Back to all articles
NCLEX Prep14 min readOctober 26, 2026
By Medoceo LMS ·
Last updated: 21 October 2026
The NCLEX ends on one of three rules: the algorithm reaches 95 percent confidence after at least 85 questions, you reach the 150-question maximum, or the five hours run out [1]. The number of questions you answered does not tell you whether you passed [3].
Computer Adaptive Testing is the part of the NCLEX candidates understand least and worry about most. It decides which question you see next, when your exam ends, and whether you pass. It is also the reason the exam feels impossible while you are sitting it.
This guide explains the mechanism: how the algorithm picks items, what a logit actually measures, and the three rules that end your exam.
The NCLEX is not a fixed paper. Every candidate sits a different exam, assembled one item at a time while they take it. After each answer the algorithm re-estimates your ability. It uses every response so far, plus the known difficulty of every item you have seen. It then picks the next item that will tell it the most about you[1].
The practical effect is that the exam keeps serving items pitched at the edge of your measured ability. NCSBN describes each item as targeted so that a candidate should find it challenging, which is where an answer reveals the most[1]. Answer well and it tends to get harder. Answer poorly and it tends to ease off. Because it keeps pushing towards the edge of your ability, a strong candidate spends most of the exam feeling uncertain. That sensation is the design working. It is not evidence that you are failing.
The NCLEX reports no score and uses no percentage. It measures ability in logits, a unit from Item Response Theory that places candidates and items on one shared scale. Every item has a known difficulty in logits; every candidate has an estimated ability in logits. The passing standard is a point on that same scale, 0.00 logits for the NCLEX-RN and -0.18 for the NCLEX-PN, both effective through 31 March 2029[2].
This is why chasing a percentage in practice misleads you. A high percentage on easy items can leave your ability estimate below the standard. A lower percentage on items pitched well above the standard can leave it comfortably above. What counts is the difficulty of what you answered, not the proportion you got right.
Every candidate answers a minimum of 85 items and at most 150, within a five-hour limit that includes optional breaks. The figures are the same for the RN and the PN exam[3][4].
Not all of those items count. On a minimum-length exam, 52 are content items, 18 are clinical judgment items delivered as three case-study sets of six, and 15 are unscored pretest items[3]. That is roughly one item in six that does not count, and there is no way to tell which. Treat every item as if it counts, because you cannot know. Clinical judgment reaches beyond the case studies too, because the test plan adds approximately 10 percent stand-alone items measuring it[3].
Older guides quoting 75 and 145 describe the exam as it ran from October 2020 until 31 March 2023[7][6]. Guides quoting 265 are older still. The five-hour limit did not change in April 2023.
NCSBN describes three conditions that end the exam[1].
Once you have answered the minimum 85 items, the algorithm checks whether it is 95 percent confident your ability sits clearly above or clearly below the passing standard. When the confidence interval falls entirely on one side, the exam stops and that side is your result[1]. This is the only rule that ends an exam early because the algorithm has made up its mind. It is why a short exam means you were resolved quickly, not that you did well.
If confidence is never reached, the exam runs to 150 items and your final ability estimate decides. Reaching the maximum means your performance sat close to the standard for long enough that the algorithm could not separate you from it. It does not tell you which side you landed on.
If the five hours expire, what happens depends on how far you got. Past the minimum, your final ability estimate decides pass or fail. Below the minimum, it is a fail, because there is not enough scored data to decide. This is the one stopping condition where pacing alone can cost you the exam.
Because the algorithm reached 95 percent confidence at the earliest moment it is allowed to. Eighty-five is the minimum, so it is the first point the confidence rule can fire. An exam shutting off there means your answers were consistent enough for the algorithm to place you clearly on one side of the standard.
Which side, it does not say. A confident pass and a confident fail both resolve at the same speed, and both end at 85. The same logic applies at any length: shutting off at 92 or 118 means confidence was reached at that point, nothing more.
Neither on its own, and the question contains the mistake. Good and bad are not what the length measures.
