Embedded Pretest Questions

Out of the 98 questions you see, 90 count toward your score and 8 are pretest questions, quietly inserted to collect data. The test maker needs to know how hard a new question actually is before using it for real, and the only way to find out is to put it in front of real test-takers and see how they do. These trial questions look exactly like the rest. Same formatting, same modules, same timer. Nothing marks them as practice, because marking them would change how people answer and ruin the data.

This is why hunting for the pretest question is a losing game. A student who hits a strange or brutally hard item and assumes it must be the unscored one is making a bad bet. Pretest items are scattered across every difficulty level, including easy and medium ones, and scored questions can be just as strange or hard as any pretest item. The only safe strategy is to treat every question as if it counts, because statistically, almost any single question you can point to probably does.

The payoff for the test maker comes later. Once a pretest question has been seen by a few hundred thousand students across many test dates, its percent-correct becomes a stable measure of difficulty. That number lets the question get calibrated and placed correctly in a future test's adaptive routing, where module difficulty shifts based on how the first module went. So the pretest system quietly feeds the same branching engine that decides what module two looks like for everyone else.