SRS and flow

Lately I’ve come to a greater appreciation of the importance of ‘flow’ when your SRS deck grows to a certain size–around the point where it starts to take 35 minutes to an hour to run through. I don’t allow my daily repetitions to take more than an hour, but however much time you spend on them can become taxing if the flow of your deck is poor. This problem is obviously exacerbated as the number of reps increase

Once upon a time I kept a monolithic deck, in the fashion of the day. What I was being drilled on, on a card by card basis, could switch subjects and foci within subjects with every card. And it did.

For the longest time I thought this was virtuous, imagining it had the effect of keeping you on your toes, mentally speaking, and actively involved in the processing of each card. For pragmatic reasons I eventually decided to split the kanji drills out into a separate deck, since then I could do the non-writing portions at my leisure and break out a pad of paper and a pen only when required.

It was this change that made me realize that a monolithic deck is mentally taxing for no good reason. If your subjects are all mixed together and you can be served any card at any time, every time you are presented with a card your mind has to take a second or more to orient itself regarding the subject matter. I used to think this kept one engaged and actively thinking, but now I believe it simply wastes precious energy and time. If you know you are studying vocabulary or clozing sentences or producing kanji, you don’t have to spend a moment every time a card is produced figuring out what you are doing.

What happens when your deck is monolithic, realistically speaking, is that you start to get a flow going and then you are presented with a flow-breaking card like a clozed sentence or kanji, you come to hate these experiences. However, if these more-complex cards are grouped in a separate deck with their brothers, you not only know what you are doing but can enter a flow-state with them as well.

I find separation into similar themes also, paradoxically, increases engagement with individual cards. When I am studying clozed sentences I read the whole thing and think about them, but when they are thrown in with the vocabulary I treat them as the nuisance they are and dispose of them as expediently as possible, usually only reading enough to solve.

Using the better-flow principle I’ve recently seen an uptick in my accuracy, a reduction in time spent, and a more enjoyable SRS experience in general. I still keep cards in the same deck that ask different things, so long as they don’t break the flow of their individual decks.

Duolingo is pretty good

Yesterday I wrote a huge post on the subject of using an SRS system effectively.

I’ve recently been messing around with the program Duolingo to learn a bit of Spanish on a very casual level (which they consider ‘serious’, judging by their difficulty levels), and it very much resembles what I would have designed circa 2012 if you asked me how to construct a proper language learning system using modern techniques. It’s seemingly a product of the quiet, grassroots movement to reform foreign language instruction. So far it’s easy to use, though I think it coddles users a bit too much for my tastes.

Though I said in yesterday’s post that SRS doesn’t have a broad application, I was referring specifically to flashcards. An entire system designed from the ground up around SRS, such as the one Duolingo uses, is ideal. They went so far as to commission an independent study to see what time savings are realized from this approach, and it is expectedly significant. I wholly believe it, given the inefficiencies of traditional instruction. I’m also impressed with the depth of the program’s implementation of SRS, which is more modern and well designed than other programs which only partially implement an SRS principle.

I anticipate within the next decade all language learning will be assisted by some form of program using the spaced repetition principle. Traditional language instruction isn’t wholly broken (especially in the case of ‘easy’ languages like Spanish or German) but it is close. I predict the classroom portion of language instruction will be devoted mostly to speaking practice. Offloading the instruction portion onto a Duolingo style SRS system would enable a more efficient use of class time.

I went searching around the site to see if I could support their efforts, but there is as yet no way to do that. It’s a free service currently in search of a business model. The program itself lends well to a pay-per-unit model, and I wouldn’t be surprised if they transitioned partially or wholly into that at some point.

Rules for SRS success


If you want to learn something complicated, like a language, chances are there is a lot of memorization involved. This is especially true if you plan to learn Chinese or Japanese. Legend holds that hanzi/kanji was designed by a cartel of scribes specifically to be impenetrable to ordinary people, as a form of anti-competitive practice. I find this story so truthy that I don’t care to debunk it, even though it’s almost certainly an invention by understandably frustrated Westerners.

