Almost everyone who learns about the spacing effect asks the same question next: okay, but how long should I wait? Most answers are a little embarrassing. Review tomorrow, then in three days, then a week. The numbers sound authoritative. They are usually folklore, repeated because they feel reasonable.

The truth is stranger and more useful. There is no best gap. There is a best gap for a particular goal, and the goal is a date: the day you will need to remember.

The experiment that mapped the problem

In 2008, Nicholas Cepeda and colleagues published one of the largest studies of spacing ever run, in Psychological Science. More than 1,300 people learned obscure trivia facts, such as the name of the lowest-ranked player on a team or the answer to an odd question. They then reviewed those facts after a gap that ranged from none at all to roughly three and a half months. Finally they were tested after a delay that ranged from about a week to nearly a year.

The design is what made it valuable. Earlier studies usually tried one or two gaps and one test date. This one crossed many gaps with many test dates, so the researchers could see the whole landscape. The shape they found has been called a temporal ridgeline. For any test date, recall rises as the gap grows, peaks, and then slowly declines. And the peak moves.

If the test was a week away, the best gap was short, about a day. If the test was a year away, the best gap was weeks long. The longer you need to remember something, the longer you should wait before reviewing it.

The ratio that shrinks

The second finding is subtler. As the test date moves further out, the best gap grows, but it grows more slowly than the delay itself. By the authors' account, the ideal gap was roughly 20 to 40 percent of the delay when the test was about a week off. When the test was a year off, it fell to something like 5 to 10 percent.

These are rough ratios, not a law. They came from trivia facts and adult volunteers, and your material may behave differently. But the pattern points toward a useful rule of thumb. Space your reviews out in proportion to how long you need the memory to last, and accept that for long horizons, the proportion gets smaller.

It also explains why expanding schedules, with gaps that stretch from days to weeks to months, feel right. Each successful review pushes the next test date further away. So the best next gap gets longer.

Why waiting works at all

The mechanism is not that time itself strengthens memory. Waiting works because it makes the next retrieval harder, and retrieval effort is what builds the memory.

Robert Bjork's distinction between storage strength and retrieval strength helps here. Right after you review something, retrieval strength is high. You can recall it effortlessly, but that ease is misleading, because little new learning happens when recall is trivial. As retrieval strength fades, a later retrieval takes real effort, and successful effortful retrieval is what increases storage strength.

So there is a window. Review too soon, and the memory is still sitting on the surface. The review feels fluent and does little. Review after the memory has fully gone, and you are relearning from scratch, with no retrieval to strengthen. The sweet spot is the moment the memory is just hard to reach but still reachable.

The asymmetry nobody mentions

Here is the part worth tattooing on your study habits. The ridgeline is not symmetrical.

In Cepeda's data, going slightly past the optimal gap cost you only a little. Falling short of it cost you more. The peak drops steeply on the short side and tapers gently on the long side. This pattern has been noted in follow-up work too.

The practical conclusion is plain. When you are unsure, wait a bit longer. An overlong gap risks a failed recall, which you can recover from and which still teaches you something. An overshort gap gives you a pleasant feeling of mastery while you quietly collect less learning per minute.

This cuts against instinct. Reviewing often feels diligent. Letting a card sit until you are not sure you'll get it feels reckless. The research says the second feeling is closer to the truth.

Your target is not 100 percent

There is a reason modern schedulers don't aim for perfect recall. If you only see a card when you're nearly certain to get it right, you review far more often than you need to. If you wait until you'll probably miss it, you spend your time relearning.

Spaced repetition algorithms deal with this by estimating, for every item, the probability you could recall it right now, then scheduling the review for when that probability falls to a chosen threshold. Many people settle on around 90 percent. FSRS, the open scheduling model used by modern flashcard tools, does this by tracking for each card how stable the memory is and how retrievable it is at any moment. A lower target gives you fewer reviews and more misses. A higher target gives you more reviews and fewer misses. The setting is a trade, not a virtue.

That is the same trade the ridgeline describes, applied one card at a time. Different facts decay at different rates, so the best gap for one card is the wrong gap for another. Hand-built schedules can't capture that. Only per-item estimates can.

What this means if you're studying for a date

You don't need software to benefit from the idea. You need to ask the right question. Not "how often should I review?" but "when do I need to know this, and when will it be slightly hard to recall?"

A student with an exam in three weeks should space material across days, not hours. Someone learning a language for a trip next summer should space across weeks and months. A professional who needs a body of knowledge for years should expect the gaps to stretch much longer as the material becomes familiar.

Notice also what the ridgeline does not say. It does not say cramming is useless for a test tomorrow. If the test is a day away, the best gap is minutes to hours. It says that cramming optimizes for a short horizon, and that the same effort spread out buys a far longer one.

Your next moves

  • Write down the date. For each thing you're learning, name the day you'll need it, whether that's an exam, a trip, or just "this year." Everything else depends on that horizon.
  • Use the proportion as a starting guess. For a test about a week away, try your first review roughly a day or two after learning. For a test months away, try gaps of a week or more. Adjust from how recall actually feels.
  • Lean long when unsure. Before re-reviewing something you recall easily, wait one more session. A miss is information, and a too-early review mostly isn't.
  • Notice effort, not comfort. If a review feels effortless, you probably came too soon. If you're guessing blind, you came a bit late. Aim for the small struggle in between.
  • Test the retrieval, not the recognition. Cover the answer and produce it from memory. A spaced gap only helps if the review is an actual attempt to recall.

Where this leaves a flashcard app

The hardest part of everything above is that the right gap differs for every card and shifts every time you review. Nobody can track that by hand. That's the job we built Recall to take over: it uses FSRS to estimate how well you still hold each card and shows it to you around the point where recall gets difficult but not lost, all offline, with your Anki and Quizlet decks brought along. If you'd like to see what a well-timed gap feels like, you can try it at recall.lumenlabs.works. But the idea works with a notebook and a calendar too. Pick the date you need the memory, and wait a little longer than feels comfortable.