The Pyramid Memory System
Today I am presenting the thought process behind my Millennium PAO (3-2-3 Person-Action-Object) and how upgradeable it is. Instead of following any kind of letter code, like most popular number systems do, I went an undiscovered new path. My goal was to create a system where I have fully control over all images without being limited by any letter code. It took me over seven years to develop it and I gave every angle of it a lot of thought. The result is a modular build category system.
I call this approach the Pyramid Memory System because it is like building a pyramid growing bigger and bigger with each layer.
The Different Letter Codes
In the beginning of every memory athlete’s career stands the decision of what number system they will use. The most popular choice is the Major System, a phonetic letter code that works in all alphabet-based languages. That was also my first 2-digit number system. Other options are named after three big names in memory sports, which are the Dominic System (Dominic O’Brien), the Ben System (Ben Pridmore) and the Simon System (Simon Reinhard).
While the Dominic System is useful for encoding a 2x2-digit PA system with 200 images, both Ben and Simon System are meant for higher ambitions. The Ben System is designed to create a 3-digit memory system with 1.000 images and the Simon System, which is very similar to the Ben System, is even going up to a 4-digit memory system with 10.000 images. All these four systems are based on specific letter codes, which translate decimal digits into one or several letters and letter-pairs of the alphabet.
After learning one of these codes, the user will soon be able to directly “read” numbers like words. That is a powerful feature which makes it very easy to trigger the associated image, terminating the necessity to remember them in the first place. Especially the Ben and Simon System are glorious with that, since there are no “hidden” letters like in the Major System. A 711 is simply a cat in the Ben System, while the Major System makes a cadet out of it. The cadet requires to be learned first, although the code makes that process fairly easy. That cat on the other hand can be translated directly without any learning, only the code needs to be understood. The letter code will allow us to quickly create a sophisticated memory system of 1.000 images and more.
The downside of all these systems is that they are completely random. While 711 is a cat or cadet, 712 would be a can (Ben System) or cotton (Major System) for example. For me the cat and the can have very little in common, just like the cadet and the cotton. That is the first aspect that I have a problem with. I like my memory systems very organized and structured. Another aspect I dislike about using a letter code is that not all images are as nice or simple as a cat or can. There are dozens of words dictated by the code which are a) very abstract and complicated or b) simply boring and stupid. And sometimes there is even a c) where the code creates words that don’t exist. Try to find a 667 with the Major System. And what is a noor (204 in the Ben System)? You’ll have to be creative and bend the rules.
The Category System as an alternative
To reduce or even avoid randomness and ugly images there is a nice alternative. The idea is to create categories with memorable, interesting and/or useful themes and fill them with fitting images. Using a category system gives one a much better control over all used images. On top of that comes that a category system is limitless upgradable (see Wardrobe Method). To create a 3-digit category system there are normally, besides many more chaotic forms, two major size options:
10 Categories with 100 Images, using a letter code
One option is using 10 categories with one hundred images each. This approach gives full control over the categories but still limits image control. Within each category a letter code is used again, to learn the group easily. The first digit could determine the category by either using a letter code or not while the second and third digits are encoded by the Major System. Memory athlete Katie Kermode from England did exactly that approach and created a 3-digit system. She can use her mixed set of together 1.000 people and objects similar to a PAO system by combining two images with an action in the middle. That creates a 3-2-3 system, which I will refer to as Katie System from now on.
It is a great alternative between image control and difficulty. Katie can still read her images and only need to differentiate between ten categories. That makes her system complex to create but easy to learn.
A category-based system seems natural to me because all the images are grouped logically (making recall much easier) and you can learn a lot of interesting real-life information. Like 100 musical instruments, for example!
Using the Major System to encode the images within each category helps me recognise the images easily, because then they’re actually linked to the number in some way.Katie Kermode
100 Categories with 10 Images, using a letter code
Using 100 categories instead, with 10 images each, allows for more control with a broader variation of themes at the same time. And it is easier to find 10 images for one category than 100.
The categories themselves can be created using a letter code again, for example 92 = Bunny (Major Code) creates the Looney Toons category (because of Bugs Bunny), which is easy to learn but will limit the range of possible categories at the same time. Within each category the images can be ordered by their first letter, using the Major System, for example Sylvester J. Pussycat (S=0) and Bugs Bunny (B=9). This approach has been created by Michael Gloschewski, the founder of the popular MemoCamp software. I will refer to it as the Michael System from now on.
