I have first used a spaced repetition program in the summer of 2003. It was a demo version of a commercial program which had taught me the German words for “juice”, “hot” and “living-room”. It took me a few more years to discover Damien Elmes’s magnificent (and free) Anki, and a bit longer to learn to use it correctly/ effectively.

I’m a big fan of using Anki for learning, and I’m interested from both a theoretical and a practical point of view in the emergent understanding arising from massive “trivial learning” of front-back flash cards associations.

Hereby are the uses I put Anki into, followed by prospective/ theoretical uses and uses I wish to put it into one day.

I. Language learning.

I have been using Anki to learn German, studying it by starting from elemental and advancing into more complex components, namely, beginning with words and working up to memorizing sentences.

II. Anatomy learning.

Prior to arriving to Berlin for my neuroscience studies, I have commenced in anatomical studying of the brain and the nervous system. It turned out to be completely useless as far as my master program went, but it was very effective for what i was trying to do, and it was nice later when reading papers that mention the cingulate or the fusiform gyrus to know immediately which regions of the brain it was.

I have used anatomy charts (pictures) whose text labels I had occluded. Luckily enough, someone had already created an anki add-on that made the process of creating such cards out of picture files a very quick process. In many cases I had several cards for the same region with different pictures, for example one with a schema and one or more with photographs.

I’d say that anki (or other spaced repetition methods) is very suitable for this task.

Also, while it is not strictly or necessarily anatomical data, I have used anki to memorize various scientific tidbits, for example various factoids about C. elegans while I was doing an internship in a lab that studied this worm. Perhaps this is not the most relevant place to make this remark, but I’ll stick it here anyway: supposedly there are critics of rote memorization which deem it to be antagonistic to understanding. The way I see it, an understanding emerges when one is able to simultaneously hold several related and mutually-enlightening facts. Anki won’t help one, by itself, to understand why “front side” -> “back side”, but with a sufficient amount of such well reviewed cards one could understand a phenomenon on some higher meta level. Therefore the way to go could be just to memorize the “axioms”, facts whose “understanding” is beyond the project at hand, such as that “Katze” is “cat” if one studies to speak German, or Ohm’s law if someone is studying electrical engineering.

III. Keyboard layout learning.

I have used Anki to help me transition from typing on a QWERTY layout to Colemak, and I’m currently using it to help me with learning to stenotype.

IV. Equations

I haven’t met someone like this in person, but I have heard that in Germany many students use (paper) flash cards to rehearse and memorise material for classes. I was no different, except that I have used anki for this.

I have used Anki only ad-hoc before exams, cramming the material if you will. I had many a semester-beginnings planned to continuously add to course-dedicated decks, but I never carried on with it. As far as cramming a week before the exam, I have rather successfully memorized equations. The way I’d do it is have some description of an equation on the front side, and the equation itself (Anki support Latex ––– very convenient for this purpose) on the back side. During reviewing I’ll write down the equation on a notebook and then compare it with the back side.

Beyond this test-oriented application, I believe that a dedication to such a deck can benefit understanding. If one can recall an equation without looking it up, one can conjure, inspect and think about it while reading other material which might shine light on why so and so factor is in the equation.

V. “Art History”

Another university-course oriented (and test oriented) use I put anki into is with an art history class. Of course there is more into art history than that, but I have used it to memorize the artist, year, place and medium for any art work that was “on the exam”. The professor asked us to know only the decade, but with anki it was easy to get the year as well. If I recall correctly, for all the exams (2?) I had for that course, I spent roughly 40 minutes everyday for two or three days prior to the exam, and recalled correctly those details for pretty much every artwork I stumbled upon on the exam.

Here, too, the usage was ad hoc for the exam, but I imagine that if someone is studying art history, it might be adventurous to continuously rehearse a single art-works deck and build up this trivial knowledge.

Envisioned projects/ experiments

I. spatial learning

One experiment I was interested in running for a long time involved learning a “map” through the memorization of simple front-back cards. My interest is two-fold. First, it has a personal practical component: my sense of spatial orientation is absolutely dismal. I thought that backing it with some anking could topically help. Second, I was curious whether some sort of emergent “cognitive map” would emerge from this process. Spatial and semantic memories are handled with very differently with the brain, so prospectively it doesn’t seem particularly hopeful from this perspective. But one never knows until someone tries!