My first experiences with the magic of spaced repetition software (SRS) were with Supermemo, a program designed by a man who is for all intents and purposes a mad scientist. I won’t bother linking to the site now, because it resembles nothing like the program I used in the ancient times. Perhaps the creator finally cottoned on to marketing and usability.

The point being that the designer of Supermemo was more concerned with endlessly offering incremental improvements to his algorithm and going down strange memory rabbit holes like tracking sleep patterns than in offering improvements to the usability of the program itself. As a developmental path it reminds me of Dwarf Fortress: brilliant, and designed by a brilliant man, but offering an ever expanding and esoteric feature set without stopping to address usability. Supermemo was, among other things, possibly the ugliest program with the least elegant UI that I ever put up with.

I eventually abandoned Supermemo and moved to Anki, which offered something which seemed hugely seductive: the syncing of cards to a mobile version of the software, enabling a person to do flashcards whenever and wherever they were. Imagine that! I thought, surely, that was the answer to memorizing anything. You could memorize all sorts of things just standing in line at the grocery. What I hadn’t anticipated was the mental tyranny I was signing myself up for. Wherever I was, whenever I was, there was no escape from flashcards. Always I was asking myself: shouldn’t I be doing them right now? My experience with the system remained rocky, and I eventually gave it up after a period of years.

I’ve finally come about a mature understanding of the software, frankly, and to spare the fools like myself the pain I’ll summarize how to successfully use SRS to learn essentially anything. Eventually.

1. The Law Of The Worst Day

“The amount of effort you put into SRS on a daily basis should not exceed the amount of effort you are capable of on your worst day.”

When you begin SRS you will be filled with enthusiasm, as any student is when he sets out on a journey. This is natural and admirable, however, SRS will continue to make demands of you every single day. Imagine the effort you could put forth on your worst day–you just got home from a grueling 16 hour shift and had a fight with your girlfriend. You could spare 30 minutes, though, to complete your SRS, couldn’t you? You could at least run through the cards quickly at half power. You can slump over on the couch with the mobile version on a tablet and pass the cards that require you to write things. I’ve done this.

The silver lining: as your experience with SRS grows, so will your ‘worst day’ tolerance. So while your enthusiasm may wax and wane, your sheer grit will inevitably grow. But always remember that even the grittiest person is subject to the Law Of The Worst Day.

Because cards are being scheduled for your future, keep an eye on how they are accumulating in the future using the prediction features of your software. Do you have a 100 card day in two days, and are currently struggling with 80 card days? Reduce your new cards. But don’t put them at zero. We’ll get to that later.

2. The Law Of The Long Road

“Use SRS in service of long term goals.”

Do not use it for things you need to know for a specific class in a specific semester, but for things you wish to know for the rest of your life. Even if you are studying a language in school, resist the temptation to fold large amounts of class material into your SRS schedule. Design an independent schedule and follow it, and use class material only if it is something you want to remember for the rest of your life. Do not use SRS to study for your classes, but use your classes to enhance your SRS. When you set goals, they should be on the order of years, not months.

The point of the Law of The Long Road is that you control what goes into your SRS regardless of what is going on in the world around you. If you let external circumstances control your SRS schedule, you are doomed. Study for your classes in a traditional manner and use SRS to strengthen your foundational understanding of your entire major. Or just anything else you want to remember. Once a girl saw me doing SRS and asked me to remember her birthday. On a lark, I made a card–I still remember it. September 7th. It’s not even in my deck anymore. Wait, was she hitting on me? Nah.

3. The Law Of The Greener Grass

“There’s no easy way, regardless of what anyone else says.”

There will always be people who claim to have a better system or vocab list or SRS program, and you will come to doubt the effectiveness of your strategy. There will always be outliers or, more likely, ordinary liars claiming amazing rates of progress with another method or even (worse) the same method you are using.

The truth is that most strategies are basically equivalent, once the major errors of judgment are ironed out. Simply choose something you like, or design your own program, and thereafter focus your attention on its dutiful execution.

There is no easy path. Resist the siren songs of alternatives. You should be able to identify weaknesses in your program on your own and correct them yourself rather than constantly jumping from fad to fad and getting nowhere in pursuit of some unachievable ideal.

