Arbitrary Actions, or Person-Action-Object?

My memory system is based on Ben Pridmore’s but with a few modifications. The core idea is the same though: decimal digits are chunked in 3s, binaries in 10s, and cards in pairs. I also have a 2-digit system and a single-card system because I don’t think I will have my 2,652 double-card images ready by the USA Memory Championships in March.

One thing that I’m trying to figure out is whether it’s worth converting my two digit system into a person-action-object (PAO) system. I sometimes have trouble creating a wide enough variety of actions to go with my characters and objects.

La naissance de Vénus by Botticelli

84 = Aphrodite

If the first number is a can of tuna fish (12), the second number is an onion (22) and the third number is Aphrodite (84), what is the tuna fish doing to the onion and the onion doing to Aphrodite that preserves the order, and how do I make it unusual enough to be memorable?

If I changed it to a PAO system, it might be something like:

12 = [person] -> using a can opener -> can of tuna
21 = antelope -> hopping -> tall prairie grass
22 = a chef -> chopping -> an onion
84 = Aphrodite -> emerging from water -> shell

12-22-84 would then be a person chopping a shell.
84-21-12 would be Aphrodite hopping on a can of tuna fish
22-21-84 would be a chef hopping inside a shell
21-84-22 might be an antelope emerging from water with onions stuck to its horns

Photo of a pronghorn antelope

21 = Pronghorn antelope

I went back to read Ben’s description of his system and there is a useful clue:

Some of these objects are people, some are things. I ‘see’ them arranged from left to right, or top to bottom, and interacting in various ways according to rules I made up as I went along, depending on which objects come together in what order.

I’m not sure exactly what I’m going to do, but I might experiment with PAO for my two digit system, and use a loose association of actions with my three digit system. I like the idea of the loose association better than being locked into a strict set of PAO rules, but I don’t know if there is time to do it before March.

UPDATE: Part 2 is here.


  • When you’re recalling 2 or more images, you’ll never have to worry about “preserving the order”. That is, if you have the number 122284, you should use images that are assigned to each column (except for the ones and tens at the end).

    If you had a permanent image for (a) the one hundred thousand range (100,000-199,999–X-Man Wolverine) and a permanent image for (b) the twenty thousand range (20,000-20,999–Uncle Fester from the Addams family), and a permanent image for (c) the 2000 range (2,000-2999–the Green Goblin), and a permanent image for (d) the one hundred range (100-199–Superman), and a permanent image for the peg numbers 0-99, then you’d NEVER have to worry about preserving the order when you construct your story, because all the images are permanently assigned to ONLY ONE PLACE.

    122284 becomes 122,284. In your story, it would OK to put Wolverine at the end, because when you translate the images into numbers, Wolverine can ONLY go in the hundred thousand column as the number 1, therefore = 1–,—. The image at the end doesn’t mean it represents the last number.

    If Uncle Fester was the last image in the story, you KNOW that the number would look like this = -2-,—, because he’s the only image for the twenty thousand range.

    Or, you can set an image for 122 on the left (or the top of something) and that would remind you it’s the first group of 3, and then set 284 on the right (or the bottom of something) and that would remind you that it’s the second group of 3.

    On the left (or top) you’d see Batman (for the 100-199 range) fighting Mother Teresa (22 = nun).

    On the right (or bottom) you’d see Darth Vader (for the 200-299 range) with his cape on fire (84 = fire).

    Which means you’ll always know which position (ones, tens, hundreds, thousands, etc.) each image represents.

  • Have you tried that with really large numbers before? It seems like in a competition you would have the same time images repeating over and over. A 1,500-digit number has 250 six-digit chunks. With the method you describe, there would be 250 combinations of just 10 characters, which seems like it would be easy to get them mixed up.

    The 10 characters:
    100,000 (Wolverine)

  • Well, I have to admit that I wasn’t thinking in terms of memory competitions, merely shorter numbers (phone numbers, credit cards, ATM, etc.).

