MCAT prep · practice questions vs flashcards · the ratio nobody quotes

MCAT practice questions vs flashcards: the exam already set the ratio

Matthew Diakonov, Written with AI

Published May 19, 20269 min read

Almost every guide on this question ends in the same sentence: use both, flashcards build the foundation and practice questions apply it. That is correct and it is also where they stop. They do not tell you the split, they leave you to guess it, and the guess most students make (review with cards first, switch to questions later) lines the ratio up with the calendar instead of with the exam. The MCAT already set the ratio. It is printed in the question blueprint.

See how one upload does it →

Direct answer · verified May 19, 2026

Each MCAT science section pairs 44 passage-based questions with 15 discrete standalone ones, and the CARS section is 53 questions that are entirely passage-based. Add it up and 185 of the 230 scored questions, about 80 percent, ask you to apply a fact inside a passage. Flashcards train the other layer: recognizing an isolated fact with no passage attached. So the honest split is roughly four of every five reps on practice questions, one fifth on cards. Cards are real work; they are just not most of the work.

Source for the question split: the AAMC “What’s on the MCAT Exam?” overview. Each science section is 44 passage questions plus 15 discrete; CARS is 53 passage questions.

0MCAT scored questions that are passage-based

0questions that are discrete / standalone

0%of scored questions need passage reasoning

0total scored questions on the exam

The blueprint sets the ratio, not your study calendar

A flashcard rep and a passage question rep are not the same exercise wearing different clothes. A cloze card shows you "the rate-limiting enzyme of glycolysis is ____" and you fill the blank. That is the discrete-question skill exactly: a fact, retrieved cold, no context. The MCAT has 45 questions like that and a card is the cheapest, fastest rep for them.

The other 185 questions do something a card never rehearses. They drop the fact into a 200 to 400 word passage with a graph, an experiment, or a patient, and ask you to use it. Knowing that phosphofructokinase-1 is rate-limiting is necessary and not sufficient; the question wants you to read a figure of rising AMP in a sprinting muscle cell and connect it to flux through that enzyme. You can have a fully mature deck and still freeze on that question, because the deck measured how fast you recognize a fact front, not whether you can run the fact through a passage.

This is why phasing the two by calendar misfires. The instinct is content review now (cards), practice later (questions). But the application skill is 80 percent of the score and it is the slower skill to build, so the moment you understand a chapter is the moment to start drilling passage questions on it, not three weeks later. The ratio is not a schedule. It is a weighting you hold from week one.

One rep on each side, compared honestly

A flashcard rep and a Studyly question rep are different exercises. Here is what each one actually does, and where each maps onto the 230-question exam.

Feature	Flashcard deck	Studyly
What one rep actually trains	Recognition of an isolated fact. You see a cloze front, you fill the blank. That is the discrete-question skill, the 45 standalone questions, about a fifth of the exam.	Both layers. Image-occlusion and flat-MCQ formats train the isolated-fact skill; case-style passages train the application skill the other 185 questions score.
Behavior on the third revisit of a fact	The card front is byte-identical to revisit one. By pass three your brain matches the sentence, not the biochemistry. The deck still feels hard; it has quietly become a recognition test.	The question stem is re-rephrased every revisit. Same underlying fact and same spaced schedule, new wording and reshuffled options, so the rep stays a retrieval instead of a pattern-match.
Source the question is written against	A community deck built for a generic curriculum (MileDown's roughly 2,900 cloze cards, for example) or a publisher's fixed set. Not the chapter you read this morning.	Generated against the exact PDF you upload: your Kaplan or Princeton Review chapter, your undergrad biochem notes, a Khan Academy lecture transcript.
Question shape versus the real exam	One shape, a front and a back. An MCAT science section is ten passages with 44 passage questions; a flat card never rehearses reading a fact inside a passage.	Four formats from one upload: flat MCQ, free-response, case-style passage that mirrors the 10-passage science-section shape, and image-occlusion flashcards.
Quality control on the question itself	None on a card you wrote yourself; a community deck is only as good as its author and its last update.	Four-axis rubric: factual correctness, clarity, distractor quality, question-type coverage. Held-out three-document blind eval scored 81.3 of 100, versus Turbolearn at 57.8.

A flashcard deck is the right tool for the 45 discrete questions. This table is about everything else.

A green deck does not mean the application skill is handled

There is a second thing the "use both" guides skip, and it is the reason a green deck lies to you. A flashcard front is byte-identical on every revisit. The first time you see "rate-limiting enzyme of glycolysis is ____" you retrieve the answer. By the third pass your brain has memorized the sentence as a visual string and matches it to the back before you have read past the word "rate-limiting." The card still feels hard, because the wording is unfamiliar to a stranger, but it is no longer testing biochemistry. It has quietly become a recognition test scheduled at a clever interval.

