One

The Zenith Broadcast

Commander” Eugene Francis McDonald Jr. favored checked suits and a cocktail made of gin and pistachio ice cream. He resided on his 185-foot yacht, the Mizpah, docked in Chicago’s Lincoln Park Yacht Harbor. As the chief executive of the Zenith Radio Company, McDonald lived as swashbuckling a life as any titan of industry could hope for. His interests ranged from Arctic exploration to searching for pirate gold.

McDonald’s primary contribution to American business was the publicity stunt. In 1934 he sent a telegram to every US tire and oil company: WATCH ABSENCE OF PEOPLE ON STREET BETWEEN ELEVEN AND ELEVEN THIRTY DURING PRESIDENTIAL TALK. The streets were indeed deserted during Franklin Delano Roosevelt’s fireside chat. A follow-up mailing touted the power of radio. The B.F. Goodrich Company agreed to sell Zenith’s radios through its chain of 1,200 tire dealers. Many radio shops had gone bust after the stock market crash, making this a lifeline for Zenith.

McDonald supplied Zenith radios to Hollywood movie sets, inventing product placement. From 1929 into the television era, Zenith radios appeared in films ranging from Busby Berkeley musicals to Night of the Living Dead. They’re in war pictures, screwball comedies, film noir, and Three Stooges shorts. In one, Curly gets hit over the head with a Zenith radio—which must have been how regular moviegoers felt. McDonald didn’t pay for the plugs. Zenith sent along two radios to each production, one for the property manager to take home as swag and the other to appear on-screen, preferably in a close-up.

McDonald’s biggest publicity stunt of all involved network radio at the peak of its influence, in 1937. In May of that year, a few words from NBC announcer Herb Morrison were sufficient to destroy an industry. “It’s burst into flames,” gasped Morrison as he watched the Hindenburg disaster unfold. “Oh, the humanity!” Thereafter no one wanted to fly a dirigible. In 1937 Arturo Toscanini picked up the baton of the NBC Radio Orchestra, and the young Orson Welles took over as the voice of The Shadow. But nothing on the 1937 radio dial was as peculiar as the show that “Commander” McDonald cooked up.

Across the nation, Zenith dealers began handing out complimentary decks of cards. A free pack of cards was hard to pass up in those Depression years, but these were not cards that anyone could play a normal game with. The backs shimmered with a hypnotic design containing the Zenith logo and the words DEVELOPED IN PARAPSYCHOLOGY LABORATORY AT DUKE UNIVERSITY.

The cards were promoting a new Sunday-night radio series. It was McDonald’s plan to capitalize on the nation’s craze for ESP (extrasensory perception). During the mid-1930s, Joseph Banks Rhine commanded the nation’s attention with his psychic experiments at Duke University. He claimed success in demonstrating telepathy, clairvoyance, and telekinesis. With piercing eyes and a dramatic sweep of steely hair, Rhine—a botanist by training—was a compelling advocate. He received mostly favorable attention in publications ranging from the New Yorker to Scientific American. As one journalist condescended, Rhine made ESP “the brief rage of women’s clubs all over the U.S.”

On a balmy June night, Rhine and his wife had dinner aboard McDonald’s yacht. McDonald sketched his idea for a nationwide test of ESP by radio. Listeners could test their own psychic powers. It would be the biggest experiment ever, providing the best possible proof that telepathy was real.

Rhine was not sure that his newborn science was ready for prime time. Skeptics suspected that Rhine was reporting successes and ignoring failures—that some of his telepaths were cheating.

The skeptics didn’t worry McDonald. As one of his associates put it, “nothing stops a crowd on a street like a fight.” McDonald played Mephistopheles, tempting Rhine with plans to monetize telepathy. He said he’d have his attorneys look into copyright and trademark protection for the cards that Rhine used to test ESP. This was the so-called Zener deck, named for a colleague, marked with five symbols (circle, cross, wavy lines, square, and star). Rhine would get a royalty on every pack sold, McDonald promised, and they’d put them in five-and-dime stores.

Rhine was ambivalent about the show. He agreed to let his name be used as a “consultant,” on the understanding that other psychologists would supervise the experiments. McDonald agreed.

The half-hour series debuted as The Zenith Foundation on NBC’s Blue Network on September 5, 1937, at 10 p.m. Eastern time. By design, the show’s name didn’t give the slightest clue to its subject matter. It was teased as “a program so DIFFERENT—so STARTLING—so INTERESTING—that it will become a regular habit with people all over the country.” The word Foundation evoked grand philanthropy on the scale of Rockefeller’s, but McDonald saw no reason why public service couldn’t coexist with profit. A flyer sent to dealers spelled it out: “The broadcasts of The Zenith Foundation have been planned to help you sell more Zenith Radios.… Make the best of this opportunity. Get behind it and push.”

