2 years of intensive learning : my experience with flashcards and the method of loci


Hello everyone,

I have recently finished 2 years studying in the french system of ‘classes préparatoires’ (or ‘prépas’). I was in the BCPST section (biology, chemistry, physics and earth sciences (there’s also math and other subjects)).

I’d like to share my experience with different ways of learning : the method of loci and an adapted flashcards system.

Compared to university and foreign school systems, this french system is somehow peculiar and quite demanding :

  • I had ~ 60 hours work weeks (including attending to courses and exams) but motivation was clearly not my thing. Many would have ~ 80 hours work weeks or more.
  • exams are frequent and do not simply assess learning. There are 2 oral exams and 1 written exam a week.
  • in particular, during the oral exam of biology (once every 2 weeks), you are given a subject. 20 minutes later, you must present it during 5 minutes in an organized way, with the help of the board you’ve illustrated during your preparation. The subject almost always requires to synthetize information, as it has never been explicitly studied in the course. Many subjects even require to synthetize information between several chapters (examples : ‘Nitrogen and metabolism’, ‘The CO2’, ‘Intercellular communications’…) Then after you’re done, the jury asks you questions about the topic.
    So you need to know your course perfectly and be able to answer any question related to it, but it isn’t enough. You must have deeply understood what you have learned, and you must be able to make fast connections. (Then you must be able to speak fast, but that’s another concern.)
  • though these exams are a cause of stress, they don’t matter. Your only relevant grades are those you get at the end of your second year, during a ~ 2 months period (the competitive examinations) consisting of many written exams then many oral exams.

Short overview :

During the first few months I tried the method of loci. I was very enthusiastic and wasn’t new at it anymore (I could memorize a deck of 52 cards in roughly 1 minute). But I failed at it. Then I tried an adapted flashcards system : on a sheet of paper I wrote questions about the course. The answers were in the course. I always reviewed the questions in order (I couldn’t do otherwise and it had some advantages). This method worked almost flawlessly and I kept it until the end (and I will keep it, reformulating my questions so as to keep only the most important things and perhaps this time putting them on Anki).

I know this contradicts the experience of other users (see possible explanations below).

Conclusion of this short overview : I would recommend this adapted flashcards system over the method of loci to anyone going into a prépa. It would be less legitimate to say so for other education systems as well, but I certainly would like to.

Long version

I would still recommend the method of loci to anyone asking how to memorize a long list of easy items (such as the countries of the world, the periodic table of elements, the decimals of pi…) especially if the order is to be memorized and the recall is to be perfect (no omission) without any clue.

However, I have found while learning these subjects that :

  • learning science is rarely about learning long lists
  • concepts aren’t easy
  • the order very often doesn’t matter
  • retrieval is almost always triggered by a clue (a question, a subject, an association of ideas)

Yet none of this means that the method of loci cannot be used with science. The main problem of this method is in my opinion that it takes too long compared to flashcards.

During these 2 years I wrote ~ 8100 questions (~ 5500 for biology and geology alone). This corresponds to 140 questions a week (holidays included and competitive exams excluded, I worked 58 weeks). Every week I would spend ~ 35 of my 60 work hours attending courses or having exams, and ~ 25 hours working by myself. Exercises and reviews left me ~ 8 hours to write questions. We arrive at 18 questions/hour or roughly 1 question every 3 minutes. This sounds little but (i) this was done while reviewing courses for the first time : some time was necessary to read and understand the content ; (ii) this pace had to be kept during 2 years ; (iii) had I sticked to the minimum information principle, I would have made a lot more questions (but how many times more, this is difficult to tell).

Now suppose that with the method of loci I would have needed as many loci as questions I wrote (this is arguable, but some questions would have needed several loci while others could be clustered into one locus. I don’t know what other estimation I could propose). I would have needed to create 20 loci a day. This sounds reasonable, but :

  • I would have needed to do this every day during 58 weeks i.e. 14 months, arriving at 8100 loci at the end ;
  • remember I only had 8 hours a week, i.e. 1 hour 10 minutes a day, to do this. And during 1 hour 10 minutes a day, I would have needed not only to create the necessary loci but also to read the courses and finish to understand them, and to implement the knowledge into the loci, which can take a lot of time. I deem this was impossible.

Though I truly spent ~ 1 hour 10 minutes a day writing my questions, I also walked in the town about 40 minutes every day, during a few months. I could have used this time to create loci and then have 1 hour 10 minutes to do the remaining tasks, but :

  • creating one locus every 2 minutes is possible but not, I think, in the long run. And the energy invested there (instead of refreshing my mind) would have been taken from elsewhere : I’m certain I would have been less efficient in some other area.
  • 1 hour 10 minutes would still have been too little : writing questions is in my opinion much faster than creating images within a memory journey.
    Even so the method of loci would have been impossible in the long run, at least for me.

Conclusion : the method of loci is too slow for learning science with a very limited amount of time.

I didn’t do this calculus at the time. Now that I make it, it confirms my personal experience.

Now some personal feelings about the two methods :

  • when you review with questions, you know you’ve covered everything once you’re done. When you review with a memory journey, you may have forgotten things if complex images were formed. Or you could write down the images you’ve created but that would really take a lot of time.
  • as I reviewed my questions, I understood the chapters deeper and deeper and I made connections between distinct chapters or even distinct subjects. I don’t know if that would have been the case with memory journeys.
  • I believe that the oral exam of biology described above would be much harder with a memory journey. When you’ve reviewed with questions, I think you can go through a chapter in your mind faster than with a memory journey, mainly because there is nothing to decode, as long as you’ve learned how the chapter is organized (by learning the titles of sections in order. Even with this, which seems better suited for a memory journey, I gave up because of the energy/time needed to create loci and images).
  • I sometimes had small lists to remember (~ 5 to 8 items) but I preferred mnemonic phrases I made up rather than loci, for the same reason of energy expense)
  • questions were not all about facts. Some were ‘prove that…’ (in math/physics) or ‘What are 2 consequences of this ?’ (I could sometimes answer by thinking if I had forgotten the facts, often it was a combination of thinking and recalling) (and this shows one of the advantages of keeping questions in order).
  • you can easily erase questions that aren’t actually relevant, reformulate them or add new ones. These actions are harder with a memory journey.

