Why you forget what your tutor taught, and the fix

Say your tutor spent 20 minutes on se me olvidó and why it isn't lo olvidé. You got it. You made a joke with it. The lesson ended and you felt like you'd earned the hour.

Nine days later it's gone, and what's worse is you don't know it's gone until you need it and produce a flat olvidé instead. The tutor gently corrects it. Again.

This is the most common complaint from people paying for lessons, and it gets misdiagnosed as a personal failing about as often as it gets diagnosed at all.

It isn't your memory

The rate at which people forget new vocabulary is steep and pretty consistent across learners. You're not unusually forgetful. You're on the same curve as everyone else, and the curve is brutal in the first 48 hours.

What separates people who keep their lessons from people who lose them is whether anything pulls the material back out of their head while it's fading. Nothing about that is aptitude. It's scheduling.

What the curve actually measures

Hermann Ebbinghaus mapped this in 1885 by memorizing thousands of nonsense syllables and testing himself at intervals, using himself as the only subject. His finding: recall drops fastest immediately after learning, then the decline flattens.

The obvious objection is that 1885 self-experiments don't replicate. This one does. Jaap Murre and Joeri Dros re-ran it in 2015 with a proper sample and reproduced the curve's shape. It's one of the oldest results in psychology that survived contact with modern methods, which is rarer than it sounds.

The practical read for a tutored learner: the steepest part of the loss happens before your next lesson. If you see your tutor weekly, the curve has already done most of its damage by the time you're back in the chair. Full detail on the mechanism is in the forgetting curve post.

Where a tutored lesson leaks

A lesson is dense. An hour with a good tutor can surface 20 or more items worth keeping: words, a correction to a verb you've been mangling, an idiom, the reason a preposition changed.

Each of those got exactly one exposure. You understood it in the moment, with your tutor's face and gestures and the sentence around it all propping the meaning up. Comprehension in context is not the same as being able to produce the word cold, 9 days later, with none of that scaffolding present.

So the leak isn't in the lesson. The lesson worked. The leak is in the 6 days afterward when nothing asks you to produce any of it.

Rereading your notes is the wrong move

Most people's fix is to reread the lesson notes. It feels productive and it's close to useless, which is an annoying combination.

Roediger and Karpicke ran the cleanest test of this in 2006. One group read a passage twice; the other read it once, then tried to recall it with nothing in front of them. After 5 minutes, the rereaders looked better. After a week, the recall group had them beaten by a wide margin.

The version that matters most for language learners came in Science two years later. Karpicke and Roediger had students learn 40 Swahili-English pairs, then varied what happened after the first correct answer. Items that kept being tested came back at about 80% a week later. Items that were only restudied came back at about 36%.

Karpicke & Roediger 2008 · Swahili vocabulary, one week later

0%

recalled when items kept being tested after the first correct answer

0%

recalled when items were only restudied after the first correct answer

Reading your notes is the 36% column. That's the whole problem with the tidy Google Doc: it's built out of the study behaviour that loses.

Students in that study couldn't feel the difference, either. Their own predictions were flat across all 4 conditions, which is worth sitting with if you've been judging your progress by how familiar the notes feel. More on that in the testing effect.

The four touches that hold a lesson

Here's the schedule I'd defend. It's built on 2 findings: the curve is steepest early, and spacing beats massing.

That second one has a big evidence base behind it. Cepeda and colleagues reviewed 317 experiments on distributed practice and found the same total study time produces better retention when it's spread out than when it's piled up.

Total: about 18 minutes across a week, against an hour of lesson you already paid for. The ratio is the argument.

The night touch isn't filler. Sleep after learning does measurable consolidation work, and the vocabulary you engaged with that day is what gets processed, which is covered in the sleep post.

One rule holds the whole thing together: produce before you check. Every touch has to start with you trying to pull the word out of nothing. The moment you read first, you're back in the 36% column.

Doing it by hand, honestly

You can run all of this with a notebook and no software, and it works. I want to be plain about that, because plenty of pages on this query exist to tell you the manual version is broken so you'll buy something. It isn't broken.

What it needs is for you to write the items down during or right after the lesson, and then actually reopen them 4 times on a schedule nobody's reminding you about. If you're the kind of person who does that, you don't need an app, and I'd rather say so than pretend otherwise.

The failure mode is specific and it isn't laziness. It's that week 3 has 3 lessons' worth of items in one document, and no signal about which of them are fading today. Manual review defaults to reviewing everything or reviewing nothing, and both lose to reviewing the 9 items that are actually about to go.

That's the one job software does better: it knows which items are due. FSRS models each item's decay separately and surfaces it near the point of forgetting. Whether that's worth paying for depends entirely on whether you kept doing the manual version past week 3.

If you'd rather have the routine on paper, the free handbook has the post-lesson debrief as a printable page, no purchase involved.

Questions people actually ask

---

Sources

Every load-bearing claim Verbamor makes is traced to its paper on the research page.