by Don Norman
The classic study of memory for epic poetry was done by Albert Bates Lord. In the mid-1900s he traveled throughout the former Yugoslavia (now a number of separate, independent countries) and found people who still followed the oral tradition. He demonstrated that the “singer of tales,” the person who learns epic poems and goes from village to village reciting them, is really re-creating them, composing poetry on the fly in such a way that it obeys the rhythm, theme, story line, structure, and other characteristics of the poem. This is a prodigious feat, but it is not an example of rote memory.
The power of multiple constraints allows one singer to listen to another singer tell a lengthy tale once, and then after a delay of a few hours or a day, to recite “the same song, word for word, and line for line.” In fact, as Lord points out, the original and new recitations are not the same word for word, but both teller and listener perceive them as the same, even when the second version was twice as long as the first. They are the same in the ways that matter to the listener: they tell the same story, express the same ideas, and follow the same rhyme and meter. They are the same in all senses that matter to the culture. Lord shows just how the combination of memory for poetics, theme, and style combines with cultural structures into what he calls a “formula” for producing a poem perceived as identical to earlier recitations.
The notion that someone should be able to recite word for word is relatively modern. Such a notion can be held only after printed texts become available; otherwise who could judge the accuracy of a recitation? Perhaps more important, who would care?
All this is not to detract from the feat. Learning and reciting an epic poem, such as Homer’s Odyssey and Iliad, is clearly difficult even if the singer is re-creating it: there are twenty-seven thousand lines of verse in the combined written version. Lord points out that this length is excessive, probably produced only during the special circumstances in which Homer (or some other singer) dictated the story slowly and repetitively to the person who first wrote it down. Normally the length would be varied to accommodate the whims of the audience, and no normal audience could sit through twenty-seven thousand lines. But even at one-third the size, nine thousand lines, being able to recite the poem is impressive: at one second per line, the verses would take two and one-half hours to recite. It is impressive even allowing for the fact that the poem is re-created as opposed to memorized, because neither the singer nor the audience expect word-for-word accuracy (nor would either have any way of verifying that).
Most of us do not learn epic poems. But we do make use of strong constraints that serve to simplify what must be retained in memory. Consider an example from a completely different domain: taking apart and reassembling a mechanical device. Typical items in the home that an adventuresome person might attempt to repair include a door lock, toaster, and washing machine. The device is apt to have tens of parts. What has to be remembered to be able to put the parts together again in a proper order? Not as much as might appear from an initial analysis. In the extreme case, if there are ten parts, there are 10! (ten factorial) different ways in which to reassemble them—a little over 3.5 million alternatives.
But few of these possibilities are possible: there are numerous physical constraints on the ordering. Some pieces must be assembled before it is even possible to assemble the others. Some pieces are physically constrained from fitting into the spots reserved for others: bolts must fit into holes of an appropriate diameter and depth; nuts and washers must be paired with bolts and screws of appropriate sizes; and washers must always be put on before nuts. There are even cultural constraints: we turn screws clockwise to tighten, counterclockwise to loosen; the heads of screws tend to go on the visible part (front or top) of a piece, bolts on the less visible part (bottom, side, or interior); wood screws and machine screws look different and are inserted into different kinds of materials. In the end, the apparently large number of decisions is reduced to only a few choices that should have been learned or otherwise noted during the disassembly. The constraints by themselves are often not sufficient to determine the proper reassembly of the device—mistakes do get made—but the constraints reduce the amount that must be learned to a reasonable quantity. Constraints are powerful tools for the designer: they are examined in detail in Chapter 4.
Memory Is Knowledge in the Head
An old Arabic folk tale, “‘Ali Baba and the Forty Thieves,” tells how the poor woodcutter ‘Ali Baba discovered the secret cave of a band of thieves. ‘Ali Baba overheard the thieves entering the cave and learned the secret phrase that opened the cave: “Open Simsim.” (Simsim means “sesame” in Persian, so many versions of the story translate the phrase as “Open Sesame.”) ‘Ali Baba’s brother-in-law, Kasim, forced him to reveal the secret. Kasim then went to the cave.
When he reached the entrance of the cavern, he pronounced the words, Open Simsim!
The door immediately opened, and when he was in, closed on him. In examining the cave he was greatly astonished to find much more riches than he had expected from ‘Ali Baba’s relation.
He quickly laid at the door of the cavern as many bags of gold as his ten mules could carry, but his thoughts were now so full of the great riches he should possess, that he could not think of the necessary words to make the door open. Instead of Open Simsim! he said Open Barley! and was much amazed to find that the door remained shut. He named several sorts of grain, but still the door would not open.
Kasim never expected such an incident, and was so alarmed at the danger he was in that the more he endeavoured to remember the word Simsim the more his memory was confounded, and he had as much forgotten it as if he had never heard it mentioned.
Kasim never got out. The thieves returned, cut off Kasim’s head, and quartered his body. (From Colum’s 1953 edition of The Arabian Nights.)