A few interesting websites many of you must already know :
20 rules of formulating knowledge
Strengthening the Student Toolbox

What about your learning experiences ?
Has anyone compared these two methods with similar learning conditions and similar expected results (i.e. something akin to the oral exam of biology) ? Or even tried one of these two methods ? Or another method ?
What do you think of my calculus and its conclusions ?
What is your fastest pace of creating loci (i) during, say, 1 hour ; (ii) during several months ?
What is your fastest pace of creating images within loci ? (complex images, related to a subject you’re currently learning)

I’m waiting for your reactions.


Although I don’t have a lot of experience with using memory techniques for practical work, I have been personally thinking that the MoL is not ideal for non-ordered information such as vocabulary words. Thanks for the long and insightful post.


Bump. I think this is an important post with findings that should be discussed. Does anyone else have any MoL real life application experiences that agree with this post or otherwise?


Thanks for this important rundown of your experience.

My first suggestion is simple:

Not all Memory Palaces are created equal. No one should expect them to be. People use the term “method of loci,” but because the approach is used so differently, it borders on meaninglessness.

We also need to combat what I think we should call “The Tyranny of Lists.”

Used well, the Memory Palace and related forms of “location-based mnemonics” might capitalize on linear structures in space, but it really has nothing to do with memorizing lists. It’s about creating mental rhizomes, or “rhizomatic” knowledge by harnessing the power of internal repetition.

In other words, the core difference between flash cards/SRS and a location-based mnemonic has less to do with the “material” (words on paper/screen vs. words encoded in chemical memory assisted by location-based mnemonics) and more to do with how you use these materials.

On top of this, there’s the “Either/Or Tyranny.”

You don’t have to have just one or the other. You can fuse index cards with location-based mnemonics in ways that explode the return on investment if you’re finding that one is not serving as well as you’d like.

In sum, people impose limitation on these techniques by playing games of comparison.

Let’s move beyond The Either/or Tryanny and start seeing location-based mnemonics for what they really are and then use them to amplify other strategies instead of pitting them against each other.

The tools work relative to each person showing up in the laboratory of one’s one mind, using them, making observations, then optimizing. This is the perfect science and there is no end to it. I applaud the OP, but kindly suggest that your Memory Palace practice could probably use some further study and finesse. You shouldn’t feel that it’s taking too long and it should be clear that you have multiple options for using it in ways that will amplify your results even further.

In all cases, follow your bliss and better memory is bound to follow!


But can you explain why linking shouldn’t be used instead of loci?


What did you use to learn English? Because it is spectacular.


You lost me at " though these exams are a cause of stress, they don’t matter." Why would you subject yourself to that much stress and preparation if they don’t matter?

I’m glad I didn’t go to school in France!

(Silent Avenger) #8

great read


While I agree that using mnemonics to memorize that much information in such a short period of time takes longer and is more cumbersome than a flashcard system, it does not mean that you are learning better in that way. Do not forget, one of the primary reasons for using mnemonics is not only for memory, but also for creativity and metaphorical thinking. The energy you put into thinking up similar sounding words that you can then translate into a memorable mnemonic image and then storing that image in pre-determined places in your imagination is significantly better for your own innovative intellect, than merely the drill-and-kill method for tests.

The type of education you went through sounds less like real education, and more like a method of turning the student into a prisoner of the system.

I like to mix study methods. Sometimes I prefer to just passively read or watch videos on the subject I want to learn, and other times I actively work hard to memorize it.


Thank you all for your replies.

Not all Memory Palaces are created equal. No one should expect them to be. People use the term "method of loci," but because the approach is used so differently, it borders on meaninglessness.

I should indeed have been more precise : when I spoke about the method of loci, I meant any method consisting of placing images (representing knowledge) into previously defined and ordered loci. This includes the most basic method with linear memory journeys, as well as Gavino’s massive memory palace for example.

Used well, the Memory Palace and related forms of "location-based mnemonics" might capitalize on linear structures in space, but it really has nothing to do with memorizing lists. It's about creating mental rhizomes, or "rhizomatic" knowledge by harnessing the power of internal repetition.

I never tried this approach. The idea of creating mental rhizomes − as I understand it, something like a mental wiki, defining the loci on the go while memorizing − seems great in principle but I don’t recall having heard of anyone using this method. Did you ? In my opinion, this approach must be extremely difficult to apply. And I don’t know how reviews would be done.

On top of this, there's the "Either/Or Tyranny." You don't have to have just one or the other. You can fuse index cards with location-based mnemonics in ways that explode the return on investment if you're finding that one is not serving as well as you'd like. In sum, people impose limitation on these techniques by playing games of comparison.

I agree with you. I don’t currently have anything memorized within a memory journey, but if I did, I would review it with spaced repetition (and obviously active recall as no one reviews memory journeys passively). You’re right, I only compared two extremes : 100 % method of loci vs 100 % method of questions. It would indeed have been better if I had tested every possible combination of the two methods, but I didn’t have much time to do this. However, I tried to inject a bit of method of loci into my questions-based workflow (e.g. by learning tables of contents within memory journeys) but it didn’t work for me. For this reason, I don’t think I could have done better with loci given these circumstances. Which leads us to…

I applaud the OP, but kindly suggest that your Memory Palace practice could probably use some further study and finesse.

