Volume 14, Number 7
This seems to be a widely-shared sentiment among Roediger's colleagues in the field of memory research.
APS Fellow Randall W. Engle, chair of the Georgia Institute of Technology School of Psychology, says he can count "up 40 or 50 user names and passwords that I have available to me for various online stores. If you do what the security experts tell you to do, you'd end up with a huge number of passwords to remember."
"We can remember thousands of words," points out Alan Baddeley, of the Department of Psychology at the University of Bristol, United Kingdom. "Presumably, we could devote large amounts of time to learning numbers as well, but that wouldn't be a particularly useful way of spending one's life."
"Many people have memory spans that are noticeably different from the mean," says APS Fellow and Charter Member Saul Sternberg of the University of Pennsylvania, "but the answer is probably that on average we are quite close to the limit, if we haven't already exceeded it."
"The mind of the beast trying to measure itself as if from outside of the beast is always a tricky business," cautions Engle, who has devoted 20 years to studying individual differences in working memory capacity.
"Seeing how much you can recall immediately is not what's important here," he says. "Interference is really the problem. When you go to retrieve the stored information, particularly if you haven't used it in a while, you have interference from all the things that you have learned in similar contexts over the years. We know that people vary greatly in their vulnerability to such interference."
Dealing with such interference, Engle says, is the job of the brain's "executive attention system" (a three-member brain-trust composed of the prefrontal cortex, anterior singulate and locus ceruleus), which is important to such activity as planning or maintaining information needed for a given task. "We think that, to the extent that you have more executive attention capability you are better able to resist interference and retrieve a phone number or a PIN or a password on demand."
In one telling study, Engle instructed his subjects to look at the center of a computer screen. When a signal flickered on one side, the eye normally was attracted there and the subjects recorded the "R," "B," or "P" that flashed on more briefly than the blink of an eye.
"All subjects scored about the same," he says, "but when they were tested on their ability to resist the flicker and see the letter on the opposite side of the screen, they varied widely. Those who were low in executive attention capability were more likely to look to the flashing cue instead of where the letter was going to pop on, and once they did that, they were slower to get back to the correct side and identify the letter."
In other words, they were less able to weed out the interference. This is much the same process the mind uses to retrieve a telephone number from the myriad bits of information stored in our brains.
Aging, of course, takes a toll. "As people get older," says Baddeley, "even if they've learned them, the numbers and codes don't necessarily spring to mind that easily. Aging doesn't degrade very much the ability to remember a phone number from looking it up in a book to dialing it, but it has serious effects on remembering the number after a few minutes of non-rehearsal."
Engle explains why. "As we get older the vulnerability to interference gets worse in two ways. First, because we have accumulated so much more interfering information in the given context. But also, the ability of the executive attention system to get through the interference declines. In fact, it starts declining at about age 19. It's a steady downhill slope throughout one's life."
Why, then, didn't somebody think to ask psychological researchers about all this while our password-dependent society was being assembled? As it happens, a few did ask, but fewer heeded the answers.
One of the earliest to ask was the United Kingdom postal system. Baddeley says he "got into it because my PhD was paid for by the British post office, in connection with their interest in developing a postal code many years ago. I had worked as a vacation student at the applied psychology research unit at Cambridge. Then, after a year at Princeton, I came back to Cambridge looking for funding for a PhD"
His mentor was R. Conrad, who had done considerable work on memory and coding. Conrad, who didn't use his first name, only his initial, may be best remembered as the one who discovered the "acoustic similarity effect," whereby a "v" will be misremembered as a "b" or "c" - letters with which it rhymes - but not as an "r," even though it looks more like "r." That's because we sound letters to ourselves to help us remember them. A string of rhyming letters, for example "bcde," is harder to remember than non-rhyming letters such as "kroy."
"You could say I've built my career on this discovery," Baddeley now says. It was Conrad who came up with a postal code research proposal, and Baddeley's work that led the UK to adopt its postal coding system of numbers sandwiched between clusters of letters.
"If the letters make up something meaningful, for instance the first letters of a town name, it makes it easier to remember," Baddeley explains. "NOR, for Norwich, was the first one they tried. Then they wanted a mix of letters and numbers after that. Initially they wanted only one number. They wanted to know where the number should be. It turned out that putting the number in the middle, between the city name and random letters - for example, NOR 1AB - was the easiest to remember."
For his PhD, however, Baddeley went considerably beyond that, inventing a codebook for the entire UK based on 10-letter pseudo words, strings of 10 letters that were not real words but could be pronounced (and remembered) like words. His plan was not adopted. Today a typical British address has a routing code that is a seven-item string: two letters, three numbers (in the middle), two more letters.
DOES THIS RING A BELL?
In the U.S., meanwhile, the telephone system was going through similar growing pains, with similar mixed results from the input of psychological science. Bell Laboratories formed its User Preference Research Department in the 1940s, headed from 1951 to 1977 by former Harvard psychologist John E. Karlin. Although the department did a few studies in the 1950s comparing consumer response to the seven-digit phone numbers with the two-letters-five-digits system that preceded it (remember "PEnnsylvania 6-5000"?), most of the human factors research effort went into developing and designing new equipment for the consumer market, not helping customers remember numbers.
Roger N. Shepard, now professor emeritus at Stanford University, was an exception. The APS Fellow and Charter Member joined the Bell Labs technical staff in 1958 with the idea of trying to use its powerful computers to simulate the evolution of perceptual-cognitive mechanisms. He discovered early on that this wasn't exactly what his new employers had in mind.