What a minimum-length exam does rule out is the middle. Ending at 85 means the algorithm found you easy to place, which says your answers were consistent. Consistent with passing and consistent with failing look identical from the outside.
The same thing it means at 85. Once past the minimum the algorithm re-checks its confidence as you go, so an exam can end across a wide range of counts between 85 and 150. There is nothing special about round numbers, and no threshold at 100 or 120 of the kind candidates sometimes describe.
Ending later simply means the algorithm needed more evidence before it was sure. That happens nearer the passing standard, on either side of it.
No, and 145 is worth pausing on because it is no longer the maximum. It was the maximum from October 2020 until 31 March 2023[6][7]. Since the Next Generation NCLEX the maximum has been 150, so an exam that stopped at 145 stopped because the confidence rule fired, exactly like one that stops at 92.
If you are working from a guide that calls 145 the ceiling, that guide predates April 2023 and its other figures are worth checking too.
Candidates ask each other this constantly, and the answers form no pattern. NCSBN publishes no breakdown of stopping points by result, and the test plan says a candidate may pass or fail regardless of the length of the examination[3]. No stopping point rules either result in or out.
Collecting other people's stopping points cannot tell you anything about your own. The variable that decides the result is your ability estimate against the passing standard, and that is the one thing nobody in the thread can see.
No. The test plan states plainly that exam length is not an indicator of pass or fail[3]. Stopping at 85 can mean a confident pass or a confident fail. Running to 150 means the algorithm stayed uncertain, which happens near the standard, on either side of it.
The only thing exam length tells you is how quickly the algorithm reached confidence. It says nothing about the direction.
Three beliefs about the algorithm circulate widely enough to change how people study. Each one can be tested against the three rules above.
This figure appears on prep sites and all over candidate forums. It is not published by NCSBN, and it cannot come from the algorithm, because the confidence rule is symmetric. The exam stops the moment the interval sits entirely on one side of the standard, and it stops just as quickly when that side is a fail[1].
What makes the claim feel right is the base rate. First-time pass rates are high, so a candidate who stopped at 85 more likely passed than failed. That is equally true at every other length. NCSBN publishes current pass rates by candidate type[5]. Whatever that figure is, it describes all candidates, not the ones who finished early.
There is no last-question rule. The exam ends when a stopping condition is met, and the item you happen to be looking at when that happens carries no special weight.
The deeper problem is that you cannot read an item's difficulty from how it feels. Every scored item's difficulty is a logit value fixed in advance, calibrated from how a large sample of candidates performed on it[3]. An item feels easy when you happen to know the content well, which is exactly what happens to a strong candidate sitting a hard item.
Two things break this. First, perceived difficulty is not measured difficulty, for the reason above. Second, a wrong answer lowers your ability estimate, so the next item is usually pitched lower. One item near the standard barely shifts an estimate already built on dozens of responses. A short easy-feeling run is not a verdict.
Not by counting right answers. Every scored item has a difficulty in logits, and your ability estimate moves according to the difficulty of what you answered rather than the quantity you got right. There is no percentage, no raw score, and no curve against other candidates on the day.
That is why two candidates answering the same number of items correctly can get opposite results. The one who answered harder items has the higher estimate. It is also why a question bank average is a weak proxy: an average only means something if the difficulty behind it is rising.
Yes, on any item for which more than one key exists, which covers matrix, cloze and highlight formats as well as multiple response. The test plan names three partial-credit methods: plus/minus, zero/one and rationale scoring[3].
NCSBN names the three methods and does not publish how each one works, and you are not told which governs the item in front of you. So the safe reading is the cautious one: on an item with several correct answers, some methods reward a partly correct response and some do not. Answer what you can defend rather than adding options to be safe.
No item is worth extra points. What differs is the difficulty attached to it, and how much that item tells the algorithm. An item is most informative when its difficulty sits close to your current estimate, which is why the test plan describes selection as aiming for an item that is not too easy and not too difficult[3]. Fifteen items on every exam are unscored pretest items and carry nothing at all[3].