McDonald was concerned that the word telepathy might deter the more hardheaded listeners, so the first broadcasts said little or nothing about it. Early episodes took up the theme of great thinkers whose ideas had been unjustly ridiculed. Over a period of weeks, the program eased into a template that remains familiar in today’s cable TV universe—dramatizations of allegedly real psychic phenomena with commentary by a motley group of “experts.”

The novel element, McDonald’s telepathy experiment, was introduced on the fourth broadcast. A panel of ten “senders” in a locked Chicago studio attempted to broadcast their thoughts to the nationwide audience. Listeners were encouraged to write down their psychic impressions and mail them in.

In the first test on September 26, the senders transmitted a random sequence of the colors black and white. To forestall any trickery, the choices were decided during the broadcast by the spin of a roulette wheel.

Narrator: It is best to write down your impression as soon as you receive it. Do not think about it or try to reason it out. Write down your impressions in consecutive order—as rapidly as you get them. The machine is now ready to select number one.

SPIN… STOP… BELL… INTERVAL… BELL

Narrator: That was number one. The machine will now select number two.…

Almost as soon as the audience responses started pouring in, it was apparent that something remarkable was going on. There were five black-or-white choices to be guessed. The majority of the radio audience was correct on all but one. Rhine must have felt pleased, and relieved, at this favorable result.

After that first test, Woolworth’s department store sold out of ESP cards and had to reorder. The card symbols were used in several of the later tests. It’s said that 150,000 packs were printed during the show’s run. They still turn up on eBay.

The next week the choices were drawn from five vegetables: carrots, beans, peas, corn, and beets. This made it harder, as there were five possibilities for each of five places in the sequence. Two times out of five, the choice that got the most listener guesses was correct. That was twice what might be expected from chance.

On the following two broadcasts, the testers again used black and white as the choices. On October 10, the majority’s guess was correct four out of five times; on October 17, five out of seven times.

For the October 24 broadcast, the choices were circle and cross. The transmitted signals were OXXOX, and the majority’s guess was right on every single one.

That’s not to say that every individual listener guessed so well. But somehow the majority choices were amazingly accurate—a telepathy of crowds? In many ways the aggregate results were more impressive than any individual’s could have been. Given that parapsychology was a game of statistical significance, the Zenith experiment was like a more powerful microscope or supercollider, able to discern smaller effects with precision. During its fifteen-week run, the series collected over a million individual guesses, making it the most ambitious test of ESP ever conducted. On many of the broadcasts, the statistical significance of the audience’s correct guesses was fantastically high. The Zenith Foundation later put out a report claiming that the odds against the results being just a coincidence were 10,000,000,000,000,000,000 to one. The radio audience didn’t need that suspiciously round number to feel that it had participated in something uncanny.

Zenith retained several distinguished psychologists to design and carry out the experiment. Behind the scenes, they were fighting tooth and nail.

As Rhine preferred to keep the show at arm’s length—easy to do from his Durham, North Carolina, lab—the hands-on role went to two Northwestern University psychologists, Robert Harvey Gault and Louis D. Goodfellow. Gault, a few years from retirement, had a longstanding interest in telepathy experiments. Goodfellow was a young psychologist in Gault’s department. He wore donnish spectacles and parted his hair down the middle. Both shared McDonald’s conviction that the radio show was a unique opportunity to test whether telepathy was real.

It was easy to replicate Rhine’s experiments, requiring only a deck of cards and a grad student with an hour to kill. Those psychologists who did so were generally disappointed. The inability of colleagues to confirm a finding is supposed to be fatal in science. In reality, it’s never that simple. Rhine’s thesis was, or came to be, that telepathy is a delicate thing. It is not 100 percent accurate, nor can it necessarily be summoned by anyone at any time. A failure to repeat Rhine’s results might simply mean that the subject(s) lacked the “gift.”

Goodfellow and Rhine bickered at long distance over details large and small. Gault was exasperated with both of them. After the first few broadcasts, Goodfellow realized something that infuriated Rhine. Goodfellow could predict the radio audience’s guesses!

It wasn’t telepathy. In a way, it was better than telepathy. Goodfellow had a simple way to predict what the American public was going to think before they knew it themselves.

Interesting as this was, it was not what Rhine or McDonald wanted to hear. Goodfellow’s opinions were a threat to their increasingly profitable ESP industry (oh, the humanity!). Goodfellow was branded an enemy of the paranormal and dismissed from the show. Meanwhile, the ESP program’s novelty wore thin, and the ratings trailed off. In early 1938 McDonald canceled the show.

