NOA Test Practice: the MCT-H, MCT-M and MCT-A Capaciteitentest

NOA scores each subtest as the number of items you got right, with nothing deducted for a wrong answer, and compares it with a norm group at the education level the organisation chose: people with an HBO level, for example, or MBO level 3. The adviser's report gives a stanine, a nine-point scale with a midpoint of 5, for your overall level, four factors (language, numbers, spatial and speed) and each subtest.

Your own report shows no number. Since 2024 NOA draws each result as a bar in three bands, Lager, Midden and Hoger, and says why: "Om verwarring met (school)cijfers te vermijden zijn de cijfers bij de stanines weggelaten." Measured off NOA's own report bar, Lager is the lowest three stanines (roughly the bottom quarter of the comparison group), Midden the middle three (about the middle half) and Hoger the top three. NOA does not state those cut-offs in words; the bar shows them.

A Midden is not a fail. NOA's report says a Midden result means the level is achievable for you and probably as hard or as easy for you as for the people you are compared with. There is no pass mark in the test itself: a hogeschool or an employer decides what it needs. Our feedback leads with the same band words and gives the stanine beside them, compared with CogniPrep's practice population rather than NOA's norm groups.

Two more things on the report. The MCT-M can also produce an IQ estimate, including a non-verbal IQ, for the adviser; we do not produce one. And NOA suggests a report is treated as valid for one to one and a half years.

Much of NOA's testing is not hiring at all. Its heaviest documented use is education admission: the hbo 21+ route for applicants aged 21 or over without the usual entry diploma, university selection and mbo intake. NOA says the MCT-H is used widely for "HBO- en WO opleidingen, 21+ toelating, talentontwikkeling, studieadvies en loopbaanbegeleiding".

What a 21+ toets looks like. At Hogeschool Utrecht, in NOA's own 2020 account, every 21+ applicant, about 600 a year, sat six MCT-H subtests: word relations, word analogies, arithmetic, number series, exclusion and components. Other institutions may choose a different set and often add subject tests of their own, so check your invitation for the list. A 21+ applicant usually gets the result straight away.

It may be proctored at home. For admission testing NOA offers online proctoring through ProctorExam: your screen, webcam and microphone are recorded, optionally the room through a phone camera, and the recording is reviewed afterwards. It runs in Chrome only.

How to prepare. Practise the six subtests in the list above until each format is familiar, because the unusual ones (components, word relations) cost the most time on first sight, then sit them back to back to rehearse the two-hour stretch. NOA allows a pause between subtests but not during one.

Harver, the volume-hiring platform, calls its cognitive module "NOA Cognitive Ability" in its own knowledge base, and its five subtests carry NOA's subtest names in brackets: Logical Reasoning (Exclusion), Spatial Ability (Components), Numerical Reasoning (Number Sequences), Verbal Reasoning (Word Analogies) and Perceptual Speed (Checking). Practising those five here practises the item types.

Two things differ. Harver reports each subtest on its own 0 to 100% norm scale rather than as a stanine, and it does not publish its item counts or clocks, so a Harver sitting may not match NOA's published MCT-M figures.

MCT-H counts and clocks are not published. The only per-subtest figures NOA has published are for the MCT-M, in a 2018 paper by three NOA staff. Both levels share every item format and the same two-hour sitting, so we use the MCT-M figures for both and say so on every instruction screen. Our difficulty curve targets the MCT-H.

Three numbers are ours. The checking subtest's four-minute clock (NOA publishes none; four minutes is the shortest subtest length it mentions), arithmetic's four options (NOA's MCT-H example sheets show four, the MCT-M uses five), and ten questions per part on the MCT-A (NOA publishes neither the length nor the stop rule).

Two scoring rules are ours. NOA publishes no partial credit on the pick-two subtests, so we give none: both picks must be right. And it does not say how checking is scored, so ours is correct minus wrong, which stops fast random tapping from paying on a two-choice task.

The adaptive test is a staircase, not NOA's algorithm. The real MCT-A uses a calibrated item bank and item response theory. Ours moves up a difficulty tier after a right answer and down after a wrong one or a timeout, which reproduces what you experience rather than how NOA estimates your level.

Forward only, and no practice gate. NOA's current platform moves on as soon as you answer; whether it keeps a Back button is not documented, so ours has none. And each real subtest opens with practice items you must answer correctly before the clock starts, which we describe but do not reproduce.

What we do not simulate. The SBT, NOA's situational judgement integrity test, because practising it would amount to coaching an integrity measure; the Competentietest, because each client builds its own from 41 competencies; the Dutch language test, which is meaningless in English; and NOA's interest, values and motivation questionnaires, which are for guidance rather than selection. If your invitation names the older MPT-BS personality test, the MP-B6 game covers the same format and groups.