How does spaced repetition work?

The forgetting curve is the shape memory loss takes over time: very steep at first, then flattening. Hermann Ebbinghaus mapped it on himself in the 1880s using nonsense syllables, and the basic shape has survived a century and a half of replication with real material.

The practically important part is the steepness of the first day. Most of what you lose, you lose almost immediately, which means a review on day one is worth far more than the same review on day ten.

Each successful review flattens the curve for the next stretch. That is the whole principle behind expanding intervals: the memory decays more slowly after every retrieval, so the next reminder can wait longer.

The ideal interval is the longest one at which you can still recall the item, which by definition you cannot know in advance. In practice, systems approximate it by expanding the gap after each success and shrinking it after each failure.

A common manual schedule is one day, three days, a week, two weeks, a month, three months. It is not optimal and it is close enough that the difference rarely matters for anything other than exam preparation under time pressure.

Chasing the optimal algorithm is the most common way people waste time on this. The gap between a crude expanding schedule and a state-of-the-art one is small. The gap between any schedule and no schedule is enormous.

Spaced repetition works poorly for book highlights, because the technique depends on retrieval and grading, and a paragraph of prose offers neither. There is no question to answer and no honest way to mark yourself right or wrong.

What people actually want from their highlights is not recall but re-encounter: to be reminded that an idea exists, at a moment when it might apply to something. That benefit comes from spacing alone and needs no grading step, which is a much lower bar and a much more sustainable habit.

Tools that apply review scheduling to highlights tend to produce people clicking through cards without reading them, which is precisely the failure the system was meant to prevent. Simple resurfacing, where a passage appears somewhere you already look, gets the useful half without the ceremony. Second Read takes that approach deliberately, putting one passage on your new tab and never asking whether you remembered it.

Anki is the most capable and the most widely used, it is free on desktop and Android, and its scheduling is the best researched of any option. The interface is dated and the learning curve is real, and it remains the right answer for serious study.

Mochi and RemNote are the friendlier alternatives, both with free tiers, and both aimed at people who want notes and cards in the same place. Quizlet is the easiest to start with and the weakest scheduler.

For reading rather than studying, the category is different: Readwise sends a daily review of your highlights, and Second Read puts one on your new tab page. Neither grades you, because neither is trying to. That is the correct design for prose, though it means neither is a substitute for Anki when you genuinely need to memorise.

Yes. The Leitner system does it with paper cards and a box of five compartments: a card you get right moves one compartment back, a card you get wrong returns to the first, and you review the front compartments more often than the back ones. It predates the software by decades and works.

A calendar and a notebook will also do. Write the material, then schedule four reminders at expanding intervals and answer from memory before checking.

The reason software won is bookkeeping rather than principle. Once you have a few hundred items, tracking which are due is the part that collapses, and that is all the algorithm is really buying you.