Thoughts on Expanding to a 10,000 Image Mnemonic System

Yesterday, I wrote down some thoughts on phonetic memory systems. The main part of my system is made up of 2,688 one-syllable words that I think of as a kind of artificial language.

A number like 211614127 is pronounced “NIT-BIR-TUK”. The artificial word, NIT, means Magneto, BIR is beer, and TUK is toucan. The reason behind the one-syllables is explained in the previous post. Basically, some cultures apparently have a greater short term memory capacity because their numbers can be pronounced more quickly.

Some people have been talking about creating 10,000 image systems. I’ve been thinking that if I were to expand my decimal system from 1,000 to 10,000 images (which I probably won’t), I would try to keep the phonetics the same.


  • 0000 = SOSO
  • 0001 = SOSI
  • 0002 = SOSU
  • 0010 = SOTO
  • 0100 = SISO
  • 1000 = TOSO
  • 1100 = TISO
  • 1112 = TITU
  • etc.

A question I have is that, if the time it takes to pronounce the syllables does have an effect on memory, would the extra syllables slow things down?

In my 3-digit system, 6 digits are stored in 2 syllables. In a 4-digit system, only 4 digits would be stored in 2 syllables.

A point against the 3-digit system is that it has an extra consonant to pronounce. Which of the following would be faster?

  • NIT-BIR = 6 digits
  • NITI-BIRI = 8 digits (flows better)

Also, do diphthongs slow things down? NAITAI-BAIRAI takes longer to pronounce than NITI-BIRI. Is it possible to eliminate diphthongs from a system, maybe by borrowing vowels from other languages?

I don’t think I’m going to try to expand to 10,000 images, because it’s already taking too long to to finish a system with fewer than 3,000, but it’s interesting to think about.


  • I’m working on a 10 000 images system with 4 digits, 10 binaries and 2 cards representing 1 image. I’m nearly finished. Now it is memorization time.

  • Sorry I meant 12 binaries. It will be tough to get all this to memory.

  • Very cool… You found 10,000 images already? Is the system phonetic?

    I think the rows of binary digits in competitions are 30 per row. I wonder if it would be faster to do the binaries in 15.

    With 10, it’s 3 images per row, one locus:

    With 15s, it’s 2 images per row:

    With 12, the images overlap and it might be easier to lose track, or slow down on every other row:
    011101011001–010011100101–000110 [overlap]

    I don’t know… just speculating…

  • Hmmmmm… If everyone is moving to 10,000-image systems, I wonder if it’s worth looking into. I’m worried that it would take me a long time to generate the images.

  • I can see why it’s easy to do binaries in 12:

    15 gets more complicated:

  • No my system is not phonetic. I started this about 3 months ago when nobody ever discussed it. I started working earnestly on it since last month.

    I tried the PAO system. It doesn’t work for me. I often confuse the actions. I tried Ben’s system but it takes me too much time to read the words, convert it to phonetics, determine the words associated and obtaining the images.

    So I decided to keep it simple. I’m using a purely numerical system. 4 digits = 1 image.

    I’m then converting the binaries and cards to 4 digits too. Like 000 = 0, 001 = 1, etc…

    So 000 001 010 011 = 0 1 2 3

    I’ll then recall the image for 0123.

    For the cards, I’m converting a card to a 2 digits number. Like Ace of spade = 01, Ace of hearts = 21
    So ace of hearts + ace of spade = 2101 (Image of 2101)

    I knew there’ll be a problem with the binaries. I tried first with 5 binaries = 2 digits. So one row will have only 3 images for 30 binaries. But I find I work faster with 3 binary instead of 5 (conversion from binaries to numbers to images).

    I’ll be working in rows of 2 with exactly 5 images.. I’m using journeys with vast spaces for the binaries with one route point having 5 images, each interacting with the other in a story.

  • For the images, I started doing them slowly. I used every people on my facebook account and all the people I know and turned them into images with their months and dates as numbers. Like someone born on the 22 of february is 222. However if I think of them as another number more easily, I just used the other number.

    I also use all the familiar objects I know which have a number assigned to them, so I can instantly recall the objects. Like my pin card for my credit card, the laptop model 7738 for my acer, etc…

    I soon ran into a problem. I had only about 1700 images in all and I have used up everything I know. So I had to think of something.

    I’m a movie buff and television series buff. So I started to list all the series and movies I’ve watched and started building images for each actor, character, and turning them into numbers. Afterwards, I started doing these for all famous persons I know. Like Einstein or Madonna or whoever.

