Spaced Repetition Glossary

Spaced repetition has a bit of a learning curve of its own, but too much of it comes from not understanding a few relatively straightforward terms and concepts. Many authors make matters worse by throwing these around without explaining them thoroughly. I’ll try to avoid getting too technical in any of my posts, but this glossary should give you somewhere to look if you get confused, as well as offer a nice overview of the most important concepts when you’re starting out.

There are three sections; the contents of each are presented in alphabetical order. Terms that have their own entries in the glossary are italicized within definitions at least once per definition. This glossary is Anki-centric since that’s the tool my series centers around, but you’ll see many of these terms, or similar ones, used elsewhere too.

General terms

Cloze deletion

The process of creating cards by hiding some portion of a sentence or image, producing a “fill in the blank” type of card. Cloze deletion cards are favored by many users because they’re often faster and easier to create than other types of cards.

Learning

Many SRS’s, including Anki, have a “learning” mode, as opposed to the standard “review” mode, to help you initially acquire new material or to relearn lapsed material. When material is deemed “learned” by finishing this process, it is said to have graduated and is scheduled for its first review.

New

Cards that you have never studied before, in learning or in review, are called new cards.

Occlusion

The portion of a cloze deletion that is hidden.

Review

The act of studying cards that are due and providing ratings for them. Reviewing a card stops the natural process of forgetting it, strengthens your memory, and reschedules it in your SRS for its next review based on your performance.

One repetition of a particular card may also be called a “review”, e.g., “I did 200 reviews today.”

SRS

Spaced-Repetition System, the general term for any application that implements spaced-repetition algorithms.

Scheduling

The process by which an SRS guesses when you will next forget a card and presents it again for review shortly before that time.

Units of information

Card

A unit of information studied in an SRS analogous to a physical flashcard, with a front (prompting you with something to remember) and a back (explaining what you should have remembered). Each card is scheduled for review separately based on your past performance on that card.

Collection

All of the flashcards you have controlled by an SRS, particularly Anki.

Deck

A subdivision of a collection used in most SRS’s. Decks allow you to organize information or separate topics that might be confusing if reviewed together, e.g., vocabulary for two different languages. (Note, however, that many people say it’s more effective and offers benefits to creativity to review all topics together to whatever extent is practical.)

Note

In Anki, a note is a set of fields of information that is used to generate one or more cards. In many cases, a note generates exactly one card, but it can also generate more than one. For instance, you might want to study some cards, like foreign-language vocabulary, forwards as well as backwards, so you have to produce the associated word both in the foreign language and your native language.

Tag

In addition to decks, many SRS’s support tagging cards to allow for additional organization. (To be precise, in Anki you tag notes, not cards.) In Anki, you can use tags to find or manage cards or to temporarily select cards matching some criteria to study from multiple decks at once.

Scheduling

Bury

In Anki, burying a card hides it until the next day or until you choose to unbury all cards. You can manually bury a card if you can’t review it at the moment (for instance, if you’re studying in a public place without headphones and the card requires sound), if you accidentally flip the card and see the answer, etc. Anki also automatically buries cards that are closely related to other cards under some circumstances, so that you don’t see them immediately after one another and have an artificially easy time answering the question.

Due

A card is due for review when your SRS thinks you’re about to forget it. A card is due when its last review date plus its interval is equal to the current date.

Ease

A number describing how difficult a card is, often expressed as a percentage. Typical eases range from 130% to 330% with an average around 250%; more details can be found in the From the Ground Up post. The ease is used to calculate how much larger the next interval should be when you review the card.

Fail

The act of rating a card as forgotten, causing it to lapse. The connotations of the term are unfortunate, because failures are part of the system by design and don’t mean anything bad or even unwanted happened; see lapse for more.

Graduation

When a card leaves learning mode and enters review mode, it is said to have graduated.