The MCAT is built to defeat exactly that. The passage rewords the fact every time, dresses it in a new figure, a new experiment, a new cell type. Memory encoded as a fixed sentence does not survive the rewording. This is the same failure mode that bites USMLE study, and it is covered in depth on the recognition-versus-recall page; the MCAT version is just earlier and the stakes are a single sitting.

None of this means flashcards are a trap. They are the right tool for a real and specific slice of the exam: the genuinely context-free facts that the 45 discrete questions test. For that layer a card is faster than any passage. The mistake is not using cards. It is letting a finished deck convince you the application skill, the other four fifths, is also handled. Here is the slice where a deck genuinely earns its time.

Where a flashcard deck genuinely earns its fifth of your time

The 20 amino acids: names, one-letter and three-letter codes, side-chain class, pKa values
Essential equations: kinematics, gas laws, thermodynamics, optics, circuits
Hormone source, target tissue, and second-messenger pairs across the bio and P/S overlap
Enzyme-class, photoreceptor, and amino-acid-derivative trivia with no passage to reason through
Anything you can answer in three seconds with zero context: that is a discrete-question fact, and a card is the cheapest rep for it

Anchor fact · you do not have to pick a side

Four formats from one upload, in about 0s

Here is the part no flashcards-versus-questions article covers, because it is a product behavior rather than a study tip. You do not actually have to choose which layer to build. Upload one PDF to Studyly, a Kaplan or Princeton Review chapter, your undergrad biochem notes, a Khan Academy lecture transcript, and about 60 seconds later you have four formats generated from that single source.

Flat multiple-choice for fast recognition reps. Free-response prompts for cold recall with no options to lean on. Case-style passages, a 200 to 400 word passage followed by questions, which is the exact shape of an MCAT science block (10 passages, 44 questions). And image-occlusion flashcards over any diagram in the source, which is the genuine flashcard layer for a labeled pathway or an anatomy figure. One upload, both sides of the practice-questions-versus-flashcards line, no second purchase.

The questions are written against the chapter you uploaded, not pulled from a bank built for a generic curriculum. On a held-out three-document blind eval scored on factual correctness, clarity, distractor quality, and question-type coverage, Studyly-generated questions scored 81.3 of 100. Turbolearn scored 57.8 on the same documents. And on every revisit the question stem is re-rephrased while the underlying fact and the spaced-repetition schedule stay fixed, so the rep cannot decay into the recognition test a static card becomes by pass three. If you want one offline review queue, Studyly exports an Anki .apkg, image-occlusion cards included.

One biochem fact, one exam week, two loops

Same underlying fact: phosphofructokinase-1 is the rate-limiting enzyme of glycolysis, inhibited by ATP and citrate, activated by AMP. Same student, same week before the test. Left is a flashcard-only loop. Right is the both-layers loop. The gap opens on test day, not before.

The same fact under two practice regimes

Monday. You drill the biochemistry subdeck. The PFK-1 card comes up: "the rate-limiting enzyme of glycolysis is ____, inhibited by ____ and citrate." You fill phosphofructokinase-1 and ATP. Correct. Card goes green. Wednesday. The card returns. The front is the same sentence it was on Monday. You answer before you have finished reading it. It feels mastered. The deck says you know glycolysis regulation. Test day. A Bio/Biochem passage describes a sprinting muscle cell, gives a figure of rising AMP, and asks which regulatory step explains the jump in glycolytic flux. The fact you drilled all week is in there somewhere, but the card never once made you use it inside a passage. You stall on a question you could recite the answer fact to.

Same cloze front every revisit, you encode the sentence not the pathway
Trains the discrete-question skill only, about 20 percent of the exam
Recognition feels like mastery, transfer to a passage is weak

The deck on the left was not lazy. It drilled the fact faithfully all week. It just never once asked the student to use the fact inside a passage, which is the only thing the exam question actually tested.

81.3

“Generated 200 questions from a 90-slide cardiology deck in 47 seconds. Studyly scored 81.3 on the same eval where Turbolearn scored 57.8.”

Held-out three-document blind eval, Jungle internal admin Quality Comparison panel, 2026-04-24

How to actually split the two

The ratio is roughly four reps of practice questions to one rep of flashcards. Here is the weekly shape that holds that ratio without turning it into a spreadsheet.

Build the flat layer with cards, and keep it small

Put the genuinely context-free facts on flashcards: the 20 amino acids, the essential equations, hormone source-and-target tables. A community deck or your own. This is your daily five-minute memory layer. Real work, just not most of it.

Convert each content chapter to questions the day you read it

When you finish a Kaplan or Princeton Review chapter, upload the PDF. You get roughly 200 questions in about 60 seconds. Drill them while the reading is fresh. That same-day pass is your first genuine retrieval rep, and it is the moment the application skill starts building.

