Fool’s Flashcard Review

The vast majority of flashcard applications use cards as the fundamental unit of knowledge. If they record the user’s performance with a given flashcard in, say, the graded slideshow method, it is usually recorded for the card irrespective of the direction of study for that card. If the flashcard application uses a form of spaced repetition or interval study

Most flashcard developers realize however, that this approach has a certain disadvantage to it based upon a necessary assumption the developer makes about the user. Either the developer assumes the user is only interested in studying a card in one given direction (e.g. Russian to English) or they assume that one’s performance in either direction is roughly equivalent (if I know it in one direction Russian to English, then I probably know it in the other direction, English to Russian, roughly as well). This is, of course, rarely the case.¹ While some language learners are primarily interested in studying in one direction, most will want to have both passive recognition and active recall of a vocabulary word, being both able to immediately recognize its meaning when read, but also recall the word when thinking of the word in one’s own native language.

Most flashcard applications allow you to control the direction of study or reverse it. iFlash (review upcoming) allows you to reorder the two or three or more sides of a card shown or even randomize the direction of cards but in interval study it records performance for all directions together. Mental Case (see my review) allows you to designate the “note reversibility” for a card, and can thus prompt you in multiple directions for a card, but again this information is not recorded independently.

There are a number of methods for confronting this problem but the most elegant approach I have seen so far is taken by the most advanced software applications such as Anki (see my earlier review) and I believe that other flashcard developers would do well to consider emulating the approach.

Anki has chosen for its the fundamental unit of knowledge a ‘fact’ which consists of, for example, a word and its definition, and in the case of three sided Asian language cards, its pronunciation. The difference between this and the card approach is that a fact that a fact can be a mother to two or three cards each with their own intervals and directions. Anki stores the fundamental ‘fact’ information independently of the users performance in the application and when you enter information for the ‘fact’ the application can be configured to automatically add two or three cards associated with that card in the various directions that the user wants to study.

For example, if I’m studying Japanese or Chinese, I may want to be able to record, independently, my performance across time in interval study in all three of the following directions: 1) be shown the English and then guess the Kanji/Hanzi characters as well as their pronunciation (and tones in the case of Chinese) 2) be shown the pronunciation (and tones) and be able to guess the characters and meaning in English and 3) be shown the Kanji/Hanzi character and be able to guess the pronunciation and English meaning. In most cases I think learners of Chinese and Japanese will at least want to review in directions (1) and (3) so advanced flashcard applications like Anki allow you to automatically create cards to that effect when the facts have been inputted. These cards then progress along their own interval study schedule as you review them.

Because, however, Anki maintains a relationship between cards and facts, when the content of the original fact is changed, so too is the content of any cards associated with the fact. This is a big advantage over other applications which might be tempted to simply let the user create duplicate cards in other directions. In another advanced interval study focused application Memosyne (see my review), for example, you can create “Vice versa” cards which then tracks the performance on the reversed card separately. However, if you then go and correct or edit a card, it will not correct the corresponding reversed card. Anki does not suffer from this problem.

Given this fact to card relationship, however, it means there is a significant separation between applications which acknowledge the relationship and those which don’t which impacts the full exchangeability of data between them. If you export data from Anki you get cards, without the preserved link between cards that are born of the same fact.

I hope that flashcard application developers reflect on the benefits of this approach and the power it gives to users in maintaining large collections of flashcards that need to be memorized in multiple directions. Implementation of this kind of feature likely requires adding a layer of complexity to the way that the data is maintained, but I have become convinced that the advantages are truly significant, especially for those who want to engage in long-term study.

1. Though I do find that active mastery of a word, going from, say English to Russian, makes it far more likely that you know the reverse. [↩]