I totally agree with you on that. I’m definitely not a memory palace master. Had I practiced the method of loci during years instead of months before beginning the prépa, and in various ways instead of obsessively focusing on playing cards, my experience could have been quite different. The fact is that I didn’t have the opportunity to get better at a technique that didn’t give instant results. So I had to abandon it quickly.
A better conclusion could be this : “if you’re not already very good at the method of loci, don’t try it in prépa. If you are, no advice. Keep in mind that this is only an attempt at generalizing a personal experience. Perhaps I was just really bad at memory journeys. But keep my calculus in mind, too.”

Let's move beyond The Either/or Tryanny and start seeing location-based mnemonics for what they really are and then use them to amplify other strategies instead of pitting them against each other.

I’m interested, since I know my use of the method of loci was far from perfect. How should location-based mnemonics be used in such cases, in your opinion ? Can you please describe what the best strategy to learn science would be, or provide links ?

Yes.. But can you explain why linking shouldn't be used instead of loci?

I didn’t say that so I don’t feel I have to. I didn’t try linking so I cannot tell. Linking could indeed in principle avoid the creation of loci. But I wouldn’t recommend using linking alone (see elsewhere for information on linking compared with the method of loci). I think linking could be better used along with the method of loci, so as to use fewer loci. Thanks for the suggestion. But I believe this wouldn’t be enough to solve the problem.

What did you use to learn English? Because it is spectacular.

Aged 15, after 6 years of slow progress in school, I spent a summer holiday learning english vocabulary 1 or 2 hours a day (in a non-optimal way, since I hadn’t heard about SRS or memory palaces at the time). Then I read books in english, without a dictionary (somewhere between 3,000 and 6,000 pages), and I began searching and reading the Internet exclusively in english. I also practiced writing thanks to school.

You lost me at " though these exams are a cause of stress, they don’t matter." Why would you subject yourself to that much stress and preparation if they don't matter? I'm glad I didn't go to school in France!

I’m glad I did − but we may both be subject to choice-supportive bias even though we didn’t choose our native country. By “they don’t matter” I meant “the marks you get there do not directly affect the grande école you’re going to get after the competitive examinations.” But the stress − which heavily depends on the person − shows that these exams somehow matter : you want to avoid failure especially in front of your teacher ; if you perform badly, your teacher is probably going to let you know he/she isn’t happy with your performance because he/she wants you to work hard ; failing at an oral presentation or an oral exam, even if your mark doesn’t matter, is far less pleasant than failing at a written exam. These exams also matter because they make you become dramatically more at ease and convincing when you speak in front of an audience, and because since you’re alone with your teacher, or with 1 or 2 other students, you can make really fast progress if you arrived prepared enough. And also because without them, a vast majority of students would work far less and far less regularly.
So the stress is − for stressed people − a side effect of the above advantages. As for preparation, it is (unfortunately, Marshall Rosenberg would say) for many a direct consequence of stress, but I consider that these exams help you to work hard, if you know that’s what you want to do. If that’s not what you want to do, they force you to work hard. But it’s a chance nevertheless, because I believe it makes you more efficient in almost everything you will do later in your life − not mentioning the fact that you’ve learnt a lot about science so you’ve improved your critical thinking and widened your knowledge.

The energy you put into thinking up similar sounding words that you can then translate into a memorable mnemonic image and then storing that image in pre-determined places in your imagination is significantly better for your own innovative intellect, than merely the drill-and-kill method for tests.

If you’re memorizing, say, digits, I definitely agree with you. But if you’re learning science, the answer may be trickier. Since the method of loci is − in my opinion − less efficient than the method of questions, using it means that you become better at creating images, visualizing in space and making a certain kind of associations, but at the same time you’re not as good in science as you could have been. Also remember one of the advantages of the method of questions is that you can partly learn by logic. And as you get better, you find yourself using logic more and more to remember things you would have rote learnt before. So you get better at making associations, too, and I would say that getting better at this kind of associations is more useful − if you want to do science or use science in your life.
Again, you can use both methods ; and if you find that the method of loci is not worth the time invested for learning science, you can still use it to learn cards, digits or − even better − random words, so as to get the above advantages ! Keeping in mind that your time is limited so you probably want to spend it carefully.
Last point : any method that hasn’t been explored to a great extent may be worth your time. This includes the method of loci, but also − unfortunately − active recall and spaced repetition.

The type of education you went through sounds less like real education

This calls for a definition. I found this one on Wikipedia : “Education is the process of facilitating learning, or the acquisition of knowledge, skills, values, beliefs, and habits.” I firmly believe I went through education.

and more like a method of turning the student into a prisoner of the system.

It’s true : you kind of become a prisoner of the system. But it’s for your own good. Think about all the times in your life when you would prefer or would have preferred to be forced the right way, instead of being able to choose.
And it has to be remembered that, although you are a prisoner of the system during 2 years, you have chosen this path before, knowing quite well what it would be like.
It’s easier to take a decision (I’m going to work hard during 2 years) once, than every day without sufficient incentives.
It’s a resolution followed by powerful means of application.
I’m not saying that the system is perfect or that it suits everyone, though. But I think it’s a fairly good way of transforming your learning efficiency if that’s what you want.

Now if I may reiterate my questions :

What about your learning experiences ?
Has anyone compared these two methods with similar learning conditions and similar expected results (i.e. something akin to the oral exam of biology) ? Or even tried one of these two methods ? Or another method ?
What do you think of my calculus and its conclusions ?
What is your fastest pace of creating loci (i) during, say, 1 hour ; (ii) during several months ?
What is your fastest pace of creating images within loci ? (complex images, related to a subject you’re currently learning)

I’m still waiting for your reactions.