4. The Law Of Minimum Data

“The best cards are a simple question to which there is only one answer, found on the reverse side.”

If you find yourself questioning whether or not you should consider your answer as having failed–for example, if you have one card with a kanji complex on the front and its reading and meaning on the back, do you fail it if you get the reading but were just slightly off on the meaning? Or didn’t get one shade of the meaning? The problem is that you have to answer two things (reading and meaning) to get the card right.

Cards should only present you one question for which there is one clear answer, and the only thing which should appear on the back is the answer to that one question. I’ve seen cards with entire reference entries on the back–this is awful. Are you supposed to read that every time the card is produced? That would take forever. The Law Of Minimum Data means that you have to split cards that are supposed to convey more than one data point (for example, both reading and meaning, or multiple clozes) into multiple cards. If you want to memorize everything on the reference card (reading, meaning, production, context, writing), you have to create five cards. Not one card with a dictionary on the back.

Yes this increases the number of cards, but those cards will be more rapidly disposed of and your SRS experience with be far more enjoyable and less ambiguous. You can run through 150 well designed cards in the time it takes you to do 30 poorly designed ones, and you will remember the material much better to boot. You’ll also be receiving much better feedback about which elements are giving you trouble.

5. The Law Of SRS Limits

“Confine your SRS ambitions to concrete, specific facts.”

One of the major diseases I caught from my affair with Supermemo and its mildly crazy creator was the concept that SRS could be used for everything. By the time I left Supermemo, the man was stumping for a form of concentrated madness called spaced reading. I tried it out for a while, and only chalked up my failure to flaws in myself. His entire life was bent around memorizing everything. I think, eventually, even his emails were going through Supermemo to be subject to spaced reading. He wouldn’t respond to people for months.

It would certainly be nice to remember everything you were taught, but the pursuit of this will drive you crazy and you will fail, unless you are autistic in that exact way.

SRS cannot convey complex ideas and place them in context. You should study complex concepts at your leisure in a traditional fashion. Deploy your SRS as a precision weapon to learn things like dates, vocabulary words, or formulas. Cloze deletion is useful only as a supplement to already existing understanding. Do not cloze delete something which requires a contextual understanding you currently lack. Cloze delete specific facts whose context you are aware of but whose specifics are difficult to recall.

Bad cloze: The fall of the Roman Empire was due to ____________.
Good cloze: Salt begins to drop out of solution after ___ percent of water has evaporated.

The reason for this is that the bandwidth of a flashcard is low. It conveys a single point of data. The more a flashcard deviates from this dynamic, the less it is a flashcard and the less applicable SRS is to it. It is neither enjoyable nor efficient to recast traditional study methods into an SRS system.

This is actually a good thing. Your SRS study time is limited by the Law Of The Worst Day, which means that on most good days you’ll have plenty of time and initiative left over to further study the subject at your pleasure. You can even study the same things you use SRS for, such as vocab.

6. The Law Of Habit

“Practice SRS at the same time every day.”

The first block of free time after you get home from work or school, for example, is a fine time. After breakfast, for an early riser. Just make this time contiguous with the block of time in which you work. Don’t watch an hour of television and pour yourself a drink and then expect to hop on the computer and study.

Anki and other programs offer the opportunity to mobilize your cards onto a phone–personally I find this an onerous responsibility. Do not let your SRS weigh upon you at all times like an albatross. You control it, recall, it is a benefit to your life. It isn’t a tire around your neck.

7. The Law Of Getting Back On The Horse

“You will fail and begin again.”

A Worst Day will come which is worse than you anticipated, and then another after that, and another. Perhaps months will pass by. However, unless you’ve given up entirely, when you come to your senses you will need to get back into SRS. A daunting task. How many cards are due? Nervously, you open the program. 800? 1200? 1500 cards…?

When (not if) you encounter this situation, of course there is no getting around doing it. So swallow your pride. You still have the same goals you did before, or you wouldn’t need SRS. The ‘mercy’ function in most SRS programs will help you space out the reps and get into a proper habit.

