There are two kinds of SAT misses. The first kind you can study away: a rule you never learned, a word you didn't know. The second kind is stranger — questions you knew, and missed anyway. You could re-solve them cold the next morning. Those misses don't come from what you know; they come from what your hand does under a clock. We call them behavioral leaks, and they're the reason smart students plateau (the full plateau playbook is its own guide): every leak survives every content course ever sold, because there is no rule to learn for a habit. There is only measuring it, naming it, and catching yourself in the act.
Here are the three big ones, each with a self-test you can run this week with nothing but a practice test and honesty.
Leak 1 — First-instinct override (switching off right answers)
First, the honest science, because the folk advice is wrong in both directions. Research on answer-changing — a real literature; look up the "first instinct fallacy" — finds that on average, students who change answers switch wrong-to-right about twice as often as right-to-wrong. So "never change your answers" is a myth.
But that's the average, and averages hide the group that matters. A meaningful minority of test-takers show the reverse pattern — and the research doesn't say who's in that group; the self-test below tells you whether you are. For them, the first read is text-supported and the review screen talks them out of it. That's not an accident of the test; wrong answers are engineered to survive a doubting second look better than honest answers do. For this group, review time is where earned points go to die.
The self-test: grade your last two or three practice sections twice — once with your first picks, once with your final answers. If your finals win, your reviewing works; keep it. If your firsts win, you have the leak.
The fix — the Receipt Rule: you may change an answer only when you can point to the exact word in the passage that proves your first pick wrong. A feeling is not a receipt.
Leak 2 — Anchoring (letting the choices think for you)
The four answer choices aren't information. Three of them are engineered attacks, built to look plausible to a reader who hasn't committed to anything yet. The moment you read them uncommitted, you stop reading the passage and start matching vibes — and plausible is exactly what the traps are built to be.
The self-test: take ten questions. On the first five, cover the choices, commit to an answer in your own words, then uncover and match. On the last five, work normally. Compare accuracy. Many students find the covered half better — which should be impossible with less information on screen, and isn't, because the choices were never information.
The fix — Predict-Then-Peek: commit before you look, every time. This also leans on a real finding from learning science (the generation effect: producing an answer beats recognizing one). It is the single highest-value habit change on the reading side of the test — and it applies beyond reading questions: deciding a transition's direction before looking at the word choices (the transitions guide) is the same anti-anchoring move.
Leak 3 — The time-sink (one question mugs the next three)
The four-minute question doesn't cost you one point. It costs you the three easy questions you rush afterward, with a drained clock and a tired brain. Time-sinks never show up where they happen — they show up downstream, disguised as "careless" misses.
The self-test: after your next timed section, list every question you spent 90+ seconds on. Then look at what happened on the three questions after each one. That cluster of uncharacteristic misses is the leak.
The fix — the 90-Second Flag: if a question isn't clicking at ninety seconds, flag it, guess, and move. Return in the flagged round with whatever time is left. Every question is worth exactly one point; spend your clock like you believe it.
How to tell a leak from a skill gap (the honesty check)
Before you blame behavior, rule out the boring explanation. If you score badly whether the choices are hidden or shown, you don't have an anchoring leak — you have a content gap in that question type, and the cure is learning, not habit-catching (start with the grammar guide if the gap is in Writing). The sorting question for any miss: "Would I get this right if I saw it fresh tomorrow?" Yes → behavior. No → knowledge.