Goodfellow independently published an analysis of the Zenith results in the Journal of Experimental Psychology. He offered a convincing explanation that did not involve ESP. Time magazine wrote that Goodfellow “pricked Telepath McDonald’s iridescent bubble.” For good measure, Goodfellow debunked some of the tales discussed on the program. In one, a psychic detective was said to have led police to the body of a murdered woman, buried in a woodshed. Goodfellow found court records showing that the body had been located on a tip from a boy who peeked through a knothole.

After that, the parapsychologists’ feud got really childish. Goodfellow, who did not entirely live up to his name, penned an attack on Rhine under a fake name.

Cadaco-Ellis, a Chicago-based publisher of popular board games, introduced a new game called Telepathy. It was created by a certain “Dr. Ogden Reed,” and the instructions slighted Rhine’s science as “full of loopholes.” One criticism was that the ink used to print Rhine’s ESP cards caused the paper to shrink.

Those royalty-bearing cards had brought Rhine no end of grief. In the interests of cost cutting, the cards had been printed on such thin paper that amateur psychics could see through them. Psychologist B. F. Skinner “guessed” twenty-three out of twenty-five cards, to the hilarity of his students. This made Rhine a laughingstock even though he had nothing to do with the cheap cards and they weren’t the ones used in his lab.

It did not take telepathy for Rhine to guess that “Dr. Ogden Reed” was Dr. Louis Goodfellow. “Is it proper,” Rhine wrote to Goodfellow, “for an academic man to use a surreptitious approach (in this case, an assumed name) to avoid having to meet the responsibility for the things he is expressing?”

McDonald was furious. He told Rhine that he should sue the toy company over the game and promised to foot the legal bill himself.

“Rhine and Goodfellow keep me supplied with carbon copies of their love letters,” wrote Gault, the show’s senior psychologist, to McDonald. “I’m not surprised that R. is up on his ear. Between you and me and the gatepost, I don’t care what kind of spanking he administers to G. The latter is an excellent technical man in the laboratory and in that capacity he is useful to me. But in some other respects he is a damn fool.”

As that indicates, Goodfellow did not have an inside track on tenure at Northwestern. He left during the war to become director of the air force’s Technical Training Command. Afterward he joined the psychology faculty of his hometown school, Penn State Altoona. He spent the remainder of his career in that comfortable backwater, teaching and doing good works in the community, though achieving little of note to the outside world. Today Goodfellow is remembered almost entirely for the Zenith experiment. He is a hero to the skeptic movement, almost on a par with Harry Houdini or James Randi. But Goodfellow did more than debunk. In demonstrating that the radio show’s mindreading was fake, he discovered an authentic form of mind reading.

Goodfellow was not even trying to do what the show’s listeners were trying to do—predict the senders’ transmitted thoughts. Those thoughts, determined moments before by roulette wheel, were truly random, as Goodfellow himself had made sure. Instead, Goodfellow was predicting the audience’s guesses about the random sequences.

On the first broadcast, the psychologists had played a trick on the listeners. The audience had been led to believe that seven choices were transmitted. Actually, there were only five. For the third and seventh transmissions, the panelists were instructed merely to count rapidly to themselves and not to think of black or white, the two allowed choices.

That was good science. And no one in the radio audience realized the deception. Had there been any authentic telepaths out there, they could have written in to say, “Hey! I didn’t get any color for #3 and #7, just someone counting.” No one did.

This foreshadowed Goodfellow’s key finding. The transmitted sequences were random; the audience’s guesses were not. In aggregate the guesses were similar for every broadcast. They followed some simple patterns.

For instance, when the choices were heads and tails, most people chose heads as their first guess. This was not a trivial effect. Nearly four-fifths picked heads. Goodfellow was able to confirm this by doing his own survey experiment with Northwestern University students, supermarket shoppers, and town businesspeople. Each volunteer was asked to invent a five-item sequence of heads or tails. (Telepathy was never mentioned.) Seventy-eight percent picked heads as their first choice of the sequence.

Goodfellow found that 66 percent chose “light” rather than “dark” as the first choice when those were the options; “white” was favored over “black” by a slim 52 percent majority. This meant that someone schooled in these preferences could guess someone else’s first “random” choice with greater-than-chance accuracy.

Goodfellow discovered that 35 percent picked the circle as their first choice when devising a sequence of the five Zener card symbols. Someone who knew this and predicted “circle” would be correct much more often than the expected 20 percent. Six of the Zenith experiments used these Zener symbols.