The more “ambitious” (or insane) project is of memorizing street layout, but a simpler and more tractable project could be the memorization of the layout of the metro/train system. Would I be able to “calculate” metro-routes from arbitrary stations in the city without google maps? It could look like this (from macro to micro):

First, have a card for each metro line, containing information about the two ends stops and the general direction of the line. For example (front side/ back side): “U1 / Uhlandstr. west-east Warschauer Str.” This is for the purpose of putting the lines in the context of the geography of the city.

Next, one can make a card for each station which prompts for the adjacent stations, e.g.,”Uhlandstr. (U1) / End-station – Unlandstr. – Wittenbergplatz”

“Wittenbergplatz (U1) / Unlandstr. – Wittenbergplatz (U2, U3) – Nollendorfplatz (U3, U4)”

“Wittenbergplatz (U2) / Zoologischer Garten (S5, S7) – Wittenbergplatz – Nollendorfplatz (U4, U4)”

Where the line numbers in the parenthesises indicate other lines that stop in the station. Alternatively, to relieve some of the amount of information learned by these cards, one can have another set of cards that use a station as a prompt and ask for the lines passing through it, e.g., “Wittenbergpatz / U1, U2, U3”.

It might be helpful to have a convention for the “unilateral direction” of lines, for example “Northwest to Southeast” so that one knows how the order of stations in the back side corresponds to spatial directions (as long as one knows already the “general”, i.e. “horizontal” or “vertical”, direction of the given line). For train lines it might be useful to also include distance in minutes between stations, info which is at least available on train station charts in Berlin. For example (made up data):

“Wittenbergplats (U2) / Zoologischer Garten (S5, S7) –4– Wittenbergplats –2– Nollendorfplatz (U4, U4)”

The same thing can be done with a street map as well, though here we’d be overwhelmed by the amount of data, as the number of streets in a city or even a single region greatly outnumber the number of metro lines. But it could be theoretically done by treating streets as lines: “Mollstraße / | Torstraße x Prenzlauer Alle (N) Karl-Liebknecht-Str. (S) west-east Platz der Vereinten Nationen”. In this case, Mollstraße begins where Torstraße changes its name to it (indicated by the “parallel” “|”) at the crossing with a street with a different name on each side of the crossing: “Prenzlauer Alle” on the north and “Karl-Liebknecht-Str.” on the south, and this street ends at a square/crossing called “Platz der Vereinten Nationen”. Perhaps a better “syntax” can be used; admittedly this one is a bit opaque, but perhaps with practice one gets used to it.

Then, one can treat street intersections as stations. So, for example, if we want to query which streets intersect “Ritterstraße” on both sides of it’s intersection with “Alte Jakobstraße”, we can have a card like this one: “Alte Jakobstraße (Ritterstraße) / Lindenstraße – Alte Jakobstraße – Alexanderinenstraße”.

Next, one can pull photos from google map’s streetview and create “bidirectional” cards for specific intersections, for example: ” {photo/s} / Ritterstraße x Alexanderinenstraße” and test both whether one recognizes the place from the picture/s, and if one can recall the photo/s of the place when prompted.

This is obviously(?) a great overkill. It’s probably easier (definitely more pleasurable) to learn a layout of a city by walking or biking through it, and I imagine someone going through something like this only in very very special circumstances (I’d love to know if anyone has done something like this!) ––– but here it is, for theory’s sake. I’d be curious perhaps to pull this off before going to a small locality I’ve never been to before, and see how this might effect my experience upon arrival.

II. Sign language

For a while now I was interested in learning a sign language. If I would it would probably be DGS (German sign language) as I’m residing currently in Germany. Until this very moment, I have regarded the main challenge being in the ankinization (creating the deck) rather than in rehearsing it. But I’ve discovered now that there’s a video-dictionary (also for many other sign languages) at spreadthesign.com8 in what seems to be a scrape-friendly layout. Since one can simply drag and drop video into the card-creating-window of anki, the process of ankinization should be rather swift. The videos are around 140kb each, meaning the media data storage of 1000 cards would be around 140mg ––– far from catastrophic.

