The Advice Everyone Gives
"Exercise great caution if you decide to change an answer," warns Barron's test-prep guide. "Experience indicates that many students who change answers change to the wrong answer." Princeton Review echoes the sentiment. So does Kaplan. So do roughly three out of four college students when surveyed. The advice is so ubiquitous it barely registers as a claim. It is simply what everyone knows: when you are unsure on a multiple-choice test, trust your gut. Your first instinct is probably right.
It probably is not.
33 Studies, One Answer
In 1984, psychologists Ludy Benjamin, Timothy Cavell, and William Shallenberger at Texas A&M University published a review of every study they could find on answer-changing behavior. They located 33, stretching back to 1928. The finding across all of them was unanimous: answer changes go from wrong to right far more often than from right to wrong. Students who change their answers almost always improve their scores.
The ratio holds at roughly 2 to 1. For every answer a student switches from correct to incorrect, about two are switched from incorrect to correct. Some individual studies found ratios as high as 3 to 1. Not a single study in the 33 found the popular belief to be correct. The authors titled their paper with a question: "Staying with Initial Answers on Objective Tests: Is It a Myth?" The answer, after 56 years of data, was yes.
Why Does Anyone Still Believe It?
This is where Justin Kruger at the University of Illinois, Derrick Wirtz at Northern Arizona University, and Dale Miller at Stanford made their contribution. In 2005, they published a paper that did not merely replicate the finding. It explained why the false belief survives decades of contradicting evidence.
In Study 1, they collected exam answer sheets from an introductory psychology course and tracked every erasure mark across 1,561 responses. Wrong-to-right changes outnumbered right-to-wrong, replicating the standard finding. But when the same students were surveyed about their general beliefs, three-quarters said answer changing hurts performance.
Study 2 revealed the mechanism. After an exam, students were asked to recall specific instances when they had changed an answer. They recalled right-to-wrong and wrong-to-right changes at roughly equal rates. The actual ratio was 2 to 1 in favor of beneficial changes, but memory collapsed it to 1 to 1.
The culprit is counterfactual thinking. When you change an answer from right to wrong, the regret stings: "If only I had stuck with my first answer." You can vividly imagine the world where you kept it right. When you change from wrong to right, there is relief but no comparable regret. You never fully grasp the alternative where you stuck with the wrong answer, because that world is just having the same wrong answer you started with. Nothing happened. No drama. No story.
This asymmetry makes right-to-wrong switches disproportionately memorable. The availability heuristic does the rest. When you ask yourself whether changing answers usually helps or hurts, your brain reaches for available examples, and the painful ones arrive first. It feels like changing answers backfires most of the time. The feeling is strong, personal, and wrong.
What This Costs
Put the 2:1 ratio in practical terms. If a student changes six answers on a 50-question exam, the expected outcome is four wrong-to-right switches and two right-to-wrong. The net gain is two additional correct answers, worth four percentage points. For a student sitting at 88%, that is the difference between a B+ and an A-. Across a full academic year of exams, the cumulative cost of following the "go with your gut" rule can shift a transcript grade by a full letter.
The cost is not equally distributed. Kruger's data suggest that students who believe most strongly in the first-instinct rule change the fewest answers. They are, in effect, voluntarily leaving points on the table because a memory illusion told them the table was empty.
The Strongest Counterargument
In 2011, Wim van der Linden, Minjeong Jeon, and Steve Ferrara published a technically sophisticated rebuttal in the Journal of Educational Measurement. Using item response theory, they modeled the probability of a correct response after an answer change as a function of student ability and item difficulty. Their finding: within any single test item, the conditional probability of a correct response actually decreased after changing. The aggregate data looked favorable to changers only because students selectively changed answers on items where they had a reasonable chance of getting the question right anyway.
The authors identified this as Simpson's paradox, the statistical phenomenon where a trend that appears in combined data reverses within every subgroup. If they are right, the 80 years of pro-changing evidence reflect smart item selection, not a genuine benefit of switching. The objection has not been fully resolved in the literature, and some measurement experts find it compelling.
But it faces a practical limitation that matters more than the statistical one: even if van der Linden is correct about the item-level mechanics, the student-level outcome remains. In every study that has tracked total exam scores, students who changed more answers earned higher scores than students who did not. Whatever the underlying mechanism, the practical advice stays the same.
What We Didn't Prove
Nearly every study in this literature examined college students taking multiple-choice psychology exams at American universities. Whether the 2:1 ratio holds for licensing exams, medical boards, or standardized tests in non-Western educational systems is largely unstudied. The research also cannot distinguish between thoughtful reconsideration and anxious second-guessing. A student who rethinks a question based on new information recalled from a later item is in a different cognitive state than one who panics and randomly fills a new bubble in the final minutes.
The 33-study review is from 1984. While Kruger et al. replicated the basic finding in 2005 and Couchman et al. extended it in 2016 with metacognitive monitoring, no comprehensive updated meta-analysis using contemporary statistical methods has been published in the past decade. The core finding has held for nearly a century, but the field would benefit from a modern reanalysis.
The Bottom Line
Three out of four students believe changing answers will hurt their score. The data from 33 studies spanning seven decades say changed answers go from wrong to right about twice as often as right to wrong. The belief survives because switching from right to wrong generates vivid regret that is easy to recall, while switching from wrong to right generates quiet relief that fades from memory. Your first instinct about your first instinct is wrong.
What You Can Do
- Change your answer when you have a reason to. If you re-read a question and realize you misunderstood it, or if a later question triggers relevant knowledge, switch. The odds favor you roughly 2 to 1.
- Do not change answers randomly. The research supports reconsidered changes, not panicked ones. If you cannot articulate why the new answer is better, leave it alone.
- Stop trusting "go with your gut" advice. Test-prep books and teachers who repeat this are passing along folklore, not evidence. After nearly a century of research, not a single study has found the popular belief to be correct at the total-score level.
- Track your own changes. On your next exam, mark any changed answer with a small dot. Compare results afterward. You will likely find the same 2:1 ratio the literature has documented since 1928.
- Recognize the memory illusion. The next time you remember changing from right to wrong, notice that you probably cannot recall a single time you changed from wrong to right. Both happened. One is just louder in memory.