(By the way I think that this forum as a whole is a goldmine of knowledge, and I really wish someone could make a literature review out of it…)


Hi, Rodent

Your post is one of the best on this forum, in my opinion, not only because it tackles the subject I am most interested in, but also because it is one of the rare that attempts to do it objectively. So I thank you upfront for sharing it.

I am very familiar with the writings of Piotr Woźniak and it was by reading him that I got into spaced repetition. The image below shows my usage of Anki in the past year: very lacking with respect to what I wish I had done, but that’s life.


I have used mnemonics sparingly throughout the year, so I can’t make as detailed an analysis as you, I am afraid. However, I have been memorizing entire books during the last couple months, so I might be able to collaborate a little on your inquiry. In fact, I have been willing to make a detailed post about “my method” here since I came back to the forum, but haven’t found the time yet: this is going to be a “short” version of it, one more geared towards simply fostering the discussion with you (and with others, of course).

If I understand you correctly, you don’t follow one of the most important precepts of Woźniak’s system of flashcard creation, and I believe this has made all the (positive) difference for your learning endeavor. By following his “minimum information principle”, one makes many more cards, and these cards are much easier to remember. However, they are much more geared towards fact retention than the “combination of thinking and recalling” that you mentioned. I believe that by sticking to full questions (possibly quite complex ones, if I am correct with respect to what you did) you better fostered your understanding of the subject-matter than by adhering 100% with the “20 rules”.

For science in general, I believe your method is one of the best ones there is. For history and philosophy (especially if one likes to situate all philosophy in time as a “history of ideas”), an integrated system of “mnemonic knowledge” seems best to me. Moreover, as you amass huge amounts of scientific information, even if they are not sequential or time-based, the organization of the information into an integrated structure for quick (mnemonic) review seems to me to be of great value. As others mentioned and you accepted the possibility, a combination of your method and mnemonics is probably the best alternative for any subject. Anyhow, I’ll keep my discussion now on how to create such an integrated system of memories in a feasible time. This is what I have been doing with books, meaning I memorize an entire book in roughly (but not necessarily) the same order as it is written. But any kind of knowledge can be memorized in the same way, provided you spend the necessary time first organizing the information.

Like you mentioned in your post, the memory palace technique is too slow and with limited scalability, because (i) it demands the creation of images, and (ii) it demands the accumulation of loci (real or otherwise). While (i) is unavoidable in mnemonics (it is its very essence), (ii) should be minimised for two reasons:

1 - the effort dispensed in creating/finding loci is totally unrelated to the understanding of the subject-matter to be memorized;

2 - scalability being limited, it will either prompt you to spend even more time “forcing” trips or video-game playing or whatever other methods you use for loci gathering, or it will have you create all sorts of artifices, turning your memory palace into a mess of imagined/real locations with little rhyme or reason, ultimately making it hard to remember. Moreover, most advice on memory palace formation presupposes that you will organize the information to be memorized beforehand so that rooms/objects conform to the subject-matter headings or data points, etc., but we all know that knowledge on any given subject is an evolving thing.

For massive knowledge accumulation, I see no other way than NOT relying on memory palaces. I should stress here before advancing that I see little doubt that the memory palace is the most effective way of remembering anything. The ability to “see” images placed at a given locus is real and one of the most impressive feats of the mind. Memory champions wouldn’t display sub-20 second speed-card records if that weren’t the case. But, for the aforementioned reasons, I believe they should be used sparingly, e.g., one locus for an entire book, and more like a general organization tool than as an indispensable need for actual memorization.

That said, what I have been using is a combination of layered reading, binary trees, and mnemonics. Perhaps by stating this way it sounds like an innovative and elaborate system, but in truth this has been commented upon in this forum and in many books in one way or another countless times — I just selected the main techniques that work best for me and combined them into a repeatable modus operandi that has been working really well so far.

(Using your style) Short overview:

You skim the book and create initial images for representing its main divisions (beginning with the title). Note that each of these images functions as a summary of the corresponding portion of the book it symbolizes, based on the level of understanding you currently have (you can and should always improve the images as you reach a better understanding). You then speed-read the book, taking note of the main facts you want to remember and, most importantly, of any metaphors that show up in the text (and which are related to points you want memorized). Now, you create this new level of images making use of the metaphors you found and any other you can come up with. Image formation can be done in any way you wish: word substitution, sound resemblance, etc., but I strongly advise on using metaphors based on meaning (in a memoria rerum style) every time it’s possible. At this second “reading layer”, you also mark the difficult/important passages which you’ll read again more carefully during a third pass. You then create more levels of images until your desired level of detail is reached. Then you link all the images into one single super-structure using the binary tree technique (more on that in the postscript) and put it into a spaced-repetition regime for constant review. As you also noted, it is important to hard-code your images somehow, so that you don’t lose information, and also for recall testing purposes, so I use a simple tree data structure in the R programming language to do it quickly and to display it at will. Finally, you create flashcards for specific types of information such as word-for-word quotations you don’t want to completely memorize but would be nice to keep in mind, and for any complex question you think is important.

I am a slow reader and, of course, this entire method makes reading much slower. But I have been (finally) able to memorize all the books I read, which is like a dream come true, in a speed of around 100 words-per-minute, all included. It is a very active and tiring way to read, but extremely worthwhile. A further advantage is that a breadth-first-search traverse of the tree gives you summaries of the entire book with increasing detail as you get deeper, whereas combining it with a depth-first-search traverse of the tree allows you to quickly get to any desired section of the book.

If there is interest in this method, I will write a “long overview” in a later post. This has already become exceedingly prolix.

Thanks again.


