The March 2026 SAT: the exam students were still arguing about a week later

Moderate. March produced roughly twice as many difficulty reports as reassurances, which by recent standards is unremarkable, and almost no complaints about timing. It was not the exam that broke anybody.

What it did produce, in unusual volume, was disagreement about specific answers. Students returned to individual vocabulary items again and again, laying out the case for a choice they had made and could not let go of, in some cases having gone looking for the source text afterwards to settle it.

The math was quieter but not easy, and it concentrated in the same place November had: geometry and measurement, with a strong showing from solids. The difference is that March hid constants inside those questions, so the shape was familiar and the task was not.

The vocabulary arguments March produced share a structure. A student picks a word that fits the general sense of the sentence, sees the keyed answer, and concludes the test was being arbitrary, because their word also fits the general sense.

It usually does. That is the design. A words-in-context question where only one option fits the general sense would be a vocabulary quiz, and the SAT stopped setting those years ago. The modern item gives you three or four options that are all plausible in a loose reading, then puts a structural signal in the sentence that eliminates all but one. The signal is usually a contrast marker, an appositive that restates the blank, or a clause explaining what the writer did instead.

So when you find yourself with two workable options, the correct move is not to weigh which is the nicer word. It is to go back and find the part of the sentence you skimmed. There is almost always a phrase you read past that makes one of the two impossible, and it is usually doing the least glamorous grammatical work in the sentence.

This is also why post-test arguments about these items are unwinnable and worth abandoning. Arguing from the general sense cannot resolve a question that was never about the general sense.

March’s hard math looked like standard measurement work and behaved like algebra. A cone whose volume is expressed as a multiple of pi, with the multiplier as the unknown. A rectangle whose area is given and whose dimensions are expressed through a constant. A right triangle whose two acute angles are given as multiples of the same unknown.

Each of these takes a shape students recognise and moves the unknown from the answer to the setup. You are not asked for a length, you are asked for the constant that makes the described configuration consistent. It is the same reversal that made October hard, transplanted into geometry.

The trigonometry item is the cleanest example. Two acute angles in a right triangle are complementary whatever the constant is, so the question is not really about the constant at all. It is about knowing that the sine of one acute angle equals the cosine of the other, and recognising that this holds regardless of the specific values.

Beyond the geometry, March ran a strong line in graph features: questions defining two functions and asking which displays some property, or where a described feature appears in the equation. These reward reading the form of an expression rather than evaluating it, and they punish students who graph first and think second, because the graph does not tell you which coefficient the question means.

The measurement questions rewarded one habit above all: identifying which length in the described solid is the one you actually have. A distance from the tip of a cone to the rim of its base is not the height, and a question that gives you both is testing whether you noticed.

The words March put under pressure were ordinary. What made the items hard was that each sentence contained a structural signal doing the deciding, and it was possible to answer without noticing it and still feel confident.

The recurring signal is a contrast between what someone assumed and what was actually the case. When a sentence says colleagues expected one thing but the person maintained another, the blank belongs to the expectation being denied, which is usually more extreme than the word a careless reader would supply. Getting these right is a matter of finding the pivot word and reading in the direction it points.

Original questions written around what this exam tested, one for each argument above. Nothing here is a real test question.

The instinct to argue about a vocabulary answer is worth converting into something useful. Every one of those arguments marks a question where you had the vocabulary and missed the structure, which is the most fixable kind of error on the exam and the one most likely to recur.

Do the same with the geometry. If a shape question felt strange, check whether the unknown had moved from the answer into the setup. That single reversal accounts for most of what made March’s math harder than it looked.