    But I ran into another problem. I was working at a rate of about 50 images per day. I’m a medical student and sometimes I don’t even have time to find images. I’m preparing for the wales competition in March and at this rate, my system will be ready in about 200 days. I won’t have a system in place for the wales competition, let alone practising or having memorised it.

    So I had to come with a solution. I found out that the DOBs didn’t matter. If I assign an arbitrary number to Jack Nicholson like 2132, then after some practice, I’ll be able to recall the image of Jack Nicholson for 2132.

    I went to sites like Fiverr and elance and outsourced the image building. Cost me about 30 bucks but I had a system in place without any efforts. All I have to do now is just memorise them. I’m still tweaking the system and replacing hard to recall images with easier ones.

    The memorization part will be the hardest one.

  • Thanks for the writeup… very interesting. Maybe I should try elance to build my flashcards.

  • I’d recommend fiverr because I find the people over there more responsive and it is far cheaper than elance.

    By the way, if you are interested I can provide you with my 10 000 images, all famous persons only.


  • You could use all the members of the U.S. Congress for 535 images. Each can be tied to a geographic region for an extra peg.

    I think anecdotal evidence will show a 4-digit (10,000) system is too unwieldy and will be inefficient in practice. I think the gain in chunking a larger number (only 1 digit more) will be more than offset by the added effort in recall. My intuition tells me that there is an unconscious sifting through of multiples of images even when just one image is consciously recalled. A 10,000 image system collapses under it’s own weight. But that is just my inexpert speculation.

  • @Yan
    Thanks, I’ll try fiverr.
    I would be interested in seeing your list.

    It does seem like it would be incredibly difficult to maintain all those images. I’m fascinated by the idea though. 🙂

    By the way, is there any interest in my adding a simple forum to the site so that people can discuss techniques and share tips/images/ideas?

  • I put a basic forum online here if people want to use it to trade tips and ideas:

    With the forum software, anyone can start a topic or post ideas.

  • @Dan: Here’s a cliché for you: Rome was not built in one day! 😉 I’m thinking long term with this system.

    One thing that I have seen people neglecting is the encoding process. This is the biggest problem I think with Ben’s system. There are too many steps involved between reading the numbers and generating images.

    I seriously weighed the pros and cons for such a system. But I’ve found it hard to speed up the encoding process for the other methods. For Ben’s system for cards, I find it really hard to memorize the letters and then transcode the cards into letters, then words first and then images, especially the double suit letters. Plus it was taking me a lot of time to come up with images based on phonetics.

    PAO with such a system is much easier but I tend to mess up the actions part.

    So I preferred to keep it simple. I think that the recall is the same but the memorization process for 10 000 images will take a lot of time. I’m trying to make each number as personal as possible so that I can instantly think up of the image when seeing the numbers.

    Also 1 letter may seem too little for 10 000 images but with each image you are storing 1 letter more than the others and this will count as a lot in the end.

    Anyway I’m not saying that this system is perfect but I’ll give it a try and work on it for about 1 or 2 months and see how I’m doing. In the meantime, I’ll try to see if I can come up with easier systems.

    @Josh: I was trying to set up a blog but you just saved me 15 minutes of buying a domain and setting up the blog! 😉

    How do I send you the list?

  • I’ve learned a few foreign writing systems while traveling around and it was very difficult in the beginning. With enough practice, I’ve noticed that I start to instantly read the foreign words by their shape.

    Example: ξενοδοχείο (hotel in Greek). I’ve seen that word so many times, I don’t need to sound it out anymore. A quick skim is enough to pick up the shape instantly.

    What I’m doing with Ben’s System is working the same way for me. It’s actually a bit weird to read numbers as if they are words, because it sometimes happens unintentionally and I think it uses a different part of the brain.

    It was painfully slow in the beginning though.

    I’ll send you my email address…

  • I’ve read Ben’s blog back to back. And on one post, I read when he was developing his system back in 2003, that he never looked at anybody’s system. He just sat down and came up with a system that fitted him best. I’m sure that with sufficient practice, any system can be mastered. I find it easier with the system I’ve described above. To each his own.

    On the memory sports yahoo group, I read the interview of Wang Feng, translated into english. Looks like he is using a 2 digits system. It’s pretty amazing that he could memorise 480 numbers with that. He must have practised a lot.