P.S.: This Wikipedia link shows how any arbitrary hierarchical tree can be turned into a binary tree using the “left-child, right-sibling” method. This means that if we are able to create mental “triple links” (very doable, believe me), which are just one more link (not two) with respect to the usual linking method everyone does, then we can maintain huge knowledge trees in our minds, trees that can be easily traversed, appended, trimmed, or modified in any way. If large sections of the tree (or one whole tree corresponding to, say, an entire book) are kept at loci, then you have a powerful system.

P.S.2: For a fantastic reference on layered reading (and a complete study method, for that matter) see “The Manual”, by Rod Bremer; and for the best info on speed-reading, in my opinion, see the classic “Break-through Rapid Reading”, by Peter Kump.

EDIT: I (hopefully) clarified the second sentence of the “Short overview”.


This is a fascinating thread of insightful posts. @Mnemoriam Yes I’m interested in reading the longer version of your system on how to read. I agree with the notion that we can sometimes expend too much energy creating loci for learning information, and in the end knowledge isn’t static so there will be a need for expansion.

Recently I’ve been practicing putting information in the Macunx arrangement developed by Aaron Ralby. Essentially it’s a way of linking pieces of information in 2D space. Because the information is always arranged in the same order, time doesn’t have to be expended to figure out where to place the next piece of information. For information that is already numbered, such as scripture with verse numbers, this is very useful for hard-encoding that number without having to add any image mnemonics. Every verse or numbered item now has a specific place relative to the other pieces of information. That 25 or 50 or 100 pieces of information in Macunx format can then be placed in any memory palace locus.

It sounds like your tree method is similar, but maybe more free-form than the Macunx. I can see where both would be useful, depending whether your information is numbered (use a Macunx with consistent locations for numbering) or not numbered (use a free-form tree of locations). I think combining the two methods could be very useful!


Hi, Slate

Thanks for your interest.

In fact, I don’t think my “tree method” is similar to the Macunx (as far as I understand it), but no doubt they can be used in tandem. Macunx seems to combine the ideas of a memory palace (even if an imaginary one) with that of pegs. The tree is just “linking on steroids”, so to speak. Therefore, once you create trees for whatever subject you want, just place it at a Macunx spot, associate it to a number peg, and you have a strong method for saving knowledge long-term.

I see you are interested in memorizing the Bible. While for such an audacious endeavor any help would be welcome (Macunx, memory palaces, pegs, etc.), I think that progressively incrementing a linked tree-like structure would be ideal.

The whole Bible (Old Testament + New Testament) has 31,173 verses, each of which has, on average, little less than 25 words. I don’t know your memorization objective, but in case it’s not memoria verborum, I firmly believe that one complex image is enough for each verse, on average (some small or monotonous verses could be combined).

That being the case, if we were to do the obvious and follow the Bible structure closely, we would have a tree with only 5 levels (Bible->Testaments->Books->Chapters->Verses), but that would mean having to form “right-sibling” link chains of up to 150 links (Psalms) or, more commonly, of more than 20 links (Leviticus: 27; Joshua: 24, etc). This is clearly not ideal. By doing our part in active reading, we can segment the text (any text) further by establishing other natural divisions and summarizing each of them by a short description to be turned into a complex image.

If we did so, we could, for instance, increase the number of levels to 8 (Bible->Testaments->Tomes->Books->Parts->Chapters->SubChapters->Verses), what could give us on average: 3 Tomes per Testament; 4 Books per Tome; 5 Parts per Book; 6 Chapters per Part; 6 SubChapters per Chapter; 7.3 verses per SubChapter, for a total of 1234566*7,3 = 31,536 verses, more than enough for the whole Bible without ever (or rarely) making more than 8 consecutive links.

And the great advantage of such method is that it is perfectly coupled with the layered reading idea, which, in my opinion, is the best method for reading ANY kind of material (with the exception of mystery novels).

So, in first skimming the whole Bible, you could understand it just enough to create the “Tome” level of divisions. That means understanding the structure of the Bible well enough just to be able to express in as few words as possible the unity of each tome; which means, in turn, being able to find natural divisions in the structure of books; which in turn means grasping the gist of each book. That’s why this method is not simply mindlessly creating images for each entry in a table of contents: you actually have to skim the whole book in order to have a cursory understanding of its contents. Or, of course, you could use some kind of CliffsNotes or Wikipedia for these first layers of reading, which, in my opinion, is a very valid idea (although I prefer to hold the real text in my hands and actually skim it).

After this, you are already apt for memorizing the whole Bible, only with a shallow level of understanding. But the beauty of it is multifold: (i) you feel accomplished for having “read” the whole text, in this case, the Bible, and even memorised it; (ii) you greatly facilitate further understanding to be gained in the next layer of reading, thus making it much faster to read next time; (iii) by creating a mnemonic representation of the content, you have in effect created a structure to link more detailed images, hence also facilitating the process of further mnemonic encoding; (iv) you can place these “high-level” trees in any memory palace or Macunx you want, further facilitating retention in memory, as well as use any kind of spaced-repetition regime while you keep getting deeper into your reading and mnemonically encoding specific portions (“sub-trees”) of the book.

I hope this example can help elucidate the whole idea while I don’t have the time to actually make a detailed post about it.

Best of luck!


Thanks for sharing! Your posts are chock-full of good stuff and I’ve re-read them several times for better understanding.

Yes I have a long-term goal of memorizing the content of most of the Bible, and as you assumed, not verbatim to a particular translation. My interest is in having a mental index of what is happening in every chapter and verse (excepting a few portions like lists of genealogy). I’ve been reading about various approaches and experimenting with possibilities on how to do this for a couple of years. Until now the best approach (and indeed effective) I had used were memory palaces based on real locations.

