Leitner system
Leitner system (aka the Leitner box system) is a method of prioritizing flashcards in learning. Well known cards are shunted to boxes corresponding with higher memory stability.
Leitner system is often incorrectly labelled as a spaced repetition system. Spaced repetition is computational in nature. Without an effort to compute optimum intervals, prioritized review is little more than an inefficient form of scheduling that may only slightly reduce the cost of high retention as compared with, for example, plain reading where frequency may correlate with importance. Moreover, retention itself is hard to predict. In other words, Leitner is a hit-or-miss system.
There are many variants of the Leitner system. In flashcard applications, the original system of boxes has mutated into variants that can simulate a rudimentary spaced repetition approach. If there are many boxes labelled with specific intervals, they can be equivalent to SuperMemo on paper (see: Algorithm SM-0). The Leitner system has been used in many older applications (e.g. VTrain). Many modern web applications may start off from implementing the Leitner system at early stages (e.g. Duolingo).
I use the term Normalized Leitner to refer to a software implementation with adjustments that can turn Leitner into a spaced repetition system, and make it possible to employ comparison metrics (e.g. to compare the Leitner system with SuperMemo algorithms). This is how Normalized Leitner works:
- boxes are associated with intervals
- first interval is set to Int1, and successive intervals are set to Int1*power(E-Factor,repetition)
- failure results in reversal to box #1 (violating this principle worsens the performance of the algorithm)
- target recall at review claim is set to 90%
In the simplest case, normalization would use Int1=1, and E-Factor=2. This would associate boxes with intervals: 1, 2, 4, 8, and 16 days (i.e. as in the schedule suggested in C.A. Mace book of 1932). In addition to normalization, the author of an implementation should make a claim on the desired level of retention. For example, if a standard normalized Mace-like Leiter system claims recall at review of 90%, its metric may differ from the same system that attempts to achieve 85% recall. Target recall claim has no effect on learning. It only affects the comparison metric. In SuperMemo, the algorithm adjusts intervals to the target level of retention. Those adjustments may change the degree of SuperMemos advantage at different levels of the forgetting index.
In SuperMemo 17, it is possible to compare the advantage of Algorithm SM-17 over the Normalized Leitner system. See: Universal metric for cross-comparison of spaced repetition algorithms.
This glossary entry is used to explain "History of spaced repetition" by Piotr Wozniak (June 2018)
For more see: Leitner system @Wikipedia
Figure: An incorrect mutation of the Leitner system where failed answers are moved back by one box only (source: Wikipedia). This variant was in use in Duolingo for a while