The mercy function is important. Remember that when catching up, you don’t turn into some superhuman who no longer experiences dips in motivation or Worst Days. In fact it is the complete opposite: the nature of your endeavor will drain you more than new material, due to the problem discussed in the next law. Use the mercy function to space out your overdue cards to a reasonable per-day limit, and pay attention to the next law.

8. The Law Of Forward Motion

“In the shadow of failure, continue to progress.”

You may find yourself inundated by cards for whatever reason. Your judgment was poor, or you let too many days lapse. In these circumstances you may be tempted to halt the delivery of new cards. Isn’t that natural?

Do not do this. As you chip away at your mountains of review every day, you will become depressed by the oppressive fog of stasis. Haven’t you seen all of this material before? Is that what you’ll be doing forever? Spinning your wheels? And you’ve forgotten so much. It’s a dark swamp to slog through, consisting as it does of nothing but reminders of your personal and constitutional failures.

In these situations it is necessary to introduce an element of forward progress, however minimal, to help with your morale. I recommend splitting your deck into two decks: one deck ‘catchup’ will contain the hundreds of cards you have to review, and another deck in which you introduce a small element of forward progress. Not like you were doing before, which got you in this situation to begin with. Something small, even two new cards a day, will help.

9. The Law Of Dominance

“You rule your SRS. Your SRS does not rule you.”

SRS is a tool you are using to forward your goals. If you’ve designed your cards well, if you get into the habit, if you’ve paid mind to your own limits and the limits of SRS, you should be looking forward to doing your SRS on most days. If you find yourself dreading the experience every day, when your Worst Day arrives you can trust you will utterly collapse.

Certainly there will be some days when you don’t want to do it, but that isn’t the SRS talking. That’s past-you, who only wants the best for you. If you find yourself starting to think of your SRS as a whip-wielding dominatrix, and looking forward to the golden day when you accomplish whatever your goal was and climb out from under her scourge, you are doing it wrong. Review the laws, your cards, and yourself, and fix it. It’s much better than the alternative.

As part of the Law Of Dominance, back up your system and block out an afternoon to become intimately familiar with the workings of your specific SRS program. Learn how to create new decks and send cards to them, how to order the queue of new cards, and how to create new cards and change the format of existing cards. If you are unafraid of your program’s workings, you will be able to improve your methods continuously, based on your own personal findings, without being blown around on the winds by the advice of others.

One reason you are superior to your SRS is that the mature cards embedded in your SRS could be reconstructed in a fraction of the time it took to learn them, using only what is in your head. Your SRS data is a reflection of knowledge you have.

Students of SRS may come to confuse their deck status as being synonymous with their knowledge, and that is not true. If you were presented with a deck of English vocabulary, for example, you would easily be able to dispose of an immense volume of cards a day, and it would come to reflect your true abilities within a few months. A person attempting to learn this material from scratch using an SRS–say, 20 to 40 thousand words–might very well take a decade or more. So obviously, you are the one on top in this situation. Just because you have a deck, the deck isn’t you. You are the deck; it is a quantification of you.

Which isn’t to say it’s worthless. If quantification weren’t a worthwhile endeavor, we wouldn’t even bother with SRS. So keep backups of your data, because you will cry if it even seems like you might lose it.

10. The Law Of Pride

“Just fail the card and move on already.”

The reason people deliver themselves cards too complex to memorize and in volumes they can’t possibly keep up with is due to an overestimation of their own abilities. This is the infamous and insidious Dunning-Kruger effect in action. After all, it’s only those lowly mortals who can’t remember things and need to follow the rules, right?

Pride is the queen of the mortal sins for a reason. Pride leads you to waste hundreds of hours idling while attempting to remember cards you have, yes, forgotten. Pride leads you to idle on new cards which you think you should remember–you’ve seen them a few times by now, haven’t you? Pride leads you to pass cards you know you failed, but are ashamed to admit to yourself that you did. Close enough, right?

Above all, pride is counterproductive, frustrating, and a gigantic waste of time. If you’ve forgotten something, it isn’t an attack on your moral character. If a card is identified as a leech, it isn’t because you are stupid. Well, you might be stupid, but even if that’s true… there’s nothing you can do about that, is there? Except continue forward, earnestly, through the snow.