Most of us will not get our head cut off if we fail to remember a secret code, but it can still be very hard to recall the code. It is one thing to have to memorize one or two secrets: a combination, or a password, or the secret to opening a door. But when the number of secret codes gets too large, memory fails. There seems to be a conspiracy, one calculated to destroy our sanity by overloading our memory. Many codes, such as postal codes and telephone numbers, exist primarily to make life easier for machines and their designers without any consideration of the burden placed upon people. Fortunately, technology has now permitted most of us to avoid having to remember this arbitrary knowledge but to let our technology do it for us: phone numbers, addresses and postal codes, Internet and e-mail addresses are all retrievable automatically, so we no longer have to learn them. Security codes, however, are a different matter, and in the never-ending, escalating battle between the white hats and the black, the good guys and the bad, the number of different arbitrary codes we must remember or special security devices we must carry with us continues to escalate in both number and complexity.
Many of these codes must be kept secret. There is no way that we can learn all those numbers or phrases. Quick: what magical command was Kasim trying to remember to open the cavern door?
How do most people cope? They use simple passwords. Studies show that five of the most common passwords are: “password,” “123456,” “12345678,” “qwerty,” and “abc123.” All of these are clearly selected for easy remembering and typing. All are therefore easy for a thief or mischief-maker to try. Most people (including me) have a small number of passwords that they use on as many different sites as possible. Even security professionals admit to this, thereby hypocritically violating their own rules.
Many of the security requirements are unnecessary, and needlessly complex. So why are they required? There are many reasons. One is that there are real problems: criminals impersonate identities to steal people’s money and possessions. People invade others’ privacy, for nefarious or even harmless purposes. Professors and teachers need to safeguard examination questions and grades. For companies and nations, it is important to maintain secrets. There are lots of
reasons to keep things behind locked doors or password-protected walls. The problem, however, is the lack of proper understanding of human abilities.
We do need protection, but most of the people who enforce the security requirements at schools, businesses, and government are technologists or possibly law-enforcement officials. They understand crime, but not human behavior. They believe that “strong” passwords, ones difficult to guess, are required, and that they must be changed frequently. They do not seem to recognize that we now need so many passwords—even easy ones—that it is difficult to remember which goes with which requirement. This creates a new layer of vulnerability.
The more complex the password requirements, the less secure the system. Why? Because people, unable to remember all these combinations, write them down. And then where do they store this private, valuable knowledge? In their wallet, or taped under the computer keyboard, or wherever it is easy to find, because it is so frequently needed. So a thief only has to steal the wallet or find the list and then all secrets are known. Most people are honest, concerned workers. And it is these individuals that complex security systems impede the most, preventing them from getting their work done. As a result, it is often the most dedicated employee who violates the security rules and weakens the overall system.
When I was doing the research for this chapter, I found numerous examples of secure passwords that force people to use insecure memory devices for them. One post on the “Mail Online” forum of the British Daily Mail newspaper described the technique:
When I used to work for the local government organisation we HAD TO change our Passwords every three months. To ensure I could remember it, I used to write it on a Post-It note and stick it above my desk.
How can we remember all these secret things? Most of us can’t, even with the use of mnemonics to make some sense of nonsensical material. Books and courses on improving memory can work, but the methods are laborious to learn and need continual practice to maintain. So we put the memory in the world, writing things down in books, on scraps of paper, even on the backs of our hands. But we disguise them to thwart would-be thieves. That creates another problem: How do we disguise the items, how do we hide them, and how do we remember what the disguise was or where we put it? Ah, the foibles of memory.
Where should you hide something so that nobody else will find it? In unlikely places, right? Money is hidden in the freezer; jewelry in the medicine cabinet or in shoes in the closet. The key to the front door is hidden under the mat or just below the window ledge. The car key is under the bumper. The love letters are in a flower vase. The problem is, there aren’t that many unlikely places in the home. You may not remember where the love letters or keys are hidden, but your burglar will. Two psychologists who examined the issue described the problem this way:
There is often a logic involved in the choice of unlikely places. For example, a friend of ours was required by her insurance company to acquire a safe if she wished to insure her valuable gems. Recognizing that she might forget the combination to the safe, she thought carefully about where to keep the combination. Her solution was to write it in her personal phone directory under the letter S next to “Mr. and Mrs. Safe,” as if it were a telephone number. There is a clear logic here: Store numerical information with other numerical information. She was appalled, however, when she heard a reformed burglar on a daytime television talk show say that upon encountering a safe, he always headed for the phone directory because many people keep the combination there. (From Winograd & Soloway, 1986, “On Forgetting the Locations of Things Stored in Special Places.” Reprinted with permission.)
All the arbitrary things we need to remember add up to unwitting tyranny. It is time for a revolt. But before we revolt, it is important to know the solution. As noted earlier, one of my self-imposed rules is, “Never criticize unless you have a better alternative.” In this case, it is not clear what the better system might be.