Then this fall Mnemon got me thinking about Gregor von Feinaigle’s approach to the artificial memory rooms where numbers always reside in consistent locations. I plowed ahead with this for a while, once I realized that being consistent with verse location in space meant that the chapter number or verse number could be hard-encoded to the space instead of having to add number images.

A month ago I began to look more at Aaron Ralby’s approach, which I initially thought was just for languages, and realized that the Macunx pattern is a lot like Feinaigle’s memory rooms. The Macunx is simpler, though, and thus more flexible than Feinaigle rooms I think. The Macunx is initially a 2D pattern, but if, for example, you picture it laying flat like a chess board, you can then build vertically at each location and it becomes a very 3D memory palace. The MacunxVR software to be released this year will allow utilization of SketchUp’s 3D Warehouse and other 3D databases to build these 3D memory palaces and have a way to review them as well.

Everything you’ve written regarding the need for understanding of the text structure in order to create a tree makes sense. I’m also in favor of breaking up some of the chunks into logical groupings such as the ‘tomes’ that you refer to. What I have in mind, though, is to not break up books and chapters of the Bible into subsections, but retain their numbering. It’s difficult to retain numbering for each chapter and verse without using a number image system many times over. BUT, if each section of the tree that has number order is arranged in space according to the Macunx pattern, then the numbering can be hard-encoded into the space itself. Here is an example:

The green numbers represent 27 chapters of Leviticus, and the orange numbers represent 34 verses of Leviticus 27. If a person wants to link each of those 34 verses to an image of Leviticus 27 they can do that, or they can just focus on the unique location in space of each of these verses, relative to the other verses, and create links between the verses if they like.

Most books aren’t marked by verse, so the need for a numbering pattern like the Macunx is less necessary. But since the Bible is already organized by numbered chapters and verses, I think it makes most sense to use some kind of hierarchical approach (tree or memory palace) that also employs a consistent use of space for numbering. So instead of having all the little tree branches extend linearly, they can extend in a circuitous pattern that follows something like the Macunx pattern. In fact the Macunx pattern is derived from the simple quincunx pattern of 5 locations, where each location is very memorable, and avoids linear placement of information as much as possible.

If I’m overlooking something in my analysis here let me know, I appreciate the feedback!


Hi, Slate

I think everything you say makes perfect sense, but I believe I lack some understanding of these “quincux/macunx” patterns. They are just geometric patterns for placing images, right? They are not memory palaces per se, not even imaginary ones; they are simply ordered patterns for placing images so that you always know their relative order just by looking at them. Is that it?

If so, where are they actually placed? Do you still need to create a real or imaginary memory palace where you would place the images using the quincux pattern? Say, for the 34 verses of Leviticus, would you still need 7 rooms (or any other structure) where you would place the quincux patterns? If that’s the case, I would say that these patterns provide little advantage over normal memory palaces.

Now, if you don’t need actual memory palaces, then even if you link 7 quincuxs like in the Leviticus example, you would be, in effect, creating a long chain of 34 linked images, wouldn’t you?

Please, clarify these questions for me, so that we can proceed with this interesting discussion with more precision.




Yes that’s correct. And more importantly than just telling you the relative order of information/images, each location of the macunx tells you exactly what number it is in that order (according to the pattern of numbering as shown in my previous sketch).

It’s very flexible I think, you can use the macunx pattern as a very simple outline of the information, or you can embed it more permanently in the design of a memory palace, real or imaginary. One thing I find very intriguing about the macunx is its ability to have some of the benefits of a memory palace–information placed spatially relative to each other–without needing to take the time to find or create a memory palace structure. I asked the creator of this macunx pattern if he found, after much practice of using it, that his mind would automatically arrange numbered information in his head according to the pattern, and he affirmed that it was so.

So I think the macunx has a lot of potential for on-the-fly learning, much like your memory trees. Even without mentally ‘placing’ a macunx in a memory palace, I think it still has usefulness for giving each piece of numbered information a home, a place in space for each item of information relative to each other. But, for information that is to be stored long-term, such as for my Bible project, I think it’s best to combine the macunx pattern with features of a regular memory palace. With the aid of 3D software, an artificial memory palace can be constructed, designed logically with macunx patterns arranged in macunx patterns for easy placement and retrieval of information.

One of the things that intrigued me about the macunx design is that every location is unique. Before, drawing from Feinaigle’s methods, I was working with 3x3 grids. The problem with information in grids, and especially as they get larger, is that it’s easy to confuse consecutive pieces of information. The macunx is designed to avoid this problem. The creator settled on the quincunx as the basis for the pattern, because when working with anything larger it was found that people tended to try to break the six or seven pieces of information into smaller groups. It also avoids placing consecutive loci next to each other vertically or horizontally.

So the macunx (short for “magical quincunx”) is a pattern of 4 quadrants, each containing 25 loci. They are usually ordered upper-left, upper-right, lower-left, lower-right. This gives a total of 100 possible loci for one macunx, though more quadrants can easily be added if needed. I also ignore unused quadrants, so if I need less than 25 loci then I just think in terms of one quadrant.

I’m giving all this explanation because it helps show why each location of a macunx is highly unique. When I think about the number 25, I immediately think about it being the lower-right location of the first quadrant. When I think about the number 26, I immediately think about it being the upper-left corner of the second quadrant. I also think of it being adjacent to the 7th location, and could make a link to that if I wished. 27 is unique too, because it anchors the upper-right corner of the first quincunx of the second quadrant. When I think of 27 I immediately go to that mental location. 28 is a ‘middle’ number, and middles of quincunxes are always interesting in my mind.

As you can see, each location of the macunx pattern is unique and memorable. The more I use the pattern, the more I find that each location has a certain feeling to it. If for a moment I ever think about how one location is similar to another location in the macunx, I stop and think about how it’s different, how it’s unlike other locations.