At lunch one day he mentioned his interest in the then emerging field of artificial intelligence to the executive director of the division that included his department, an engineer named John R. Pierce. As he recalls it now, the eminent engineer leaned back in his chair and exclaimed: "Ah yes, AI! There's no holding that area up; it keeps hitting new lows!"
(Perhaps this is the time to note that in 1995 Shepard was awarded the National Medal of Science for his pioneering work in cognitive and evolutionary psychology.)
In an attempt to give AT&T a practical solution to a practical concern in the early 1960s, Shepard and an assistant conducted an experiment in which they asked subjects to look up and dial seven-digit phone numbers arranged in one of two sequences - the normal format, in which a familiar three-digit exchange is followed by four random digits, and in reverse order, with the familiar exchange number listed last. They found that putting the familiar numbers last instead of first cut in half both the time it took to look up and dial the number and the number of dialing errors.
"In addition to making life easier for telephone users," Shepard writes in a chapter soon to be published,1 "such a change could have saved the Bell System vast amounts of money. Moreover, such a change was just then becoming implementable thanks to … computer-controlled switching. But when I circulated a draft of our report to the applied and engineering divisions, the feedback I uniformly received was to go back, as one respondent put it, to my 'ivory tower.' The system, they said, was too big to make such a change - especially one that would in any way affect the subscribers."
Next came direct dialing, and after that an explosion of toll-free "800" numbers. As the system eventually used up the possible "800" numbers, Bell Labs wanted to know what other prefix could be assigned to toll-free calls.
Thomas K. Landauer of the University of Colorado, also an APS Fellow and Charter Member, conducted a literature search and found "a fair amount of research on what numbers and what patterns of numbers are easiest to remember. It showed that repeated numbers are easier to remember, and that for some unknown reason, things with the number 8 were a little better." That, coupled with some of his own work, led him to recommend in a white paper submitted to the North American Numbering Plan organization, the entity that decides such things, that 888 be the new toll-free "area code."
It was adopted, but "I don't know what sort of discussion and negotiation if any went into 888," Landauer says now, "or even if it really had anything to do with my recommendation."
Landauer also weighed in "mildly" in favor of overlaying new area codes over old ones in geographical areas whenever the old codes "filled up," instead of subdividing the area, "on the grounds that most people would keep their familiar numbers and the rest would have a gradual transition, mostly composed of new people and businesses" moving in.
"There wasn't good experimental psychology on this issue," he concedes, "except the clear knowledge that changing from one number to another would be hard. Every expert psychology or learning text tells that story."
O NO, CANADA
The lack of good research appears to have been the story through much of the decision-making history of our number-coded society.
Baddeley, who helped design the British postal code, comments ruefully, "Over the years, I've watched with interest as the codes got longer and longer. There's been virtually no research since [the 1950s, when he did his study]. Nobody thinks much about it. They just went on ignoring what we had learned since then. Occasionally, consultants are asked about it, but mostly people just stick more digits on as they need additional capacity."
Canada's postal code, by all accounts, is Exhibit No. 1 in how not to design a code intended for memorizing. Alternating letters and numbers, "it is reliably worse than even randomly assigning letters or numbers," says Baddeley.
Sternberg agrees. "It is one of those cases in which well-known and big effects related to human memory have been ignored in the design of codes that people have to remember. The postal codes in Canada may reduce the most common error in pure digit sequences-the transposition of adjacent symbols-but they were introduced well after psychologists had learned how difficult it is for humans to process strings of symbols from alternating categories."
Roediger concurs that the Canadian system "seems to have been deliberately designed to be hard to memorize. For example, M5S 1A1 is the postal code for the University of Toronto, or used to be when I taught there. It would be relatively easy to design more memorable codes by putting letters and numbers together. Missouri license plate numbers, for example, are numbers then letters - 868 AST is much easier to remember than 8A6 S8T."
The design of codes and password systems could do with a dose of scientific inquiry, most researchers agree. "I think it would be worth funding someone to do a thorough go-through of the literature and maybe do some further research," says Baddeley.
"Direct experimental lab research would help," says Landauer, but he laments that funding sources for such work are not obvious. "Few industrial labs are doing fundamental research or even applied research of this depth anymore."
Much of the relevant psychological research in the United States is done at universities, says Sternberg. "For many years, there has been a barrier between some parts of psychology and applied psychology. The barrier, I believe, results partly from a snobbish attitude toward applied research by some academic experimental psychologists.
Among other things, says Sternberg, "Academic psychologists are sometimes quite unaware of the job opportunities for their students outside academia."
Sternberg says that when he worked in the research area of Bell Labs he learned that the number of experimental psychology PhDs employed in the development area had grown to about 200, an "astonishing" number.
"I asked one of the directors who hired many such people how this had come about. He told me that when systems or hardware were being designed, engineers sometimes made human factors decisions that seemed, from the viewpoint of common sense, to be errors, such as locating a part that must be replaced periodically in a piece of equipment so that it is impossible not to burn your hand on another part when replacing it. He said that when people trained in psychology were involved, these errors did not occur."
Involve psychologists, reduce errors: Now that's something to remember.
RICHARD HÉBERT is a freelance writer living in Chicoteague, Virginia.
Copyright © 2001 American Psychological Society. All Rights Reserved.