The August 2026 SAT: when the answer is that you can't tell

Harder in Reading, argued over in Math. Students describe Reading and Writing Module 2 as draining rather than tricky, with dense paragraphs back to back. Several say it cost them energy they needed later. If you felt fine and then faded, you weren’t alone.

Math divides people sharply. One group calls it easier than most practice tests. Another ran out of time and guessed on the last handful. Both groups describe the same turn, somewhere around question fifteen of Module 2, after which the items get noticeably harder.

If you came out unable to recognize the questions other people describe, that’s worth knowing about. It came up constantly in August and it says nothing about how you did.

A cluster of the hardest items had the same answer in different clothes. The information in front of you doesn’t determine the thing being asked about, and noticing that is the whole question.

The histogram question is the cleanest. Two samples of 27 values each, grouped into intervals ten wide, and you want the smallest possible difference between their means. Most people computed with interval midpoints and got a number near 4. But a bar tells you an interval, not a value, and within a ten-wide interval you can slide values around. The achievable totals overlap, so the means can be made equal and the smallest possible difference is zero.

The two-statement geometry question runs the same idea. You get one angle in each triangle and a pair of equal sides, then two statements to test. Neither is forced. One angle leaves the other two free to split the remainder however they like, and equal sides fix a ratio without fixing a length. The answer is neither, and you get there by trying to build a second figure that breaks each statement.

In each case a confident-looking number is available, and it’s the answer to a question nobody asked. Midpoints answer what the means would be if every value sat in the middle of its bar. That’s a different question, and it has a tidier answer, which is exactly why it feels right.

The same idea carried Reading and Writing, and students found it harder to see there. Several questions turned on a number that doesn’t measure the thing it appears to.

One paired two texts about how often studies confirm their starting hypotheses. The first author reads a gap between fields as a sign of looser method. The second doesn’t dispute the figures. He points out that one field publishes negative results and the other historically didn’t, so the rate reflects what editors accepted. The rate is accurate and measures the wrong thing.

A soil-chemistry passage did it with a sampling depth. Raising pH was supposed to cut metal uptake, and it didn’t. The measurements came from the top ten centimeters. If the roots draw from lower down, or if root-zone chemistry overrides the bulk soil, your pH reading describes soil the plant isn’t using.

Statistics led the section by volume, with geometry close behind. The items students remember are less about formulas than about what exactly is being counted.

The fastener question is the one people got wrong twice over. A rail 12x feet long takes a fastener every four sevenths of a foot, with one at each end. Dividing gives 21x, and that counts gaps. Fasteners run one ahead of gaps, so the answer is 21x plus 1. Invert the division and you get 48x over 7, which is the same error wearing a different face.

The sphere question asks you to hold two solids at once. Two spheres touch at a point, the larger has seven times the radius, and you know the smaller one’s volume. Work back to a radius of 6, scale to 42 for the larger, then add diameters rather than radii, because each named segment runs all the way across its sphere. That gives 84 plus 12, or 96.

Two more shapes recurred. A rectangle inscribed in a circle, where the diagonal is the diameter, and a diagonal twice the short side makes it a 30-60-90. And a pair of lines sharing a constant, with a point on each, where substituting both points and eliminating the constant leaves you a single relationship.

Vocabulary matched data questions for volume, and the hardest ones share a structure. A sentence concedes that some description would overstate things, then insists the effect was real anyway.

The most argued question ran exactly that way. It would be an exaggeration to call the effects blank, and then the losses turned out to be real and lasting. Students split between detrimental and negligible. Moderate effects can still be harmful, so calling them detrimental overstates nothing. Calling them negligible does, which is why the insistence that follows only makes sense with negligible in the blank.

Another described a cooling system for a data center. It uses less water in a year than one household, and it sits where the aquifer is already overdrawn. The sentence says the facility would do little to blank the shortage. Exacerbate and assuage point opposite ways. A near-zero user fails to make things worse, so exacerbate is the one the logic forces.

Beyond vocabulary, explanation questions carried the section. One asked why one group of apes showed a much stronger hand preference than another. The only stated difference was daily human contact, and the passage had already told you how right-handed people are. Another asked what follows when flower shape predicts visits and nectar predicts nothing. No relationship at all is stronger than a weak one, so any answer leaving nectar a secondary role overstates the finding.

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

August rewarded a question you can ask on any form. What does this information actually fix, and what does it leave open? A bar fixes an interval. One angle fixes one angle. A publication rate fixes what got published. Each of those leaves more room than it appears to, and the room is where the points were.

Practice it on questions you already answer correctly. Before you solve, write down one thing the given data does not determine. You’ll find that on most questions the answer is nothing, and on the hard ones you’ve just found the point of the question.