a liability. Knowing the topic makes the "true but unsupported" trap more tempting, not less.