    On the other hand, even Dominic O’Brien’s system would call for repetition of images, seeing as how, in a very long number, there are bound to be repeats of pairs of numbers. And with a number of 1,500 digits, the chances of a six-digit chunk being repeated exactly are more remote than the chances of a two-digit repeat.

    However, even for a 1,500 digit number, there are plenty of ways to get around this. One obvious way is to set each six-digit number in its own place using either the journey method or a memory palace. This will keep them separate as you walk along looking at the images.

    If you have Wolverine for every six-digit number beginning with a “1”, it hardly risks confusing it with another six-digit number beginning with a “1” because (a) the digits after the “1” won’t likely be the same, (b) they images will be in separate locations.

    If you had 100 phone numbers with the same area code, what difference would that make? The seven digits following the area code will always be different, thus creating different images. Even if there’s ONLY a one-digit difference between two phone numbers, that one digit is sufficient to force a new image to appear, thus no confusion.

    In addition, remember that you ATTACH the six-digit story to a SPECIFIC location (journey or palace), so that it gets locked in place. Two of the same numbers attached to two DIFFERENT locations prevents confusion.

    Yes, there would be 250 combinations of just 10 characters, but each story would be different and each story would be set at a unique location. And don’t forget, with this method, you can constantly change the order of images. You don’t have to create a linear story that positionally corresponds with the number, since each image represents ONLY one column. If you have, say, 3 six-digit chunks that begin with “1”, you don’t have to place Wolverine at the beginning of the story—you can place him anywhere (since he can only represent the “1” in the sixth position). Thus, repetitions are either avoided or kept to a minimum. Besides, they’re always attached to a unique location.

  • The whole point of mnemonics has always been to REDUCE a large amount of information into a SMALL piece (or pieces) of information.

    If you create a unique image for every single number from 1-10,000 (or beyond), then first, I don’t think you’ll be able to get them into your long-term memory, and, second, you’re simply INCREASING the information instead of reducing it.

    Yates, Carruthers, Lorayne, O’Brien–all those who chronicle or create mnemonics agree that it’s the collapse of “more” into “less” that’s one of the fundamental keys, one of the reasons FOR the systems and one of the reasons the systems work.

    You have (or plan to have) 10,000 images. For the same number range (1-10,000), I never have to use more than 4 images, and sometimes only 3, after having memorized only about 128 images. In other words, I can leverage 128 images into any number between 1 and 10,000.

    Let’s extrapolate. For the number range 1-1,000,000, you’re going to need one million images. I’ll need only about 138.

  • Interesting idea… Dominic O’Brien’s system does have some repeated numbers, but they aren’t too common. If the Dominic System is expanded to a PAO system, the repeats are even fewer.

    For competitions, I’m working on the Ben System, which fits 9 digits in each locus (30 digits for binaries). I only need 1,000 images for 1 billion numbers: 000,000,000 to 999,999,999…

  • Ok has anyone done something like this / would this work?
    P A O O, so each 2 digit number is represented by a person doing an action to some object connected to another object?
    This allows you to code 8 digits to 1 locus (2 for P, 2 for A, 2 for O, another 2 for the other O) instead of the regular 6 digits with normal PAO, but you only need to remember an additional 100 items in your image strings. In this way, it is better than a 3 digit PAO system because you don’t need to remember 1000 things, however you can only code 8 digits to each locus instead of 9 digits.

    22 = Darth Vader kissing a pumpkin with Pikachu on it
    36 = Buzz Lightyear punching a cat with a Charizard on it
    91 = Obama throwing a basketball with a (some pokemon) on it
    40 = Santa Claus licking a bowling pin with a (some pokemon) on it
    Then 36409122 = Buzz Lightyear licking a basketball with a Pikachu on it
    So now you are turning 8 digits into 1 picture, but you only need to remember 100 strings of images. Pretty cool: you can memorize numbers up to 100 million with very little infrastructure!
    Please let me know if this would work, thanks a lot.

  • I recommend asking your question in the Forum. I think that Larry has experimented with different systems like that:

  • what about attaching a color to the person and a color to the object for for an additional two digits to one image?

  • You could give it a try and then post in the forum about how it works…