  • It sounds like he trains a lot. These are from some of the Chinese newspaper articles on the Yahoo Group, translated with Google Translate:

    This is hard to prepare for the World Championship of training class, Wang Feng, “coming to the fore” the. At that time, Wuhan, intolerable heat, several members of Life Sciences, rented a small classroom, 5,6 hours every day practicing a variety of memory skills on the inside, tired, lying on the table a break and ate lunch. “High temperature, dry training, actually persevered.” More than a month, Wang Feng made rapid progress, unhappy memory of the master in January reached one of three criteria — “No one within two minutes of regular playing cards Memory . ”

    Reporter has learned that this competition before, Wang Feng and master Yuan Wenkui village in Wuhan, rented a house less than 20 square meters, the house only a bed and two tables. They begin at 8 am from the day has got to 17:30, sometimes even “late training.”

  • @Yan
    I agree that the encoding intermediate step is only necessary to make things easier in the beginning and no doubt hampers things beyond that stage. Some memory athletes say this encoding drops away after time and when you see a number you see the peg image (504 = laser, not 504= l+s/z+r= laser). I first started with the memory system and did the encoding. I then tried to move to the Dominic System. I had trouble coming up with people for the combo of letters. So I just made up a list of the 100 most memorable people (Donald Trump, Oprah Winfrey, George Clooney, etc.). While I didn’t pursue it too far before falling back to the major system, I did notice my recall of certain pegs was easy and instantaneous and unencumbered, unlike what I experienced with the major system translation/encoding.

    I think a 10,000 image system is suited for someone beginning in elementary school, where they would have the free hours each day and the years to develop and the absence of distractions.

  • Wow, it is incredible and encouraging that Wang Feng uses a 2-digit system. This goes to my earlier contention that it might be better to use a simpler system (2nd level) and become perfectly proficient at this rather than adopt a more complex system (3rd level) and spend a significant amount of time reaching mere competency.

  • For me, the Ben System is easier than “504= l+s/z+r= laser” because you don’t have to guess vowels. 504 is just “LOR” (or another sound, depending on the sounds assigned to digits). LOR could be Lori, lore, lord, Loren…

    I wonder if anyone has tried training children in a complex mnemonic system from an early age. If I ever have kids, I’ll try it… 🙂

  • @Dan: You are right about the amount of time it will take commit all the 10 000 images to memory. I’m still deciding whether to press on with it or go with a 2 digits system. Ben’s system or any phonetics system is not definitely not for me.

    If you take a look at the history of the WMC, most wins have been by 2 digits system with the exceptions of Andi Bell with PA0 (1000 images) and Ben. Clemens Mayer, Gunther Karsten, Wang Feng and O’Brien all mastered the 2 digits systems. I think Johannes Mallow is using some kind of phonetic systems.

    Well I’m looking into creating a 100 images system for 2 digits and then combine 4 digits to create a hybrid ingle image with 2 images. It will be far easier to memorise and recall and it will follow the same numerical system I’m using for binaries and cards.

  • @Yan: Sounds like you are looking for the Dominic System.

    I think the genius of the Dominic System is not fully appreciated. A person and an action is coextensive, while an object is never coextensive with the person or action. It does seems that the Dominic System is really a 4-digit system (1511 = Einstein playing tennis), while a PAO system is always tripartite (the object’s relationship with the action and the object’s relationship with the person). At best a PAO system is simply a PA+O system.

    Again, I believe a younger person (Dominic was in his 30’s and 40’s when he was competing) working at becoming perfectly proficient with the Dominic System could reach the outer limits of human memory performance. This tendency to move to more sophisticated techniques is not the way to go, IMO.

  • @Dan:

    Like I stated above, I tried the PAO which is an extended version of the Dominic System but I couldn’t quite get the actions down. Sometimes I find it easier to use a different action or interaction between the images to aid memorisation.

    I beg to differ though. Dominic system may be easy for memorisation and recall but I doubt that it is perfect for the hour phases.

    Someone which has mastered a 3 digits system will be able to recall more numbers with the same number of images than someone with a 2 digits system. That is why you are seeing developments of new and advanced techniques.

    You can’t say definitively that one system is better than the other. It all depends on the person using it.

  • Actually I got my first proper introduction to mnemotechnics and memory sports when I stumbled upon Dominic’s system. I was trying to come up with methods for quick study and memorization. I have studied his system in depth.

    I have to say that most people know only about PA in his system. But there are some other peculiarities too.

    1. He does not use any object.
    2. He does not use any recall. He does only one pass.
    3. He has only one Point of View during his journey. That is he sees his route points like through a television set. He doesn’t rotate or change the angle of view.
    4. He uses only one image at a route point with a Person and Action to memorize a card. I doubt he will break Ben’s records of 28 packs in 1 hour with this.

