The SAT Is Digital and Adaptive Now. Here Is What That Changes for a Score

The SAT that students sit today is a different instrument from the one their older siblings sat. It is taken on a device rather than on paper, it runs closer to two hours than three, and — the part that actually changes how a score works — it adapts.

What multistage adaptive means

Each of the two sections is delivered in two modules. Everyone gets the same kind of first module. Performance on that first module then determines which version of the second module you receive: a harder set or an easier one. The exam is not adapting question by question — that is a different design — but stage by stage, which is why it is called multistage adaptive testing.

The practical effect is that two students can answer the same number of questions correctly and receive different scores, because they answered different questions. A correct answer on a hard item carries more information about ability than a correct answer on an easy one, and the scoring model uses that.

• Fewer questions are needed to reach the same measurement precision, which is why the test got shorter without becoming vaguer.
• Raw score conversion tables of the old kind do not apply, so “I got 45 right, what is that out of 1600” no longer has an answer.
• Skipping strategically is a worse idea than it used to be, because the first module determines what the second module can measure.
• The reported scale stayed at 400 to 1600 on purpose, so scores remain comparable with earlier years for admissions use.
• Section timing and question counts are published by the test maker; anyone quoting different figures is working from the old paper exam.

Why the scale did not change

Keeping 400 to 1600 was not conservatism. Admissions offices, scholarship rules, state reporting and years of institutional data are all built on that scale, and moving it would have broken comparisons the whole system depends on. The work of keeping a new format on an old scale is done through equating — statistical procedures that place scores from different forms onto a common metric so that a given score means the same thing across administrations.

That is ordinary practice in large-scale testing and it is also where the technical scrutiny goes. Equating is what makes a score from one form comparable with a score from another, and it is the thing a testing programme has to document publicly for that comparability to be checkable.

What it is not

The SAT is an achievement test. It is designed to measure what a student has learned in reading, writing and mathematics, against a curriculum. An IQ test is designed to estimate reasoning ability with as little dependence on taught content as its designers can manage. The two correlate — almost all cognitive measures correlate with each other to some degree — but they are built for different purposes, normed against different populations and validated against different outcomes.

This distinction gets blurred constantly, including by conversion tables that claim to turn an SAT score into an IQ. Those tables are not produced or endorsed by either the test maker or the publishers of intelligence scales, and the populations involved make the conversion unsound: SAT takers are a self-selected group of college applicants, not a representative sample of the general public, so a percentile in one group does not translate to a percentile in the other.

For students the change is mostly logistical: a shorter exam, on a screen, where the second half responds to the first. For anyone reading scores, the useful correction is smaller and more specific — stop thinking in raw right-answer counts, because the exam stopped scoring that way.

Common questions