The shape of the trap
Words-in-context questions live in the Craft and Structure domain, one blank per mini-passage, four real words underneath. Here is the structural fact the test is built on: all four choices are engineered to fit the slot. They are the right part of speech, they scan smoothly, and each would sound fine read aloud without thinking. Grammar will not save you here; grammar was never the question.
So the trap has a precise shape. The wrong choice is a genuine word whose meaning contradicts the context clue, dressed up to pass the ear test. You are meant to pick it by sound, or by reaching for the word's most common dictionary meaning, instead of by the one signal that actually decides the answer: the clue the sentence gives you about what the blank must mean. Miss the clue and every choice looks equally plausible, which is exactly the state the question wants you in.
Why the common meaning is bait
The most familiar meaning of a word is a default, and the SAT prices that default against you in two ways.
The first is topic association. A word that "goes with" the subject feels right even when the logic points the other way. Put a lighting question in front of you and "dramatic" glows, because lighting and drama travel together in your head. If the sentence actually says the lights were kept plain so nobody would stop to look, "dramatic" is the trap, and its only job was to sound at home.
The second, meaner one is the secondary sense. Many tested words carry more than one meaning, and the everyday sense is not always the one the sentence needs. This is the wrong-sense trap. Think of "check" (to verify, but also to halt), "arrest" (to detain, but also to stop), "temper" (a mood, but also to soften). When the answer rides the less-common sense, your default meaning makes you skip past it; when the trap rides your default sense, you grab it. Either way the fix starts with refusing to read the choices first.
A quieter cousin is tone mismatch: a choice with roughly the right meaning but the wrong temperature, a warm, approving word dropped into a plainly critical sentence. Right idea, wrong register, still wrong.
Predict-Then-Peek: define the blank before you look
The one method that dismantles all three is Predict-Then-Peek, and words-in-context is where it earns its keep. The move is simple and the discipline is everything: cover the four choices, read the sentence, and decide in your own plain words what the blank has to mean. Not the fancy word. Your word. "Something like slowed down." "Something like praised." Then, and only then, uncover the choices and match.
Why cover them at all? Because the moment you read four options uncommitted, you stop reasoning from the sentence and start reacting to the words, which is the leak called anchoring: letting the answer choices think for you. The choices are not information; three of the four are attacks. It is the same anti-anchoring move that decides transition questions, where you commit to a direction before reading a single word, and the same first-instinct override you practice on the questions you keep talking yourself out of.
Your prediction does not need to be elegant, only committed. A rough, honest paraphrase of the clue beats a polished vocabulary word, because the clue is the only evidence the question actually gave you.
The plug-back check
Prediction narrows the field. The plug-back check closes it. Take your surviving choice, drop it into the blank, and read the whole sentence, both sides of the clue, as one finished thought. Does the sentence still say something true and coherent? Does the word satisfy the exact signal, the contrast, the cause, the definition, that pointed you here in the first place?
This is where wrong-sense traps die. A word can match your prediction loosely and still break the sentence when you read it back in full, because the clue demanded a precision the word does not have. If the sentence only half-works, if it fits the front of the clue but ignores the back, the word is wrong no matter how comfortable it feels. The trap survives a glance. It rarely survives a full re-read.
Two examples, predicted then checked
First, common-meaning bait.
Because the museum wanted visitors to move through the new wing quickly, the designers kept the lighting ___: bright enough to see by, but never so striking that anyone stopped to linger.
Cover the choices. The clue is the second half: bright enough to function, never striking. Predict something like plain, functional, no-frills. Now the choices: dramatic, utilitarian, dim, ornate. "Dramatic" and "ornate" are the topic-association traps, the words that live near "lighting design," and both contradict "never so striking." "Dim" fights "bright enough to see by." Only "utilitarian" matches the prediction and survives the plug-back. The familiar, showy words were bait.
Second, a hidden sense.
The medication did not cure the infection, but it did ___ the spread long enough for the patient's own defenses to recover.
Predict from the clue: not a cure, but it bought time, so the blank means something like slowed or stopped. Choices: accelerate, monitor, check, diagnose. "Accelerate" is the opposite and "diagnose" is off, so the real fight is "monitor" versus "check." "Monitor" is topic-familiar (medicine monitors things) and scans fine, but monitoring a spread is watching it, not slowing it, so it fails the clue "long enough to recover." "Check" is the answer, in its less-common sense of halt or restrain, the sense your default reading wants to hide. Predict-Then-Peek keeps you open to it; the plug-back confirms it.
The filter
Put the whole thing in one loop you can run under a clock: cover the choices, read the sentence, name the clue, say the meaning in your own words, uncover, match to your word, then plug your pick back in and read the full sentence to confirm it satisfies the clue. Reject any word whose meaning fails the prediction, however familiar or smooth it sounds. This is one branch of the larger wrong-answer taxonomy, and on words-in-context it does most of the damage.
Here is the self-test to run today. Pull one words-in-context question you missed recently. Cover the choices and write, in your own plain words, what the blank has to mean and which clue told you so. Now uncover and look at what you originally picked. Almost every time, the miss traces to one of two things: you never wrote a prediction and matched by feel, or your pick was the word's familiar meaning rather than the one the clue demanded. That is your leak on this question type, and naming it is how you stop repeating it. (If instead you can re-solve it calmly the next morning, the miss was behavior under the clock, not vocabulary, and needs a different fix.)
Practice it now: the free words-in-context drill at /?drill=words-in-context makes you commit to the meaning before it shows the words, then flags exactly which trap caught you. No signup, works as a guest.