The advice to "study differently for STEM vs humanities" is half-right. The principles — retrieval, spacing, feedback — don't change between subjects. What changes is the shape of the retrieval. A physics exam and a history exam both test memory, but they ask for completely different outputs: one wants a computation, the other wants an argument. Studying that doesn't match the output is practice for a different test.

Here's how to keep the science and change the prompts.

The principle that doesn't change

Retrieval practice works because it strengthens the exact operation the exam performs. That's the key: not "strengthens memory" in the abstract — strengthens the operation. The exam asks you to produce something (an answer, a derivation, an argument), and practice works best when it's the same kind of production.

This is why rereading fails universally — it's the same passive recognition regardless of subject, and it trains nothing the exam will ask for. And it's why "just do retrieval" is incomplete advice: retrieval in the wrong shape is still mismatched practice. Reciting history dates prepares you for a quiz on dates, not for the essay exam that asks you to deploy them as evidence.

STEM: practice the operation, not the topic

STEM exams rarely ask "what is X" — they ask "use X to solve this." The knowledge is the method; the retrieval is applying it to a problem you haven't seen before.

What works:

• Problems that require method selection. Not "20 integrals" — mixed problems where the hard part is recognizing which method applies. The exam doesn't label the technique; practice shouldn't either.
• "Why" prompts alongside the problems. Understanding why the method works is what lets you adapt it to the unfamiliar problem the exam will definitely include. One "explain why this works" per concept beats another five near-identical problems.
• Re-solving, not re-reading. The wrong move: reviewing the worked example and nodding along. The right move: covering the solution and producing it yourself. The gap between "I followed that" and "I produced that" is exactly what the exam measures.

Humanities: practice the argument, not the facts

Humanities exams ask for analysis — a position defended with evidence, under time constraints, in prose. The knowledge layer (dates, names, texts) is the raw material; the exam tests what you build with it.

What works:

• Timed outlines instead of full essays. A 10-minute outline — thesis, three evidence points, counterargument — is the retrieval practice without the writing time. You can space and repeat outlines; full essays take too long to be the repetition vehicle.
• Evidence-recall prompts. "List three examples of X with their significance" — the retrieval of deployable evidence, not just facts. The exam asks you to use the evidence, so practice retrieving it in argument-ready form.
• Thesis drills. "Take a position on Y in one sentence." Weak theses sink essays; strong ones organize them. Practicing the move separately builds the skill faster than discovering it mid-essay.

What stays the same across both

The differences are in the prompts; the schedule and the feedback are universal:

• Spacing works everywhere. Re-solving problems and re-outlining essays both benefit from expanding intervals — the mechanism doesn't care what you're retrieving.
• Interleaving trains discrimination. Mixed problem sets train method selection; mixed essay prompts train argument adaptation. Both are the same principle: the exam doesn't tell you which category a question belongs to.
• Feedback closes the loop. STEM has it easy — the answer is right or wrong. Humanities needs it built in: compare your outline against a strong example, or have someone check whether the evidence actually supports the claim.
• Timed practice is the final layer. Both types of exam are timed; both benefit from practice under the real constraint. The last week before the exam is when timing becomes the training.

The one-method-fits-all failure

The mistake students make isn't using the wrong technique — it's using one technique for everything. Usually rereading, because it's the lowest-effort option that produces the comforting feeling of coverage.

The fix is a two-question check before every session:

1. What does the exam ask me to produce? (compute / argue / analyze / apply)
2. Is this session practicing that production?

If the answer to the second is "no, I'm just reviewing," the session is recognition practice for a production test — regardless of which subject it is.

Where LearnX fits

LearnX generates prompts in the shape the exam needs — problems and derivations for STEM, argument-and-evidence prompts for humanities — from your actual course materials. The Exam Sprint handles the spacing and re-testing automatically, and the feedback layer works for both formats: correctness checks for STEM, explanation checks for humanities. The mechanism stays the same; the prompt form matches the subject.