As I have solved my perceived issues with anking a sign language in the course of writing this paragraph, I’m excited to try it out.

III. Programming

I have been interested in the capability of anki assisting in learning specific programming languages. Beginners as well as well-seasoned programmers spend some time looking up functions while programming. It seems to me like one can gain some advantage by anking subroutines, so that one spends less time looking them up online9 and possibly make them more “mentally available” such that one easily recalls the existence of the subroutines at moments when they could be handy. One can add subroutines as one encounters them ––– if one looks a subroutine online, the extra step of copying the text into the add-card window is rather minuscule so long as one keeps having to do that in mind.

I have tried doing this only once during a single day, learning SQL. I haven’t put this knowledge into “practice” yet (unless one counts the online practices of an online guide to be practice) so I cannot really comment yet on how useful this might be, but I imagine that one day I might try this out. My guess is that such decks would be relatively thin and therefore easy to maintain (i.e., dedicatedly rehearse).

IV. Games

Another project I had in mind in the past was to use Anki to gain skill in a game. Two scenarios I had in mind were with the games Go and Starcrft. There are rather different, the former being a turned based board game, the latter a real time strategy computer game. In either case my musings would probably remain theoretical, at least as far as I’m concerned, as despite thinking how much fun I’ll have playing something, it never seems like the right time to do it.

I’ll start with the latter, which by being more complex renders the ankization simpler. Possible card sets can be be for different build orders, statistics of units (costs, building speed, firing range) and so on –––- basically any information that can be put on a text-guide for the game can be distilled, I think, into an anki deck. One can train perception by having cropped screenshots of the game as the questions, the answers being the number of units in there e.g., “9 hydralisks, 5 mutalisks”, “7 zealots, 4 siege tanks “, “17 SCVs”. For this to work one would have to have a rather large and varied set of such cards. An additional set can be used to memorize which side wins (and with how many survivors) when a certain amount of units engages in a simple frontal fight with another group, without any micro being employed e.g, “5 marines and 3 firebats vs 18 zerglings” as the question and “3 marines survive” as the answer (this example is not based on real data).

All of these cannot supplant actual practice. However, as there’s a general progression in skill as a player plays, I think one can use Anki to learn some “basics” and “advanced basics” faster than learning the same things by repeatedly playing matches. With some stuff already covered by the anki deck, a player can “start” playing at a higher level, learning higher-level aspects of the game instead of worrying about not knowing some trivial facts about the game.

Go required a very different approach. It is turn based, and all of the components of the game are visually in front of both playears all at once. There are few possibilities for question/answer cards; for example learning some concepts and training “reading” groups, that is, determining whether a group is dead or alive. Beside this, one can use “indexing cards”, cards that do not contain the answers themselves, but only refer to them (or not at all). In this case Anki only takes care of the scheduling of rehearsal.

Such indexing cards can be of Go problems, for example. These can contain as the question-side a diagram of the board, and the learner can copy the situation onto a board and try and solve it. Alternatively, there are pages with interactive go problems, so one is relieved of the task of setting up the board; a question card, therefore, could simply contain a link to such a page (and thereafter be dependent on the availability of the page. Perhaps it’s better to back the link up with a diagram of the board). If one employes such a learning scheme, I’d suggest, if possible, to learn all four 90-degrees rotations of every problem, to slightly mitigate issues of “overfitting” (i.e. memorizing the answer moves without the rationale behind them).

Another thing that can be done is to have a certain succesion of moves (perhaps again a link must be used for the question) from the middle of a game between professional players as question that prompts for the rationale of the last placed piece. Here the learning is particularly “fuzzy”: we want the learner to have a feel for certain kind of moves and thinking that perhaps, otherwise, would take many many played games to learn.

All of these are meant to enable a person to play more sophisticated and interesting games at an earlier point. I am not a Go player so my imagination regarding how Anki can be applied here are limited by my ignorance.


9. Which can translate into a considerable time and concentration save if your interest, like me, tends to be piqued by the side “thumbnails” on stack overflow such as questions about the possibility of piracy based economies, or any questions asked by a group of people who treat RPG manuals as if they were the scripture.