Why CARS Scores Plateau
CARS may be the only MCAT® section that punishes you for remembering. Once you have read a passage, it stops measuring your reading. A look at why CARS scores stall, what the research on case comparison suggests, and practice built around the single word that decides a question.
I have come to think CARS is the only section of the MCAT® that punishes you for remembering.
Consider what happens after you read a CARS passage once. You know how the argument ends. You know which answer was right. The next time you see that passage you will get it right again, and you will have learned nothing about reading. The passage is spent.
It is a bit like rereading a mystery novel to practice detective work. It also explains a great many stuck scores.
The Plateau Has a Shape
The section is 53 questions across nine passages in 90 minutes, drawn about evenly from the humanities and the social sciences. The AAMC sorts the questions into three skills, and the weighting is worth knowing before anything else.
Foundations of Comprehension asks whether you understood what was said. Reasoning Within the Text asks how the pieces of the argument fit together. Reasoning Beyond the Text, the largest share, hands you something new and asks what the passage would make of it. Jack Westin's CARS guide calls that last category the one students most consistently underperform on.
The plateau shows up everywhere once you start looking. One Student Doctor Network poster described six months of CARS work across passage books from several publishers, early gains, and then a flat line. A student write-up hosted by UMass describes an 11-point total improvement on a retake, after which the writer still judged CARS to be lagging badly. Both were putting in serious hours.
My guess, and it is a guess informed by the forums, is that a lot of people plateau because their practice quietly stops measuring reading.
When a Passage Becomes Familiar
There is a short thread on SDN that captures this well. A student realized the CARS question packs overlapped with older AAMC passages they had already done, and asked whether a second pass was worth it. They said it felt like memorizing. Another poster replied that picking answers from memory is essentially wasted time.
That exchange is the whole problem in miniature. Official AAMC passages are the best practice material for this section, and forum consensus ranks them well above anything a third party writes. They are also finite. Once you have read one, you know how it resolves. A second attempt becomes a test of recall, and recall feels exactly like comprehension from the inside.
The same thread offered a reasonable workaround: wait long enough to forget, then return with your attention on the reasoning path. That helps. It still leaves you rationing a small supply.
Drills Built Around One Word
Education research offers a better tool. In a 2013 meta-analysis, Alfieri, Nokes-Malach and Schunn pooled 336 tests and found that learning through case comparisons produced better outcomes than studying single or sequential cases, with an average effect of about half a standard deviation. Set two nearly identical things side by side and the difference between them becomes the lesson.
In CARS, the difference that decides a question is often a single word. A connective. A reporting verb. A hedge.
Read those two and the subject matter is the same down to the comma. The answer turns on one verb. Shown means the author accepts the finding. Claimed means the author is reporting a view and keeping some distance from it. Questions like this sit inside a great deal of Reasoning Within the Text, and they are very hard to pass on memory, because the version you remember is exactly half the pair.
This is how the CARS drills on PurpleCAT are built. Each item belongs to a family of near-identical twins that differ by one structural cue, or to a small set of questions that each ask something different about the same short passage. Answering from memory of the topic gets you the wrong answer about half the time. That is deliberate.
A fair caution belongs here. Comparison learning has limits, and some studies in the same literature found analogical comparison helped less than self-explanation on certain near-transfer tasks. I would treat cue drills as a companion to full AAMC passages, which remain the closest thing to test day you can get.
So here is what I would suggest. Treat each first read of an official passage as precious. Between those reads, drill the small cues that decide the Within and Beyond questions. When you review a missed question, write down the exact word that should have changed your answer. And be patient with yourself. Reading improves slowly, and once it improves it stays.
Sources
- AAMC, What's on the MCAT Exam?
- Jack Westin, MCAT CARS Strategy 2026
- Alfieri, Nokes-Malach and Schunn, Learning Through Case Comparisons: A Meta-Analytic Review (2013)
- Proceedings of the Cognitive Science Society, analogical comparison and transfer
- UMass, Student MCAT preparation write-up
- Student Doctor Network, CARS Qpack
- Student Doctor Network, MCAT Advise
- Student Doctor Network, CARS