The answer is smaller than you think
In school, "inference" means reading between the lines: theme, subtext, what the author implies about the human condition. Throw that definition out. SAT inference questions live in the Information and Ideas domain, and they behave like logic problems, not literature. The stem usually reads "Which choice most logically completes the text?" or "Based on the text, [something] would most likely...". Both ask the same narrow thing: given only these sentences, what is forced?
Forced is the whole game. The right answer is often so modest it feels too obvious, a near-paraphrase of a consequence the passage already set up. Students talk themselves out of it because it is not impressive enough. Meanwhile the exciting answer, the one that connects the passage to something you already know about the world, is the trap. The test is not rewarding insight. It is rewarding restraint.
Words convict, vibes acquit
Here is the one method to run on every inference question: the Receipt Rule. A receipt is the exact clause in the passage that forces a claim to be true. No receipt, no conviction.
Run it in two moves. First, before you look at the choices, Predict-Then-Peek: in your own words, say the smallest thing the passage now makes true. Second, take each choice back to the text and try to underline the words that convict it. If the only defense you can give is "well, that makes sense" or "that is probably true," you are holding a vibe, and vibes acquit. Eliminate it, however reasonable it sounds. (The mechanics of both moves live in our guides to the Receipt Rule and Predict-Then-Peek.)
This single habit defeats the two distractor families that catch nearly everyone: answers that are too strong, and answers that are outside the text. Handle those two and you have handled most of the wrong choices you will ever see on an inference question.
The too-strong filter
The first family overshoots. The gist is correct, but the degree is not. Watch for the loud words: always, never, proves, eliminates, only, must, entirely. The passage hedges, and the choice does not.
The tell is temperature. Match the temperature of the answer to the temperature of the text. Suppose a passage reports that "most participants who took the supplement described milder headaches." A too-strong choice says "the supplement eliminates headaches." The idea is right and the intensity is wrong: "most" is not "all," and "milder" is not "eliminates." The defensible version, "the supplement may reduce headache severity for many people," is the one the SAT rewards. When two choices say the same thing and one is louder, the quiet one usually wins. (Full pattern in too-strong answers.)
Out-of-text traps: the plausible ones hurt most
The second family, and the single largest source of inference errors, is not a hard idea you missed. It is an easy, true-sounding statement the passage never actually makes. The distractor writers take a fact adjacent to the topic, something you could believe from general knowledge, and offer it as if the passage backed it. Your brain, primed by the passage, nods along. That reflex has a name: anchoring, letting the answer choices think for you instead of the text.
The fix is a question swap. Do not ask "could this be true?" Almost anything could be true. Ask "does the text say it?" If you cannot point to the clause, it is out of the text, and out is out. (See out-of-text answers and its behavioral root, answer anchoring.)
A worked example (try it before you read the walkthrough)
Botanist Lena Ortiz sampled the nectar of a single desert shrub throughout one day. Nectar she collected at dawn was noticeably sweeter than nectar she collected at midday, yet the shrub produced the same total volume of nectar at both times.
Which choice most logically completes the text? Ortiz's measurements suggest that:
A) the shrub releases more nectar at dawn than at midday. B) the sugar concentration of the shrub's nectar changes over the course of a day. C) this shrub produces sweeter nectar than any other desert plant. D) pollinators are drawn to the shrub at dawn because its nectar is sweeter then.
Predict first, before touching the choices: same volume, but sweeter at dawn than at midday, means the sugar is more concentrated at dawn. The forced claim is that concentration changes across the day.
B is that claim, almost word for word. The receipt: "sweeter... at dawn" plus "the same total volume." Convicted.
A contradicts the receipt. The passage says the volume did not change, so "more nectar at dawn" runs against the text, not merely past it.
C is out of scope and too broad. Ortiz measured one shrub across one day. There is no comparison to other plants anywhere in the passage. It sounds grand and carries zero receipt.
D is the seductive one. In the real world, sweeter nectar probably does attract pollinators. But try to underline the words about pollinators. There are none. D is out-of-text: true in the world, absent from the passage. This is exactly where anchoring bites, because the choice hands you a story and your brain fills the gap.
Notice that the winner, B, is the smallest of the four. It adds nothing the sentences had not already forced.
Run the method, then test yourself
The full loop is short. Read the passage. State the smallest forced claim in your own words. Then convict each choice with the exact words that prove it, killing anything you can only defend with a feeling, anything louder than the text, and anything the text never mentions. What survives is almost always the modest, near-paraphrase answer you were tempted to skip.
Here is the self-test to run today. Pull one inference question you missed on a recent practice section. Cover the choices and write the smallest forced claim in your own words. Now uncover the choices and, for your original pick and for the correct answer, try to underline the exact words that convict each one. Most students find the same thing: their wrong pick had no receipt, only a vibe, and the right answer was smaller than the one they wanted. That gap is your leak, and naming it is how you close it. (If your misses cluster on questions you could re-solve calmly the next morning, the problem is behavior under the clock, not reading skill.)
Practice it now: the free command-of-evidence drill at /?drill=command-of-evidence runs inference and evidence questions with per-trap feedback, so you see exactly which family caught you. No signup, works as a guest.