Interval

The amount of time between reviews of a card, typically expressed in days, months, or years. A card’s interval is recalculated every time you review it, based on its previous interval and its rating at that review time. Intervals typically grow over time, unless you forget the card, at which point the interval is zeroed out or reduced substantially.

Lapse

When you forget a card you previously knew, it is said to have lapsed. Lapses should not be considered failures and do not mean you did anything wrong: forgetting and therefore spaced-repetition algorithms are probabilistic, so the algorithm will necessarily put the due date too late on occasion. In fact, the algorithm is tuned to produce a certain percentage of lapses, usually 10% by default. You can adjust the parameters of most good SRS systems to reduce the percentage if you wish, but this comes with diminishing returns. See retention.

Leech

An unusually difficult card that takes up an unreasonable amount of your study time. It’s important to take appropriate action when your SRS identifies a leech, or you will spend a large fraction of your study time on a tiny handful of cards.

Mature

A card with an interval of at least 21 days. Mature cards generally have the highest retention percentage, as they are least subject to small fluctuations in your memory or review time. You can think of a mature card as one that you’ve fully learned and are now merely focused on maintaining. See also young.

Overdue

A card which has a due date in the past. In other words, you theoretically should have reviewed it already, but did not. We’re all human, so this will happen from time to time, but allowing cards to become overdue substantially increases your chance of forgetting them, so is to be avoided whenever possible.

Rating

A subjective description of how difficult a card was to remember, provided by the user after looking at each card. Most SRS’s use two (forgot/remembered), four (forgot/hard/good/easy) or six (totally forgot/glimmer of recognition/almost remembered/hard/good/easy) ratings. Anki uses four. The easier you rate a card, the longer it will be before you see it again and the faster its interval will grow in the future.

Many SRS’s sometimes represent ratings as numbers, starting with the most “forgotten” rating as 0 or 1. For instance, Anki labels Again as 1, so if a rating is shown as 2, that is equivalent to a “hard” rating.

Reschedule

The act of manually changing the next due date of a card. You might want to do this if you have information the SRS doesn’t (for instance, you just imported a bunch of material you already know very well), or if you change a card so completely it doesn’t make sense to allow it to maintain its review history.

Retention

A measure of the percentage of material you remember at the time it is due; in other words, the percentage of reviews which are not lapses. Higher percentages mean you remember more material both at review time and at any moment you’re called upon to produce it, but they also mean you have to spend significantly more time studying.

Most SRS’s aim for about 90% retention with their default settings, but a good SRS will let you tweak the parameters in an attempt to change this percentage, if you’re so inclined. Maximum learning efficiency in terms of material remembered per unit of time spent comes around 85% retention. Aiming to increase the percentage beyond 90% may make sense for extremely critical information, but doing so for all your cards is a bad idea, as it leads to rapidly diminishing returns: aiming for 99% retention, for instance, would likely result in having to review all of your cards almost every day.

Keep in mind that the retention figure is not the percentage of your material you remember at a given moment in time – that’s significantly higher. To understand why, notice that because your memory decays monotonically over time until you review (i.e., under normal circumstances you won’t become more likely to remember a card by waiting longer to review it), the due date is the moment you’re least likely to remember the information, provided that you don’t allow cards to become significantly overdue. It follows that if you pick a card from your collection at random, chances are it’s before its due date and you’re more likely to remember it than if it were currently due. But retention is always calculated when your card is due or overdue, so your retention percentage will underestimate your chance of knowing the card at the point of this random test. The math suggests that if your overall retention is 90%, it turns out that you typically know about 95% of your material.

Suspend

Suspended cards never appear for review even when they are due. You might want to suspend material if you aren’t ready to study it yet, or you aren’t willing to spend the time on it right now but there’s a chance it might become much more important to you later, or you suspect it’s incorrect or not worth learning but don’t have the time to deal with it yet. Anki automatically suspends leeches under many circumstances.

Young

A card that has graduated from learning but hasn’t yet become mature.