How the digital SAT adapts

Every candidate begins with the same first module. In Reading and Writing that is 27 scored questions in 39 minutes; in Math it is 20 scored questions in 43 minutes. The questions are drawn from a pool of mixed difficulty and are identical in kind for everyone.

When the module ends, the testing application evaluates the performance and routes the candidate to one of two second modules: one built from easier questions, one from harder ones. The second module then runs to the same length and timing as the first.

The two subjects are routed independently. Reading and Writing module 1 decides Reading and Writing module 2, and Math module 1 decides Math module 2. A strong performance in one subject has no effect on the other.

This is the part that is rarely explained, and it is the part that matters.

A score of 800 in a subject represents a level of performance. To demonstrate that level, a candidate has to answer questions difficult enough to distinguish it from the level below. The easier second module does not contain such questions.

The consequence is that a candidate routed to the easier module cannot reach the top of the scale, no matter how they perform. Answering every question in an easier module correctly produces a strong score. It does not produce a perfect one, because a perfect score is a claim about something the module never tested.

The harder module works the other way. Its questions are more demanding, so a given number of correct answers is worth more, and the full range remains available.

The first module carries weight out of proportion to its length. It is half the questions in a subject, but it decides both its own marks and which marks are available afterward.

Three things follow.

Do not save effort for later. On a linear test, spending too long on early questions costs you later ones. Here, the early questions decide which later questions you get, so the usual advice to move briskly through the opening and settle in afterward works against you.

Answer everything in module 1. There is no penalty for a wrong answer, and a blank in the first module reduces the measured performance the routing depends on. A guess costs nothing and may keep the harder module in reach.

Do not try to read the routing. Candidates sometimes report that their second module felt easy and conclude that they did badly. Question difficulty is not obvious from the inside: a question that feels straightforward may be one you happen to be good at. Trying to infer the route mid-test produces anxiety and no useful information.

An adaptive test measures more precisely in less time. A linear test has to spend questions on every ability level to be sure where a candidate sits; an adaptive one narrows the range first and then concentrates its questions there. That is how the SAT became two hours shorter without becoming less reliable.

The cost of that efficiency is the ceiling effect described above, which is the trade-off the design accepts.

The same reasoning now applies to the TOEFL iBT, which adopted multistage adaptive Reading and Listening in January 2026 and shortened accordingly. Where the SAT routes once per subject, so does TOEFL — and the ceiling effect works the same way there.