When he described what the mistakes actually felt like from the inside, he used a phrase I have now heard from probably two hundred students: he was debating over the nuances between the answers.
That sentence is the whole diagnosis. When you sit there weighing B against D, you are asking which one is better. Better is a matter of taste, and there is no taste in the room to appeal to. The question has no ground truth in it. The ground truth is forty words up the page, in the passage, which you have already stopped looking at.
Read it like it's going to be on a test. Oh, wait.
Here is what I think is actually going on, and it took me an embarrassingly long time to see it. Strong scorers and stuck scorers read the same passage and take away different amounts, but not because one of them is smarter. It is because one of them is reading the passage and the other is reading to get to the questions. If the passage is just the thing standing between you and the item, you will skim it for gist, arrive at the choices with a vague summary in your head, and then be forced to pick whichever option sounds most like your vague summary. Two of them always will.
The test is built out of small words
I went and counted this, because I wanted to know whether I was just repeating something I had heard. Across 444 real College Board reading passages, about half carry a hedge like may, suggests, or proposed. About half carry a scope word like most, only, largely, all but one. Half carry an attribution: researchers suggest, conventional wisdom held. Just over half turn somewhere, on a however or an although.
Nothing is in all of them. That is the finding, and it is also why my student's errors looked random to him. They looked random because the passages are not all the same shape. He was hunting for one pattern in a pile of things that do not share one.
Take a sentence like this, from a passage we wrote for the tool:
For decades, ecologists attributed the decline of eelgrass meadows almost entirely to nutrient runoff.
There are two words in there doing all the work, and neither of them is eelgrass. Attributed tells you this is what a field believed, not what the author is claiming. Almost tells you runoff was most of the story and not all of it. Miss those two and you will happily select an answer choice that says runoff played no role, because it will feel close enough to what you remember reading.
Corvus: every wrong choice is wrong at a specific word. Find it and stop arguing.
That is what wrong answers are made of, and they are rarely outright lies. A wrong choice usually takes something the passage genuinely said and quietly upgrades it: a may becomes a proves, a most becomes an every, something a researcher cautiously proposed becomes something that was demonstrated. If you never registered the small word going in, you have no way to catch the swap coming out.
So we built the boring version
Close Read is a passage and a Next button. That is the entire tool.
You get a real SAT-length passage on the left. You press Next, and a phrase lights up with a note beside it explaining what that phrase is doing and why it matters. Then the next one. Nine steps, about two minutes, and the passage fills up with marks as you go, so what you are left at the end is an annotated page instead of a memory of one.
On the eelgrass passage, the notes go roughly like this. For decades, ecologists attributed: this is what the field used to think, not what is true. Almost entirely: almost, not entirely, hang onto that. That explanation, however: here is the turn, and most passages turn exactly once. Such estuaries: not 42 estuaries, 42 of those estuaries, the strange ones. Cautions: she is hedging her own finding, which is a much smaller claim than saying the crabs caused it.
None of that is clever. It is what a decent tutor mutters while reading over your shoulder, which is a thing almost no student has ever actually had done for them.
Why there are no questions in it
This was the argument I had with myself for a week, and I want to be straight about it, because the version with questions is the version that is easier to sell.
Judge Strix: the court notes that you have already done four thousand questions.
Everything in SAT prep is already a quiz. You have done thousands of questions. If questions were going to teach you to read closely, they would have by now. What a question does is test whether you noticed the small word, and if you did not notice it, a question cannot show you what noticing would have felt like. It can only tell you that you got it wrong, which you already suspected.
So Close Read does the one thing the question format structurally cannot: it shows you the read. You watch someone do it, at normal speed, with the reasoning said out loud. That is not a new idea, incidentally. It is how apprenticeships have always worked and how writing gets taught everywhere except test prep, where somewhere along the way we decided the only respectable teaching tool was another item with four options under it.
Watching is not practicing, and I am not going to pretend otherwise. Close Read builds the habit of noticing. It does not by itself build the reflex, and if you run through five of these and go straight back to shopping the answer choices, you will get exactly what you have been getting. The way it is meant to work is that the muttering becomes your own: you start hearing almost, not entirely in your head on a passage nobody is annotating for you. That takes doing a few and then deliberately trying it cold.
One more thing worth saying, since it is the least obvious part. Bluebook already hands you a highlighter and an annotation tool on test day. Every student gets them. Almost nobody is ever taught what to put them on, so they end up highlighting whole sentences at random, or nothing at all. The marks in Close Read are the marks worth making on the real thing.
It is free, it takes two minutes, and there is nothing to get wrong. If reading has been the part of this test you cannot get a grip on, start there instead of buying another book of questions.
