The honest answer-changing research
Start with the science, because the folk advice is wrong in both directions. Decades of research on answer-changing find that for the average test-taker, changed answers move from wrong-to-right roughly twice as often as they move from right-to-wrong. So "never change your answer, trust your gut" is a myth. On average, reviewing helps, and students who refuse to touch a single answer leave points on the table.
But averages hide the group that matters here. Inside that population number is a real minority for whom the pattern runs the other way: their reviewing is net-negative, and their second-guessing destroys more correct answers than it saves. The research does not tell you who is in that group, only what the average is. Which side you personally sit on is not a rule to memorize, it is a measurement you take on your own tests. That is the entire point of this page, and it is the first of the three behavioral leaks that survive every content course ever sold, because there is no rule to learn for a habit.
Measuring which group you are
Do not guess whether you are switch-prone. Count it. This is the self-test, and you can run it tonight with tests you have already taken.
Grade your last two or three practice Reading & Writing sections twice. Once using only your first picks, the answer you committed to before any review. Once using your final submitted answers. Then compare the two scores. If your finals beat your firsts, your reviewing is working: keep doing it, and stop reading here. If your firsts beat your finals, you have the first-instinct override leak, and the review screen is your most expensive habit.
One honesty check makes this test trustworthy. It only counts as the leak when your first pick was the correct answer and you moved off it. Changing a wrong guess to another answer is not the leak, that is normal churn or a good save. The signature you are hunting is specific: an answer that was right, reopened during review, and submitted wrong. If you cannot re-solve those questions cold the next morning, the miss was knowledge, not behavior, and the fix is study, not this page.
The Receipt Rule on review
Here is the fix, and it has a name: the Receipt Rule. You may change a committed answer only when you can point to the exact word, phrase, or line in the passage that proves your first pick wrong. A feeling is not a receipt. "This one is nagging me" is not a receipt. The only currency that buys a change is textual evidence you can put your finger on.
The reason this works is not willpower, it is how the test is built. Wrong answers on the SAT are engineered to look plausible to a doubting second reader, and several are designed to survive exactly the uncommitted re-read you do on the review screen, when the passage is no longer fresh and the clock is winding down. A vague sense that "another answer looks good now" is not new information. It is the trap doing its job on a tired brain. The Receipt Rule forces you to convert doubt back into evidence before you act on it, which is the difference between fixing a genuine misread and getting talked out of a right answer. For more worked cases, see the full Receipt Rule protocol.
What to check, and what to never reopen
The review screen is a navigation tool, not an invitation to re-litigate the whole module. Each Reading & Writing module gives you 27 questions in 32 minutes, and at the end the screen lists every item as answered, unanswered, or flagged. Use it for two jobs only.
Check these. Any question still marked unanswered gets a real guess first, because there is no penalty for guessing and a blank is a guaranteed zero. Then the questions you parked with the 90-Second Flag on your first pass: you never committed to those, so returning is solving, not second-guessing. A quick scan that your intended answers actually registered rounds it out, in case a tap landed on the wrong choice.
Never reopen these. Any question you answered confidently, with a reason, on the first pass, and did not flag. That is the donation box. You already did that work when the passage was fresh. Reopening it now, with less context and more fatigue, only introduces doubt the trap was built to exploit. If your finals lost to your firsts in the self-test above, treat an unflagged, committed answer as closed. The urge to "just check it once more" is precisely the leak asking for the keys.
The final-minutes protocol
Put it together into a sequence you run every time you reach the review screen, in this order:
- Cover the blanks. Fill a real answer on every unanswered question first. Insurance before anything elegant.
- Work your flags, not the field. Go only to the questions you parked on purpose. These are fair game because you never committed to them. This is the same discipline the two-pass strategy is built on: bank the cheap points first, then spend recovered time on the ones you chose to defer.
- Hold the line on committed answers. Do not open an answered, unflagged question to browse it. If a specific doubt hits one anyway, do not act on the feeling. Go find the receipt.
- Apply the Receipt Rule to any change. Point to the word that proves your first pick wrong. Found it: change with confidence. Cannot find it: the doubt is noise, and your first read stays.
- Spend leftover time as coverage. If minutes remain after your flags are done, confirm no blank slipped through. Do not go hunting for answers to overturn.
This is not about being rigid or refusing to think. Genuine misreads happen, and the Receipt Rule is how you catch them. It is about refusing to trade a right answer for a shrinking clock and a plausible trap. It also pairs with anti-anchoring: commit to an answer in your own words before you read the choices, using Predict-Then-Peek, and you reach the review screen with fewer soft answers to talk yourself out of.
Practice it now: the free simplesat first-instinct drill puts you on a live review screen, tracks every answer change by direction, and shows you whether your switches help or hurt, the self-test above run automatically. No signup, works as a guest: /?drill=first-instinct-lock