So to your question, yes you can use a macunx to create a long chain of 34 linked images, but the result will be much more than a mere chain. In a linear chain of images with no geometric pattern, one broken link could result in irrecoverable loss of the remainder of the chain. But if that chain was forged in the pattern of a macunx, there will be many more links to hold the images: memories of certain images being in certain corners of certain quincunxes, or in the middle of particular quadrants, thus giving their relative location AND exact number. In addition, mental links can be forged between totally unrelated pieces of information, such as with the adjacent loci of 7 and 26 as mentioned earlier, or 19 being above 51, the start of the 3rd quadrant.

I hope this helps with clarification. I’m still relatively new to using macunxes so I’m still learning ways to use them, but from what I’ve seen thus far I find them to be both simple and powerful.

Do you actively visualize the branches of your memory trees extending in a certain way in space? Maybe as “left-child, right-sibling”? Before I learned about spatial learning I never gave any thought to how I remembered things, such as days and dates. Then one day I realized that I already have a spatial way of remembering these things! My mental days of the week go from right to left, and then turn a corner at every weekend like a game of Candy Land. And my yearly calendar is a circle of months that go counter-clockwise! These certainly aren’t the ways I would intentionally create a mental daily planner or yearly calendar, so where did such designs come from? I think at some point early in my childhood my brain just arbitrarily chose these patterns as a way to store the information in space, and they’ve stayed with me ever since. I bring this up because I think it could be indicative of how spatial relationships will be forged in the brain, even if we think we’re only making mere links between information.



Thanks for your detailed explanations. They were very clarifying indeed.

I find it difficult to do. If you can put 5 images into a single locus, such arrangement could help in discerning the order of the images within that locus. However, I can’t place 5 images into a single locus, all I can do is link 5 images in a single locus and linking can’t follow strict geometric patterns (not for me, at least).

The value I see in the macunx is simply a pre-established geometry for creating an artificial memory palace, one that has the added advantage of providing an inherent numbering for the loci. However, I don’t see much difference with respect, for instance, to a “tricunx” where each pattern is a triangle where you place each image at each vertex and then keep joining adjacent triangles to the sides as if forming a fractal-like structure. This also has an inherent ordering and, with practice, you would also know in each “quadrant” you are and what is the number of each image. It might be a little more difficult, but this was just a quick invention, there might be many other better possibilities. Moreover, you can always know the number of most loci in a conventional memory palace, real or not. The ancients advised placing pegs in loci which are 5 or 10 multiples but even that is unnecessary. I know most of the loci of my memory palaces for cards by number after a few weeks of practice. Now, you could argue that as my memory palaces grow large it would be harder to know that, but the same thing would happen with your macunx; if you join 4 4-quadrant macunx for a total of 400 loci it would begin to get quite complicated to know in each quadrant and corner you are and then to decode it into an exact number. Even in your simple 34-loci orange Leviticus example above, it is not very easy to know when you are at the 31st locus, for instance.

My point here is that I find artificial memory palaces already a difficult thing to use, even more so (much more so) if it is a regular artificial memory palace with no distinction whatsoever from locus to locus. I understand when you say “The more I use the pattern, the more I find that each location has a certain feeling to it”, but that’s because our minds are impressive tools, not that this technique actually provides patterns that are “unique and memorable”, like you said in the sentence previous to the one I just cited. It is your impressive capacity as a human being that lets you use this technique and your particular impressive capacity as Slate. I state this last fact because I myself don’t have the visualization capacity to use such a plain and monotonous geometric structure to do much with it. You can, so it is useful, and I won’t contradict you. But not for me. I remember a girl in this forum (I think she was a “she”) that used an imaginary book as a sort of a memory palace. She would place image pegs on each page so that she knew the number of the page and then she would add the images for the day on that page. I can’t do that. Either I’ll have to link the images from page to page together with the number pegs or that (the book) would be useless to me. That impressive fellow Lembran Sar remembers every day that passes by using a mental calendar as a memory palace. I can’t do that either. So I congratulate you for being able to use the macunx pattern for long-term retention. I could probably use it just for quick memorization of 5-element lists, but then I always have a familiar (real) memory palace where I could easily place them instead.

Yes. If you not only link a long chain of images but also place them 5 by 5 in a quincux geometric pattern, that would indeed be more memorable and less prone to a chain reaction of forgetfulness. But, again, at least for me, I can’t “force” a quincux pattern to my linked images. Linking demands a sort of freedom of imagination that makes such a geometrical arrangement very disagreeable to enforce. That said, I do “force” some geographical awareness in my “triple linking” strategy for the trees, which brings me to your question.

Yes, “in a certain way”. The tree is made by local triple links. The final geometric pattern of the entire tree is irrelevant. All that matters is the capacity to maintain mental local triple links – the rest unfolds naturally. And I don’t say that maintaining such links is trivial. No, it demands repetition and care. And a little method. With respect to the method (with explains the “geographical awareness” I mentioned above), allow me to reproduce a fragment of what I wrote on another thread recently:

So, I see each right-sibling link indeed “extending in a certain way in space”, and if I can keep adding other right-sibling links in the same spatial context I will, but that’s not a need. What I need is to differentiate between a right-sibling link and a first-child link so as to be able to traverse the tree in any way I want. The difficulty lies in knowing if there is indeed a right-sibling or a first-child link in a given image (node) of the tree. For that, until now, I need to rely on spaced repetition. If you need a material example of such linking, I can give you in a later response; this is already too long.

With respect to losing a link and forgetting a huge portion of the tree, that is minimized by summarizing the information of lower levels by creating intervening levels just I suggested you do with your Bible project. The linked chains will never be too long if you take this care and in the event you do forget a link, just go up a level to where you were before and restart from there.