Three things follow from the mechanism.
A fixed-length practice test cannot locate you against the standard. If a session always ends at the same count, its ending carries no information. On the real exam the ending is information, because it is produced by a confidence rule.
A rising percentage is only meaningful if difficulty rises with it. An average climbing on easy items tells you nothing. The question to ask of any question bank is whether it is pitching harder items as you improve.
Late-exam accuracy is what decides a borderline result. Nothing scores stamina directly. But accuracy that holds for 60 items and slips at 120 will not show up in short practice blocks. Under the maximum-length rule those later items are exactly the ones deciding the outcome.
The NCLEX assembles itself around you, measures you in logits rather than percentages, and ends on one of three rules: confidence, full length, or time. Exam length tells you nothing about your result. Practice that does not behave this way cannot tell you where you stand, however good the questions are.
Yes. The algorithm selects the item that will tell it most about your ability, which for a strong candidate means items pitched near or above your current estimate. Finding the exam hard throughout is the expected experience for someone passing.
Yes. The exam ends as soon as the algorithm is 95 percent confident your ability sits entirely on one side of the passing standard, and a clear fail reaches that point just as quickly as a clear pass.
Not by itself. Stopping at 85 means the algorithm reached 95 percent confidence at the earliest point it is allowed to, and that confidence can be in a pass or in a fail. Neither takes the algorithm longer to reach than the other. The widely repeated claim that stopping at 85 gives you an 85 percent chance of passing is folklore rather than an NCSBN statistic.
Your ability estimate falls after a wrong answer and rises after a correct one, so the difficulty of what you are served tracks how you have done. But how an item felt to answer is not how it was scored, so an easy-feeling stretch is not evidence that you are failing.
No. Reaching the maximum means the algorithm could not separate you from the passing standard with 95 percent confidence within 150 items. Candidates sit close to the passing standard from both sides. Your final ability estimate then decides, and it can land above it.
No. On a minimum-length exam, 15 of the 85 items are unscored pretest items being trialled for future exams, and candidates are not told which ones.
If you have passed the minimum of 85 items, your final ability estimate decides the result. If you have not reached the minimum, the exam is recorded as a fail because there is too little scored data to judge.
Usually, yes. The algorithm keeps selecting items near the limit of your measured ability, so a candidate performing well is served hard items throughout. An exam that felt easy is more concerning than one that felt brutal.
Yes. Both use the same adaptive method, the same 85 to 150 item range and the same five-hour limit[4]. The passing standard differs: 0.00 logits for the RN exam and -0.18 for the PN exam[2].
1. National Council of State Boards of Nursing. Computerized adaptive testing (CAT). Available from: https://www.nclex.com/computerized-adaptive-testing.page
2. National Council of State Boards of Nursing. NCLEX passing standards. Available from: https://www.nclex.com/passing-standard.page
3. National Council of State Boards of Nursing. NCLEX-RN test plan effective April 2026. Chicago: NCSBN; 2026. Available from: https://www.nclex.com/files/2026_RN_Test%20Plan_English-F.pdf
4. National Council of State Boards of Nursing. NCLEX-PN test plan effective April 2026. Chicago: NCSBN; 2026. Available from: https://www.nclex.com/files/2026_PN_Test%20Plan-F.pdf
5. National Council of State Boards of Nursing. NCLEX pass rates. Available from: https://www.ncsbn.org/exams/exam-statistics-and-publications/nclex-pass-rates.page
6. National Council of State Boards of Nursing. Summary of modifications to the NCLEX-RN and NCLEX-PN examinations starting Oct. 1, 2020. Available from: https://www.ncsbn.org/public-files/SummaryofModificationstoNCLEX.pdf
7. National Council of State Boards of Nursing. NCLEX-RN test plan effective April 2019, in force until 31 March 2023. Chicago: NCSBN; 2019. Available from: https://www.ncsbn.org/public-files/2019_RN_TestPlan-English.pdf