Every student who's ever failed a quiz has an error log of sorts: a pile of marked papers with red X's, a folder of wrong answers, the vague intention to "look at them later." It feels like accountability and it's a graveyard — the misses are recorded, filed, and never returned to. The red X documents the failure; it doesn't repair it.

A working error log converts each miss into a future attempt. For each wrong answer, you write four things: a new stem (a cousin of the missed question, not the same one), the trap in a few words (what the question was really testing), the correct move (what you should have done), and the next date (when it comes back). If a line can't produce a re-test, it's a diary entry, not a repair.

The pile of wrong quizzes is recognition: "I remember getting that wrong." The dated cousin stem is retrieval: "can I get this right now, on a new version?" The log works when last week's miss becomes this week's closed-book win — not when it's been carefully filed.

Why the pile of wrongs fails

Keeping the marked quizzes feels like confronting your mistakes — but it's confronting the record of them, not the mistakes themselves. The red X tells you where you failed; it doesn't tell you whether you can succeed now. Rereading the wrong answer is the same fluency problem as rereading anything: the material is right there, "I remember that" is easy, and nothing tests whether the miss is repaired.

Feedback plus delayed retrieval beats restudying the red mark — the miss is information about a gap, and the gap closes when you retrieve the correction, not when you re-read it. Naming the trap is the metacognitive piece: "defaulting to z" is a pattern you can catch next time; "wrong formula" is just a label for the failure. And the date is what turns the log from a museum into a schedule — a miss with a re-test date is a repair plan; a miss without one is a souvenir.

The miss-to-stem workflow

Use this week's misses. The marked quiz from the last few days — not a semester of accumulated failures. The recent misses are the active gaps; the old pile is a different problem.

Write the trap in a few words. Not "got it wrong" — what the question was actually testing that you missed. "Defaulting to z when σ is unknown" names the pattern; "wrong formula" just labels the symptom. The trap name is what you catch next time: "this looks like the z-default trap."

Write the correct move. What you should have done — "unknown σ and small n → t with df=n-1." The move is the correction; it's what the re-test will check. Not the full worked solution — the decision rule, the one that would've picked the right path.

Write a cousin stem, not the same question. Same trap, different numbers — "n=11, s=4.2, build a 95% interval and name the table." The cousin forces the move without letting you recopy the old answer; if you can solve the cousin, you've repaired the miss, not memorized the answer.

Set the date and return. 1/3/7-day spacing — the cousin comes back as a closed-book question on the scheduled day. The date is what makes it a log instead of a diary: a miss that returns is a repair in progress; a miss that's filed is a failure remembered.

A 40-minute session on the clock

The weak log line is the diary entry — "wrong formula, remember t" — which names the failure but produces no repair. It's a note to self that can't be tested. The strong line names the trap (defaulting to z), states the move (unknown σ + small n → t), and generates the cousin (new numbers, same decision). That line can come back on the scheduled date as a closed-book question — which is the whole point of logging it.

Why the date is the mechanism

An error log without dates is a list of regrets — each entry documents that you were wrong, none of them tests whether you'd be right now. The date is what converts the record into a retrieval: the cousin stem comes back, you attempt it closed-book, and the log closes the loop. Without the date, the miss is remembered; with it, the miss is repaired.

The spacing matters too. Testing the cousin the same day proves nothing — the correction is still fresh. Testing it at 1, 3, 7 days is the retrieval that consolidates: the miss that's still repaired after a week is the one that'll hold on the exam. The date isn't just scheduling; it's the memory mechanism.