I see your point of keeping the number of the verses, so simply using my method would not be ideal for that. Maybe you could indeed add the idea of the macunx like you suggested, but I myself still think that adding number pegs to upper branches telling you the beginning number of the lower branches would suffice. Then, as you follow the right-sibling links, you would count them and automatically know where you are. But I have never tried this so I wouldn’t know for sure. That’s an interesting idea to try.

Have you begun your Bible project or are you still thinking about what method to use? Maybe it would be interesting for you to make preliminary runs of small segments of the Bible and actually try different variations of methods. I’d be very interested to hear your findings.



Hi Mnemoriam, Thanks again for yours and Happy New Year!

You give a very apt description of the function of the macunx. Indeed instead of a quincunx as the basis, other geometric patterns like a triangle could be used in the same way. I recall that the creator of the macunx acknowledged this very point. So I think use of other geometric patterns is worthy of consideration, particularly if they lend themselves better to a particular subject of study, but this point doesn’t detract from the concept in general.

For me the important thing is to settle on a geometric pattern and practice using it over and over again, so that in time my mind can more automatically place numbered items in their appropriate locations relative to each other. I find the macunx pattern sensible enough, its basis of quincunxes makes it easy to count by 5s, and the 4 quadrants give a nice round 100 locations.

For my Bible project it won’t ever be necessary to join that many macunxes together, seeing as there is an average of a little more than 26 verses per chapter. Psalm 119 is a major outlier with its 176 verses. As for being able to tell when you are at a certain locus of the macunx, I don’t deem this as difficult, just something that requires a bit of practice.

I agree. This is why I am greatly looking forward to begin the 3D modeling process of my memory palaces, as I mentioned earlier. I could start doing this now with software like SketchUp, but my experience has been that such software has difficulty handling scale. So I will wait now for the release of MacunxVR which is being designed for this very purpose and will allow infinite scaleability and use of 2D images and 3D models. Then these memory palaces will be as memorable as any video game (actually moreso because of the creative process required to build them), and easily reviewed by visual walk-through, making spaced-repetition less necessary.

And so it is with all memory techniques, I believe. As individuals we are probably drawn to certain techniques more than others.

I’m glad you mentioned Lembran, because he is one of my inspirations. He is using a mere monthly calendar, which I believe is a much more difficult geometric shape than the macunx because there are 7 columns by 5 or 6 rows of cells, all adjacent to each other with no space between them forced by the geometry. Yet, with practice, he’s remembered something from every day of over 6 years of his life. So, as we often advise new mnemonists, go with the technique that inspires you! Clearly some practice and determination has turned a simple monthly calendar into a powerful artificial memory palace for Lembran.

It’s interesting to hear your feeling on this. For me, forcing the information into the macunx pattern is very enjoyable and interesting. Many of the verses are already familiar to me, but now that I’m forcing them into the macunx pattern (as images), I have these ‘aha!’ moments when I ‘see’ where they belong in space. They are no longer floating about, but they are anchored in relation to their surrounding verses, much like the pleasant feeling of anchoring a bit of information to it’s locus in a non-artificial memory palace.

I appreciate all you’ve shared about your linking/tree method. I feel like I’m getting a little bit closer each time to understanding it, but a full example would definitely help to clarify. Only when/if you have time though. Particularly I’m unsure of how much you use memory palaces. If you decide to memorize some book, do you only use one locus of a memory palace to store the entire book as a tree?

At this point I’m settled on the prospect of using the software, when it becomes available, to model my memory palace in 3D. I believe this will greatly assist my mind in visualizing the information in macunx format (which then facilitates easy chapter/verse number retrieval), as well as offer a very visual way to review the scripture and see what areas need to be worked on.

But I’m interested in any technique like yours that may offer a more practical approach for retaining information encountered on a daily basis. The structure and content of the Bible is very established, so I think it lends itself well to a more rigid structure like the macunx. But most books aren’t marked by verse number, and most information in general is fluid, so your linking method is very intriguing.

I am still in the infancy of practicing the macunx, and I feel it has good potential but I can’t speak for results until I’ve given it sufficient time and opportunity. If all else fails I can always go back to using regular memory palaces, which have worked well for me in the past, and I can be more consistent about my use of numbering.

Since the 3D software I’d like to use is not yet available, and even then the modeling will consume considerable time, I can use this time to improve my ability to create these structures mentally. I think using some of the strategies you’ve suggested for linking and creating composite images as headings for specific sections of the macunxes will be very useful.


Hello, Rodent!

Thanks for sharing your experience, I liked it very much.

You said you used an adapted flashcards system, and that maybe you would use Anki in the future. My question is: did you use it by now (from August up to here)? I would like to know your experience with that.

From what I understand, you were very concerned about time issues (as I think every student must be), and I guess once you use Anki for a long time, the time of reviewing the cards due on a day can be very huge and intense. I don’t know if it is worthwhile.

Also, in this sense, when you didn’t use Anki in the past, how often did you review the questions you created for self testing? You went through them in order you said, but how often? (I personally don’t like the spaced repetition idea, and I’m guessing you didn’t use it too).



Thanks Mnemoriam and Slate for detailing your very interesting techniques! This might indeed be a really good way to memorize the Bible, or any book. As for myself, I’m currently favoring Internet articles and Wikipedia pages over books (this might change again in a few months); that’s where incremental reading is especially powerful. I’m using the incremental reading Anki add-on. But I may (read your posts again more slowly and) try your techniques in the future. By the way Mnemoriam, I also think that an example of your system would be awesome!

Yes, this might be true. This got me thinking about the optimal complexity of items − a vast question, it seems to me.