The “crazy” third option: no code at all
For me both approaches are not going far enough. I want total control over my categories in the exact order and size as I please. And I want total control of my images within each category as well. So I ultimately decided to pass on any code and learn my images with memory techniques instead. I was very aware that this approach is way more complicated and requires a lot of training to achieve the same with it compared to a letter system. But memory sports has shown me that in the end all that matters is to have your images available when you need them. And in that moment I want to have the best images I could think of. Since I am planning on staying in this sport for the rest of my life, I can and should afford to invest all the time it takes to get the system I am 100% satisfied with.
The most important argument for only using perfect images is the fact that memory sports is a modern form of Alchemy, a transmutation of mind. Our brain is constantly rewiring throughout our lives. Using memory techniques all the time is influencing this process heavily in my opinion. Taking full control of the imagery being used in this development is my own responsibility. Therefore I embrace the extra effort and think in the long run, to accomplish something very powerful and personal: my perfect memory system.
The 1.000 People System
How I created the 1.000 People
So I wanted to build a category based 3-digit PAO system without using a letter code. I started to create my 1.000 people list first. First of all I created 13 major categories and filled them with sub-categories of tens.
- Video Games
- Super Heroes
- TV Shows
Since I had full control I wanted to pick only the characters I found interesting and memorable. So I ended up with an odd 13 major groups, some of them bigger than one hundred and some much smaller. I ordered them to my own logic, trying to make them harmonious. Within each sub-category I was often putting the main character first, followed by the side-kick and ending with an evil or special character. I wanted to understand each sub-category by its mere logic before I even have started to memorize it.
I did spent years to collect all my characters and order them perfectly together, coupled with researching the most brilliant digital image I could find for each one of them. But even though I took my time, this was still the easy part of it. To memorize them I had to find a way to use my old 2-digit memory system. How could I reuse my 100 Major System images over and over again, to learn my new 1.000 people?
How learned the 1.000 People
The first idea was using a certain keyword for each major category and then couple that with the Major System image plus the new person. The keyword for Cartoons would have become a pencil, for example. Then the pencil and 78 = coffee would be associated with the specific cartoon character. This can be referred to as Keyword Method. The problem with this is that I would use the pencil in one hundred stories. Those are a lot of repetitions that would probably create confusion and would be pretty monotonous as well.
The second idea was using a keyword for each sub-category. That would create only ten repetitions for each keyword, a number that does sound much more reasonable. But learning all these one hundred keywords without a code would have been tricky. Possible, though, but I really wanted to avoid that. I could have associated them with my basic Major System, but that didn’t feel right either because it would have forced me to learn and always practise using one hundred images I actually don’t need for my system.
The third idea was the right one for me: The number 678 would be created with the 67 and the 78, overlapping the middle digit. That is the so-called Overlap Method. I decided to use my old 2-digit action list for the 100 sub-categories and my old 2-digit object list for the one hundred images within each group of one hundred. Now I was able to use my old system, without learning any unnecessary images. In combination with that new person I was about to memorize, these three elements created a PAO, something I was used doing anyway for a very long time. The repetitions were just ten times for each action (ten in a row) as well as ten times for each object (every one hundred).
Example 1: 050
My action for 05 = kicking. My object for 50 = lasso. Kicking triggers the category “Hannah-Barbara Productions”. Knowing the sub-category helps a lot already. But the main trigger is the story itself, because both action and object together create a question and memory hook:
Who is kicking the lasso? Papa Smurf.
Example 2: 051
We are still using the same action 05 = kicking as above. The object for 51 = latte (German for lath). The category is therefore still “Hannah-Barbara Productions”. The question is now:
Who is kicking the lath? Fred Flintstone.
Example 3: 061
The sub-category moves now from 05 to 06 = guarding, triggering “Filmation Production”. The object 61 = shadow.
Who is guarding a shadow? Marshall BraveStarr.
Example 4: 161
Here we have switched to a new major category: Disney. The sub-category is 16 = painting which triggers Disney Afternoon. The object is again 61 = shadow.
Who is painting a shadow? Toadwart.
The 1.000 Object System
How I created the 1.000 Objects
For the objects I used more or less the same structure. First I created the ten major categories. For that I decided to pick subjects that will either teach me something or are just great object categories for memory sports:
- Periodic Table of Elements
- Countries of the West/North
- Countries of the East/South
- Commodity Items
- Toys and Instruments
- Sci-Fi and Fantasy Items
- Esoteric Wisdom
As you can see, I used exactly ten major categories. That was easier compared to the people because I was less picky. Within those I tried to create sub-categories of tens, when it was possible. Some of them don’t even have sub-categories like the periodic table but later they will merge with the people sub-categories anyway.
