You understood the passage. Then you read the four choices, one of them sounded right, and you picked it — confidently, and wrong. In review the correct answer is obvious, which makes the miss feel like a personality flaw instead of what it is: a habit with a name, and a fixable one.
The choices are an attack surface
The four options under an SAT question are not neutral information. One is the answer. The other three are distractors — written, tested, and kept precisely because real students pick them. A wrong answer nobody picks is a failed wrong answer; the ones that reach your test survived by catching people. Check any official practice test: the wrong answers aren't random; each is aimed at a specific mistake.
And they're aimed at a specific reader: the one who arrives at the choices committed to nothing. Letting the choices propose answers instead of bringing your own is called anchoring, and it's one of the three big behavioral leaks — misses that come from what your hand does under a clock, not from what you know.
Reading the choices uncommitted changes your job
The moment you hit the choices without a prediction, the question you're answering silently changes. You stop asking what does the passage say? and start asking which of these sounds plausible? Those are different questions, and the second one is rigged — plausible is exactly what distractors are engineered to be. A good trap is fluent: familiar words, an echo of the passage's vocabulary, a statement that's even true. Fluency feels like correctness, but fluency is a property of the sentence; correctness is a property of the evidence.
There's a working-memory cost too. Four full choices are a lot of text. Once they're competing in your head, the passage — the only thing that can settle the question — fades behind them. You end up comparing the choices to each other instead of comparing each choice to the text, and choice-versus-choice is a comparison with no right answer inside it. That isn't reading; it's vibe-matching, and it works just often enough to keep you doing it.
The question types anchoring eats most
Anchoring concentrates where a prediction is possible before you look — exactly where the test builds its familiar-first traps:
- Words-in-context. The trap is the common dictionary meaning: real, familiar, wrong here. Original example: "The startup assumed the failing company's debts along with its office lease." Predict first and you'll say something like "took on." Peek first and supposed — the everyday sense of assume — sits there sounding reasonable. The words-in-context guide covers this type end to end.
- Transitions. The trap is a famous transition word running the wrong direction — a however where the logic continues, a therefore where it flips. If you haven't decided the direction yourself, every familiar word looks like a candidate.
- Inference. The trap is a statement that's true, sensible, and nowhere in the passage — something you believe rather than something the text established.
- Command of evidence. The trap is a quote or data point that's perfectly accurate but answering a different question than the one asked.
Four "content problems," one behavioral root. Students drill vocabulary lists for months to fix what is actually a sequencing error: eyes on the choices before a commitment exists.
Why process of elimination doesn't save you
Process of elimination is the most-taught SAT strategy, and by itself it does not fix anchoring — because eliminating requires a standard to eliminate against. Without a committed prediction, the only test you can run on each choice is "does this sound wrong?", and surviving that test is the one thing every distractor was built to do. Elimination without prediction is anchoring in slow motion: the choices still set the terms; you're just rejecting them one at a time instead of grabbing one immediately.
Elimination after prediction is a different tool. Commit first, then kill everything that fails the prediction — now each choice is measured against the passage instead of its neighbors, and what survives has earned it.
The self-test: covered vs. visible
Run this today with ten practice questions from the types above.
- First five: cover the choices with your hand or a card before you see them. Read the passage and the question, then commit — write your own word, or the direction ("contrast," "supports the claim"), or a one-line answer. Only then uncover and pick whatever matches.
- Last five: work normally.
- Compare accuracy.
If the covered half is better — or even equal — you've found the leak, because doing better with less information on screen should be impossible if the choices were helping you. One honesty check: if you missed both halves about equally, that's not anchoring — it's a content gap in that question type, and the fix is learning the type, not changing the habit.
The fix: Predict-Then-Peek
The fix has a name and its own full guide: Predict-Then-Peek. The core move is exactly what the covered half of the self-test forced — commit before you look, every single time: a word for words-in-context, a direction for transitions, "what would prove this?" for evidence. It leans on a real finding from learning science, the generation effect: producing an answer is more reliable than recognizing one. The full protocol — what "commit" means for each question type, and how the habit survives time pressure — lives on that page; run it there instead of improvising half of it.