Be honest about yourself. Even if you could remember something if you pained over it for 30 seconds, that’s just the same as having forgotten it. So just swallow your pride and fail. The nice thing about SRS is that you are offloading the task of remembering what you need to remember to a system. If you pain over every little thing, the time savings is significantly reduced. Just relax and let the system do its work, because it is inevitable that you will remember it in time if it is a well-designed card. A well designed card is hardly any work at all–there’s a clear answer, and you either got it or didn’t. You should dispose of it, right or wrong, in under fifteen seconds.

Pride leads you to believe that you should be doing things faster and better, and leads to concurrent shame at your lack of performance, and neither of these feelings is true or productive. Pride is the main reason many adults don’t learn new things, so don’t let it stop you.

11. The Law Of Endlessness

“SRS never stops.”

Even if you learn 10,000 words of a foreign vocabulary, aren’t there many more? Aren’t there other languages you wish to learn?

Endlessness dispels the notion that SRS is something you use to succeed in college or to otherwise study for a specific event. It’s not best used in that way; college is designed around traditional study methods, as are most discrete, constructed learning experiences. SRS is for your life, and I didn’t come to appreciate it for what it was, and didn’t start having fun, until I’d left formal education. Users of SRS systems justify killing themselves on the notion that it is a temporary pain that will end when college is over, and that this is an ennobling experience. That theory doesn’t work, because SRS is not temporary.

Endlessness also means something even more heartening, which is: you are not someone who is four years away from reaching a goal. You are someone who knows more than he did yesterday. You will start to experience this feeling when (if you are studying a language) you start to read native material in the wild. You will see words appear that you learned that week, and see that you are not someone who doesn’t know X number of words, but someone who does know Y number, and tomorrow will know Y+8, and so on. You will continue to get older whether you have learnt anything or not.

People compare it to a marathon, or to climbing a mountain, but those have finish lines and peaks. Goals are incidental, and only exist as signposts. As Newman once said, the reason postal workers go, well, postal, is that the mail never stops. SRS doesn’t stop either, so keep that in mind.

Perhaps you knew all of this already, intuitively. I didn’t, however, so this is the fruit of bitter experience. Either way, good luck.

Trust the machine

For work I have to go to a lot of new places, and the necessitates relying, at least partially, on a GPS. My practice for using a GPS is to pull up the map, get the gist of it, and navigate myself 90% of the way there. I then activate it for the last 5 miles of local roads, which I generally don’t care enough to memorize. I listen to music and don’t appreciate the lady breaking in to tell me something I already know or at least should know.

Little did I know this might just save my life someday. Well I wouldn’t say I didn’t know at all; I had an inkling. ArsTechnica writes on the ‘Death By GPS’ phenomenon, which is hard to get all that worked up over. More interesting is when they start touching upon how we build internal maps of our surroundings:

The participants were linking their perceived location to its position on city maps that they retrieved from memory—which, like most maps, were oriented to the north. The cognitive process the participants went through was, in a sense, more complicated than necessary, since they had spent a greater chunk of their lives navigating the city than looking at maps of it—and most of the locations they were asked to point out did not appear on maps of Tübingen. Some participants reported that they had not looked at map of the city for decades. The knowledge they had acquired just by navigating their city day after day was multisensory and tinged with memories of real experience, whereas a map is flat, and the only sense it appeals to is visual. Yet, when asked to organize the information they held, they still reflexively translated it into broad survey knowledge, a bird’s-eye view. They willed themselves onto a map.

There’s a particular exit to the highway in Greensboro that I, for some reason, always exited in the wrong direction. Then one day I spotted the tallest building in Greensboro, the towering 20 story Lincoln Financial Center, peeking out in the distance while approaching this exit. Suddenly the map in my head rotated into place and I never exited the wrong way again.

Personally I love technology, but I’m far too experienced with it to ever trust it. Maybe part of the problem is that, in the modern world, so much of it works so well that it’s easy to feel a sense of security. For a paleonerd who grew up when it hardly worked, ever, it’s easier to be suspicious.