For the abstract lists like the periodic table of elements, the countries and especially the esoteric wisdom I used replacement images. All elements are represented by their most common use. The countries are represented by a famous sight like a building or statue, for example. Some parts of the list of esoteric wisdom seem a bit crazy and super difficult for memory sports. Well, I will make it work for me.
How I learned the 1.000 Objects
That was probably the easiest part of the entire system. All I did to memorize them was weaving them into my new PAO stories, which I created to learn the 1.000 people. That made my original stories even more memorable and helped me to learn the entire system much better.
Example 1: 050
Remember the story from above: Papa Smurf kicking a lasso. Now I add the new object, which is tin, the 50th element in the periodic table of elements, represented by a tin can. The new story is now:
Papa Smurf kicks a lasso into a tin can.
Example 2: 051
Remember the story from above: Fred Flintstone kicking a lath. Now I add the 51st element antimony, represented by mascara. The new story is now:
Fred Flintstone kicks a lath, which catapults mascara into his eye.
Example 3: 061
Remember the story from above: Marshall BraveStarr guards a shadow. The 61st element is promethium, which is used for satellites. The new story is now:
Marshall BraveStarr guards the shadow of a satellite.
Example 4: 161
Remember the story from above: Toadwart paints a shadow. Here we are now in the major category for food. The sub-category for 16 is sweets. The new object is popcorn. The new story is now:
Toadwart paints popcorn onto his shadow.
The Millennium PAO is ready
Since I find 1.000 actions a bit impractical and a 3-2-3 PAO gives me a nice 8-digit memory cluster, I have finished my new Millennium PAO system – for now. 8 digits in one image fit perfectly into a row of 40 digits, which are the standard at all traditional World Memory Sports Council championships. I can also memorize Historical Dates now with a 3-digit image (either person or object), which is the best way to do it. The hard part is now to get fast with this system.
I am using the free spaced repetition flashcard software Anki. To get really fast with my system I am doing several different training methods. One is to practise the 100 categories. Another one is to practise the new 3-digit images directly, which means all 1.000 stories I created for the new people and objects. Another one creates virtual sub-categories where I couple all object-stories that use the same 2-digit object, for example all lasso stories: 050, 150,…,950. That last training gives me a nice interconnection between similar numbers.
At the moment I am already able to trigger half of my system in a split second, just looking at the 3-digit number. I neither need the category nor the story anymore. The other half will be that fast as well sooner or later. I confuse my categories very little. With more practise I am certain that I can reduce that close to zero.
The DoubleCard System
The Logic of Poker Cards
Since that is done, I can now create a DoubleCard memory system out of my 2.000 images (I don’t include my actions for this). I need to add another 704 images to create a true 2.704 DoubleCard system. But what is the smartest way to do that, with my existing system? In fact, using a category system with 1.000 people and 1.000 objects is close to perfect for that task. To understand that, we need to look at the logic of Poker cards:
- There are 52 cards
- There are four suits with 13 cards each:
- There are ten 1-digit number index cards (when we assign the Ace as a 1 and the 10 as a 0) in each suit:
- 10 = 0
- Ace = 1
- There are 3 people cards in each suit:
Re-using the number sub-categories
When we couple two cards with each other, each first card has 52 possible pairs with the second card. Dividing these 52 into to the four suits creates 13 cards each. Altogether there are 208 of these groups with 13 cards each (52 cards x 4 suits = 208). 208 card categories x 13 cards = 2.704 possibilities. I call these 208 groups the card categories.
Looking at the fact that there are already these ten 1-digit number cards, we can assign these directly to the according numbers within each of our 100 sub-categories. When we assign all 1.000 people to 100 of these new card categories, we can simply add the three missing people for the Jack, Queen and King in each category. That is awesome because we stay within the logic of our Millennium PAO categories. Nothing changes, except that we increase our people by 100 (categories) x 3 (Jack, Queen, King) = 300.
We do the same thing for all 1.000 objects and fill another 100 of these new card categories. Again we stay exactly in our previous categories and just add the missing three images for Jack, Queen and King. When we are done with that, we have updated our system to 1.300 people and 1.300 objects and filled 200 out of 208 card categories.
For the remaining 8 card categories we just invent 8 new categories from scratch and fill them: four with people and four with objects. Now we have images for all 2.704 DoubleCard possibilities.