Some things can only be solved by massive cultural changes, which probably means they will never be solved. For example, take the problem of identifying people by their names. People’s names evolved over many thousands of years, originally simply to distinguish people within families and groups who lived together. The use of multiple names (given names and surnames) is relatively recent, and even those do not distinguish one person from all the seven billion in the world. Do we write the given name first, or the surname? It depends upon what country you are in. How many names does a person have? How many characters in a name? What characters are legitimate? For example, can a name include a digit? (I know people who have tried to use such names as “h3nry.” I know of a company named “Autonom3.”)
How does a name translate from one alphabet to another? Some of my Korean friends have given names that are identical when written in the Korean alphabet, Hangul, but that are different when transliterated into English.
Many people change their names when they get married or divorced, and in some cultures, when they pass significant life events. A quick search on the Internet reveals multiple questions from people in Asia who are confused about how to fill out American or European passport forms because their names don’t correspond to the requirements.
And what happens when a thief steals a person’s identity, masquerading as the other individual, using his or her money and credit? In the United States, these identity thieves can also apply for income tax rebates and get them, and when the legitimate taxpayers try to get their legitimate refund, they are told they already received it.
I once attended a meeting of security experts that was held at the corporate campus of Google. Google, like most corporations, is very protective of its processes and advanced research projects, so most of the buildings were locked and guarded. Attendees of the security meeting were not allowed access (except those who worked at Google, of course). Our meetings were held in a conference room in the public space of an otherwise secure building. But the toilets were all located inside a secure area. How did we manage? These world-famous, leading authorities on security figured out a solution: They found a brick and used it to prop open the door leading into the secure area. So much for security: Make something too secure, and it becomes less secure.
How do we solve these problems? How do we guarantee people’s access to their own records, bank accounts, and computer systems? Almost any scheme you can imagine has already been proposed, studied, and found to have defects. Biometric markers (iris or retina patterns, fingerprints, voice recognition, body type, DNA)? All can be forged or the systems’ databases manipulated. Once someone manages to fool the system, what recourse is there? It isn’t possible to change biometric markers, so once they point to the wrong person, changes are extremely difficult to make.
The strength of a password is actually pretty irrelevant because most passwords are obtained through “key loggers” or are stolen. A key logger is software hidden within your computer system that records what you type and sends it to the bad guys. When computer systems are broken into, millions of passwords might get stolen, and even if they are encrypted, the bad guys can often decrypt them. In both these cases, however secure the password, the bad guys know what it is.
The safest methods require multiple identifiers, the most common schemes requiring at least two different kinds: “something you have” plus “something you know.” The “something you have” is often a physical identifier, such as a card or key, perhaps even something implanted under the skin or a biometric identifier, such as fingerprints or patterns of the eye’s iris. The “something you know” would be knowledge in the head, most likely something memorized. The memorized item doesn’t have to be as secure as today’s passwords because it wouldn’t work without the “something you have.” Some systems allow for a second, alerting password, so that if the bad guys try to force someone to enter a password into a system, the individual would use the alerting one, which would warn the authorities of an illegal entry.
Security poses major design iss
ues, ones that involve complex technology as well as human behavior. There are deep, fundamental difficulties. Is there a solution? No, not yet. We will probably be stuck with these complexities for a long time.
The Structure of Memory
Say aloud the numbers 1, 7, 4, 2, 8. Next, without looking back, repeat them. Try again if you must, perhaps closing your eyes, the better to “hear” the sound still echoing in mental activity. Have someone read a random sentence to you. What were the words? The memory of the just present is available immediately, clear and complete, without mental effort.
What did you eat for dinner three days ago? Now the feeling is different. It takes time to recover the answer, which is neither as clear nor as complete a remembrance as that of the just present, and the recovery is likely to require considerable mental effort. Retrieval of the past differs from retrieval of the just present. More effort is required, less clarity results. Indeed, the “past” need not be so long ago. Without looking back, what were those digits? For some people, this retrieval now takes time and effort. (From Learning and Memory, Norman, 1982.)
Psychologists distinguish between two major classes of memory: short-term or working memory, and long-term memory. The two are quite different, with different implications for design.
SHORT-TERM OR WORKING MEMORY
Short-term or working memory (STM) retains the most recent experiences or material that is currently being thought about. It is the memory of the just present. Information is retained automatically and retrieved without effort; but the amount of information that can be retained this way is severely limited. Something like five to seven items is the limit of STM, with the number going to ten or twelve if the material is continually repeated, what psychologists call “rehearsing.”
Multiply 27 times 293 in your head. If you try to do it the same way you would with paper and pencil, you will almost definitely be unable to hold all the digits and intervening answers within STM. You will fail. The traditional method of multiplying is optimized for paper and pencil. There is no need to minimize the burden on working memory because the numbers written on the paper serve this function (knowledge in the world), so the burden on STM, on knowledge in the head, is quite limited. There are ways of doing mental multiplication, but the methods are quite different from those using paper and pencil and require considerable training and practice.