    You are wrong to assume that he was too old and not proficient at his own system. He worked on this system for 4 years before competing in the first WMC. He trained every day, logged every details, errors made, changed images, etc. In fact he created the system and he was WMC during 8 years with this. But pitch him now against a Ben, a Feng, a Mallow or even Gunther and I doubt he will make the top 5.

    To be fair though, his techniques about faces memorization are the best I’ve read. I’ve tried all sorts of techniques like coding skin color, eyebrows, noses,etc and even Lorayne’s method but none can beat Dominic’s face memorisation system for the moment.

    Plus he has got a very good imagination to be able to form images only once without recall. I’m incapable of doing that.

  • The Dominic System is great, but this is the reasoning I used when I switched:

    (It was a very difficult decision because I had already been using the Dominic System for 7 months.)

    • PA uses up a lot of journeys, between 26 and 52 loci. The Ben System fits one deck into 9 loci. Journeys are valuable, because they can’t be used that often due to ghost images.
    • I think Ben has a brilliant system for binaries and that is why he runs circles around most people in that event. One line of 30 binary digits fits in one locus.
    • When memorizing huge numbers or many decks with just PA, I think images will begin to repeat. Over the course of one competition, the images are bound to repeat across different events.
    • Dominic O’Brien is superhuman. I don’t think PA is the most efficient system, but Dominic can make it work. Or, another possibility is that he adds objects but didn’t tell his competitors about it in his books. I started adding objects to my Dominic System before I had ever heard of PAO, just to compress more data into the journeys. It seemed like a logical step.
    • I like having a lot of images, because I think it stretches my creativity.
    • I wanted to experiment with ideas about phonetic systems. It’s more difficult in the beginning, but I think it’s going to pay off in the end — not just for speed in memorizing numbers, but for memorizing other things too.

    In the end, a lot of it is probably about what system inspires a person enough to practice more than everyone else. 🙂

  • My mistake… I think images might not repeat much with a PA system. I was thinking of the problems I was having with repeated images, but I wasn’t using PA.

    With 2-digit PA there should be 10,000 possible image combinations, and with PAO, 1,000,000. Exact image combinations shouldn’t repeat much either way.

  • Josh wote:
    “I wonder if anyone has tried training children in a complex mnemonic system from an early age. If I ever have kids, I’ll try it… ”

    Yes, definitely. I would give anything if I had been forced to learn such a system at an early age.

    I am reading “Battle Hymn of the Tiger Mother” by Amy Chua who says Chinese mothers have it right and do better for their kids by forcing them to learn piano and violin, not allowing them to watch tv or play video games, and berating them if they get a bad grade in school (e.g., an A-).

  • How is Battle Hymn of the Tiger Mother? I’ve read some reviews of it and it seems like her methods were pretty brutal.

    I agree that it’s a good idea to keep kids from watching TV.

    I made a new page in the wiki for compiling ideas on How to Teach Memory Techniques to Children.

  • I’m surprised this post hasn’t had a comment for so long. Talking about so long, I worked out such systems years ago. Not that i’m about to share them. Being forced to do anything will result in failure. It’s been said that a man convinced against his will, is of the same opinion still. The same is true for a child or adult. A burning desire should be encouraged but, that’s not the same. You all have a burning desire. Interesting post. I’m surprised I haven’t stumbled across it before. p.s. Big system aren’t necessarily complicated nor do do they always need memorizing of 10,000 images. If the system is good then, it matters not if you memorize 10 random binaries or numbers or whatever. If the system is efficient, it is simple in terms of translation. I developed such a system. It is simple.

    p.s. Location is of paramount importance in all systems. The moment you enter a room you are conditioned by evolution to inspect the locations for threats, dangers, sex, food, water, etc. Breaking up a location into locus requires attention in a specific area. Nobody instinctively walks into a room and goes around selected objects that have been chosen as locus. It’s a conditioned response. Not even in their imagination. Everybody has to check the location out for threats, sex, etc.

  • Interesting that you guys are trying to improve the system, it’s great. I think I found a way to improve it, but it requires more investment time into it. What really interested me was Metrodorus of Scepsis how he used the Zodiac to remember word for word from speeches or books he read. In my mind I was like “that’s freaking impossible,” but when Yan was talking about increasing speed in memorization that really turned on a light bulb for me. Another light bulb was the shapes of the words from John. My thought was what if you changed words into shapes and then etc etc etc…idk anything is possible. Just a thought…

  • My system is an adjective + a noun using major system to create a 4-digit system
    1511 = Tall tide
    1662 = touchy chain
    6211 = Shiny tide
    For a six digit system, Adj+ Noun + Verb
    151138 = Tall tide move