The first 52 card categories
That sounded very complicated, didn’t it? It probably is, but think about how perfectly it fits? But it gets better – and more complicated as well. The next step is to assign these 208 card categories in the best way. And here comes some simple math into play. We can actually calculate the value of each of the 208 card categories. We use the 1-digit values of the number index of the first card. We multiply that always by four, for the four different suits, no matter which suit is on the first card. Now we look at the second card and add the value of that suit. I am referring here to the common method of translating Spades into 1 (because it looks like a tree or leave with a straight central line), Hearts into 2 (because the right outer border looks exactly like part of a two), Clubs into 3 (because it has three bubbles) and Diamonds into 4 (because it has 4 pointy ends).
The total result is a number between 01 and 40. So far we have covered all number index cards for the first card (0-9). We translate now the Jack into 11, Queen into 12 and King into 13 and do the same kind of calculation. That leads us to 52. From then on it would get complicated. So let’s keep the math simple, right? We will therefore only count up until 50. Doing that with the first 50 categories of people, using the first suit value 1 = Spades for the suit of the first card, we assign a rough quarter of all our 208 card categories. These first 50 are very easy to read (with a bit of practise). The remaining 2 categories that we cut out so far are the King of Spades (1st card) + all Clubs and all Diamonds (2nd card). For these two card categories we assign two of our brand new 8 card categories, where all 13 images each are entirely new. Since we started with the people, we use two of the new people categories.
0 (1-digit value for the 10) x 4 (four different suits) + 1 (value of the Spades of the 2nd card) = 01
7 x 4 + 2 (value of Hearts) = 30
11 (value for the Queen) x 4 + 3 (value of Clubs) = 47
12 (value for the King) x 4 + 4 (value of Diamonds) = 52; this one is higher than 50 and therefore is using one of the brand new card categories with people
The second 52 card categories
The next 50 card categories should now use the second part of my people sub-categories from 51-00. They could start with the suit of Hearts (value = 2). But since that is now a red card, compared to the black Spades, I decided to use the black Clubs instead (value = 3). That makes it easy to identify whether we are using people or objects, because all people now start with a black card, while all objects start with a red card. The math we do now is staying the same but we add 50 to the result. For the 101 and 102 we use the two remaining new people card categories.
0 (1-digit value for the 10) x 4 (four different suits) + 1 (value of the Spades of the 2nd card) + 50 = 51
7 x 4 + 2 (value of Hearts) + 50 = 80
11 (value for the Queen) x 4 + 3 (value of Clubs) + 50 = 97
12 (value for the King) x 4 + 4 (value of Diamonds) + 50 = 102; this one is higher than 100 and therefore is using one of the brand new card categories with people
The remaining 104 card categories
We have now assigned exactly 50% of the entire image set, using all and only the people. We repeat the same logic from the beginning, but using the red suits Hearts (01-50) and Diamonds (51-00) and assign now the objects. For the remaining four categories we will use the four brand new object card categories. The entire thought process is now the following:
- Looking at the suit of the first card: black is a person, red is an object
- Calculating the worth of both cards: the result gives the category
- Checking the index of the second card: the index gives the exact image within that category
Training the DoubleCard system
Similar to the Millennium PAO I use Anki to learn and practise that system. I created 208 flashcards were I train the card categories by translating 1st card + the suit of the 2nd card. Additionally I practise all single images step by step, beginning with the 600 new images for the Jacks, Queens and Kings, followed by the eight entirely new card categories with their 104 images.
The rest of the 2.000 images are fairly easy but the card-image-recognition needs to be practised for high speed as well. Another thing that should be practised is the backwards translation, which is not very difficult but should be super fast as well.
The Sky is the limit
The future of the Pyramid System
The great upgradeability leads me to an even bigger dream: A true 3-3-4 PAO number system with 10 digits in one image cluster. That requires upgrading my 100 actions to 1.000 actions after all. That will be tricky but it should be possible. The hard part will be the 10.000 objects. Following the logic of my category system I would need another 9.000 objects packed into 900 sub-categories with ten objects each. An extreme endeavour but who says I need to be done next year? I have all the rest of my life to finish this and learn amazing things by doing so. Even if I never finish this, it would be worth to start it.
Well, that’s basically it. If you would like to see my entire Millennium PAO and my DoubleCard System in imagery, you can view my two galleries I have created for them. Thanks for staying that long with me and getting an insight into one of the craziest memory systems ever created.
This article and all images in it are for educational purposes only!!
Copyright Disclaimer under section 107 of the Copyright Act 1976, allowance is made for “fair use” for purposes such as criticism, comment, news reporting, teaching, scholarship, education and research.
Fair use is a use permitted by copyright statute that might otherwise be infringing.
Non-profit, educational or personal use tips the balance in favour of fair use.