Weight your science reps toward case-style passages

Each science section is 44 passage questions to 15 discrete. Weight your practice the same way. Drill the case-style format, the 200-to-400-word passage followed by questions, not just flat MCQs. It rehearses the exact shape the exam scores without buying another bank.

Re-drill misses under rephrasing, never by re-reading

When a missed question returns, the wording is rotated and the options reshuffled. Five retrievals of one fact under five surface forms is the regime that survives a reworded exam stem. Five reviews of one fixed wording is not, no matter how green the card looks.

When a flashcard deck alone is the right call

There is a real case where a flashcard deck alone is the right call. If you are under two weeks out with no content foundation, raw fact acquisition is your bottleneck, and a high-yield community deck (MileDown's roughly 2,900 cloze cards across seven subdecks is the common pick) is the fastest way to load facts into a head that is currently empty. In that situation, load the deck and spend whatever time is left on full-length practice tests. The ratio argument assumes you have weeks, not days.

Two more honest caveats. If your weak section is CARS, neither cards nor content questions move it; CARS is a passage-reasoning skill and the only thing that builds it is CARS passages. And if you already have a mature deck plus a solid content base, do not rebuild any of it. Keep the deck for the flat layer and add only the passage-style question layer for the application gap. The claim on this page is narrow: do not let a finished deck tell you the four-fifths skill is done.

Drop one chapter and get both layers in 60 seconds

Upload a Kaplan or Princeton Review chapter, your undergrad biochem notes, or a Khan Academy lecture. You get flat MCQ, free-response, case-style passages, and image-occlusion flashcards from the same source. Free tier on app.jungleai.com, no credit card.

Start with one chapter →

Frequently asked questions

Should I use practice questions or flashcards for the MCAT?

Both, but not in equal measure and not in separate phases. The MCAT's own blueprint sets the ratio: 185 of 230 scored questions are passage-based application and only 45 are discrete standalone facts. So practice questions should carry roughly four of every five of your reps, and flashcards earn the other fifth (the 20 amino acids, the essential equations, hormone tables). Hold that weighting from week one rather than reviewing with cards first and switching to questions later.

Is the standard 'use both, they're complementary' advice wrong?

It is not wrong, it is incomplete. Flashcards do build a foundation and practice questions do apply it. But the advice almost never names the split, so students default to a 50/50 or a calendar-based phasing that under-weights the application skill, which is 80 percent of the score and the slower skill to build. It also skips a second problem: a flashcard front is identical every revisit, so by the third pass the card is testing recognition of a sentence, not knowledge of the concept.

What percentage of the MCAT is passage-based?

About 80 percent. Each of the three science sections has 44 passage-based questions and 15 discrete standalone questions across 10 passages, and the CARS section is 53 questions that are entirely passage-based with no discrete questions at all. That works out to 185 passage-based questions of 230 total scored. The structure is published by the AAMC in their 'What's on the MCAT Exam?' overview.

Are flashcards basically useless for the MCAT then?

No. They are the right tool for a real slice of it: the 45 discrete questions that ask an isolated, context-free fact. For the amino acids, the equations, hormone source-and-target pairs, and other things you can answer in three seconds with zero context, a card is the fastest possible rep. The mistake is not using cards. It is letting a finished deck convince you the other four fifths of the exam, the application skill, is also handled.

Why don't my MileDown cards translate to full-length practice tests?

Two reasons. First, a cloze card trains recognition of an isolated fact, which is the discrete-question skill only; it never rehearses using that fact inside a passage with a figure or an experiment, which is what 80 percent of the exam asks. Second, the card front is byte-identical on every revisit, so within three or four passes your brain matches the sentence rather than retrieving the concept. The card feels mastered while the passage version of the same fact stays untrained.

Can Studyly replace my Anki deck, or do I keep both?

You can do either. From one PDF upload Studyly generates image-occlusion flashcards and flat MCQs for the flat-fact layer plus case-style passages for the application layer, and it exports an Anki .apkg if you want one offline review queue. If you already have a mature deck you like, keep it for the flat layer and add only the passage-style question layer. If you are starting fresh, one upload gives you both sides without a second tool.

How is this different from just grinding UWorld or AAMC questions?

UWorld and the AAMC official packs are excellent and fixed. They are written for a generic curriculum, you do each item roughly once, and a second pass becomes a recognition test on the wording of the first pass. Studyly generates questions against the specific chapter or notes you upload, and re-rephrases the stem on every revisit so a re-drill stays a real retrieval. The two are complementary: official banks for breadth, Studyly for repeated retrieval on the exact content you are weak on. There is more on the supply ceiling of fixed banks on the practice-question-volume page.