Don’t Start at NCLEX Difficulty: What the 85% Rule Says About Question Banks

There is a ritual in nursing school: buy the serious question bank, open it in semester two, get 45% of the questions right, and conclude you are bad at this. The bank is fine. The questions are excellent. The strategy — starting your practice at full NCLEX difficulty — is the part that fails you, and it fails you in a way learning science can actually explain.

The 85% rule

In 2019, researchers published “The Eighty Five Percent Rule for optimal learning” in Nature Communications (Wilson et al., 2019). The paper derived an approximately 85% optimum for particular classification-learning models and compared the prediction with artificial and animal learning tasks. It did not test nursing students, NCLEX questions, or NurseSavvy. So the honest lesson is not “keep every score at exactly 85%.” It is that useful challenge is neither perfect ease nor unstructured failure.

Now do the math on the ritual. A student meeting NCLEX-level items on fresh content runs at low accuracy and must process several unfamiliar cues at once. That can make the rationale do too much teaching after each miss. “Struggle” is not automatically learning. The Bjorks’ famous phrase is desirable difficulties (Bjork & Bjork, 2011) — difficulty in a productive window, not maximal difficulty as a virtue.

What the hard-bank model gets right

Credit where due, because it is real: UWorld-style banks are superb at their actual job. In the final stretch — content learned, judgment forming — full-difficulty items with deep rationales are exactly the desirable difficulty the science prescribes, and UWorld’s rationales in particular have earned their reputation. Case-study and NGN practice from the big players is genuinely good integration work. The error is not the tools; it is the calendar. A final-stretch tool, used as a first-contact tool, puts you at 45% for months and calls it rigor.

What building up actually looks like

The alternative is a ramp, and it has two halves:

Components first. Learning-flow cards are deliberately sub-NCLEX: one fact, one tap, climbing from recall to spotting cues to acting to evaluating — the same clinical-judgment functions the NCLEX tests, trained one at a time instead of five at once (working memory is small; splitting the load is cognitive-load theory 101, Sweller, 1988). This is where you live near 85% while the knowledge goes in.

Then integration, adaptively. Real NCLEX-difficulty questions — but served by an engine that tracks your accuracy per topic and moves you through six difficulty tiers as you earn them. You still get to full difficulty; you get there on a schedule your own data sets, per topic, so cardiac can be at tier five while endocrine is still climbing.

The tiers are not vibes — here is the bank’s calibration against real student responses, published with its n and date:

Accuracy falls monotonically as tiers rise, which is what “the tiers are real” looks like in data — full methodology and caveats here.

“But the NCLEX is hard — shouldn’t practice be hard?”

Eventually, yes — and by test day, it will be. The question is sequencing. No coach starts a couch-to-marathon plan at mile twenty; the marathon still gets run. Ramping is not protection from difficulty; it is a structured route to difficulty: you spend your error budget where errors teach (near your edge) instead of where they demoralize (far past it). And a weekly free CAT readiness exam keeps you honest the whole way — full difficulty, real formats, measured, unlimited.

The one-line version

Hard questions are the destination, not the on-ramp. Build with cards at your edge, integrate with adaptive questions, prove it weekly at full length — and save the low-accuracy grind for later, when the rationales are consolidating knowledge rather than introducing every component at once.