Goodfellow also found that some sequences of responses were greatly preferred over others. For most broadcasts there were five two-way choices. I’ll use H and T as shorthand, with H representing the first-picked choice, whatever it was. The least popularly guessed pattern was HHHHH. No mystery there! The audience had been told to expect a random sequence. Five of the same thing is the least random-looking possibility.

This brings up the distinction between “random” and “random-looking.” HHHHH (and TTTTT) is just as likely to show up in five fair coin tosses as any other sequence. You can’t say it’s any less random—though it sure looks less random. Perceptions of randomness are based on how mixed-up a sequence is. The Zenith show’s most frequent guesses fit the pattern HHTHT. This is H and T back and forth, with one extra H thrown in to mix it up. That syncopated rhythm was characteristic of all the popular responses.

The public liked the split between heads and tails to be as even as possible. With five choices, the closest you can come to fifty-fifty is to have three of one and two of the other. All the most popular answer patterns were three/two splits.

Well-shuffled patterns (like HHTTH or HTTHT) were preferred to oil-and-water patterns like HHHTT or HHTTT. But shuffling could be taken too far. The least popular three/two split was the perfectly alternating sequence HTHTH.

The radio audience guessed HHTHT almost thirty times as often as TTTTT. This was true throughout the show’s run, regardless of the sequence being transmitted. Though individuals might have sent in different answers for each broadcast, the overall popularity of patterns remained fairly consistent. The guessers were picking the same sequences, again and again, without realizing it.

This analysis accounted for the results without any need to assume telepathy. Whenever the correct sequence happened to start with a favored symbol and thereafter follow a favored pattern, there were lots of direct hits. When the patterns didn’t look so random, America’s amateur telepaths went off their game.

On November 21, using circles and crosses, the correct sequence was OOOOOX. The majority of the audience guesses were wrong on four of the six choices. It seemed evidence of “negative ESP.”

On December 12, the choices were heads and tails, and the correct answer was TTHHH. Because a lopsided majority picked heads for the first pick, there were few perfect scores.

Goodfellow showed that ten of the fifteen transmitted sequences happened to be popular ones, and five were unpopular ones. That accounted for the high rate of successful guesses. It could have just as easily gone the opposite way, with more unpopular responses.

What no one, not even entrepreneur McDonald, appreciated was that there might be value in what Goodfellow had discovered: a way to predict what the public will think and do.

The random, the arbitrary, and the made-up are all around us and sometimes take on great importance. We are all currently engaged in a Zenith experiment, and the stakes are our privacy, our wealth, and our very identities. I’m talking about the passwords that lock and unlock our digital lives. The computer user believes that she has utter freedom to choose a password. For practical purposes, she doesn’t. She is limited by the way her mind works, and by the fact that her mind is not so different from anyone else’s.

It’s not just that many use those common passwords we’ve all been scolded not to use. The deeper issue is that even prudent users favor the same patterns of obfuscation (such as adding “123” onto the end, alternating capitals and lowercase letters, and other schemes only a bit more clever). This cuts the exponentially vast range of potential options down to manageable size. Password-cracking software does what Goodfellow did, only billions of times faster.

For a time, AT&T’s vision of a wireless future did not rule out telepathy. Thornton Fry, head of Bell Labs’ mathematics division and the man who hired Claude Shannon, was in that diehard minority of scientists who believed that J. B. Rhine might be onto something. About 1948, Bell Labs built an ESP machine. It was a device for generating random sequences that a would-be psychic would attempt to guess. By taking the place of Zener cards, the machine could exclude the possibility of cheating or unconscious signaling that had dogged Rhine’s research. Rhine himself saw the machine on a visit to Bell Labs and fell in love with it. He immediately wrote to the president of Duke University, hoping to get Bell Labs to build him one, too. It never happened. Today, random generators are free on the Web, but back then this was expensive hardware.

In 1953, the year of Shannon’s mind-reading machine, Bell Labs focused on a more modest goal: designing a push-button keypad for the telephones of the future. It assigned the task to the famous industrial designer Alphonse Chapanis, who happens to be of some importance to our story.

Chapanis is known as a founder of ergometrics, the discipline of human-based engineering. In a famous tale, Chrysler chief executive Lynn Townsend once pulled Chapanis aside to ask what he thought of a sporty new model. The steering column had a decorative boss—a spike—extending an inch or two beyond the steering wheel.

“Mr. Townsend, do you know what you’ve designed here? You’ve designed a spear aimed at the driver’s heart.”

Townsend’s reply was “Doc, it’ll sell.”

In at least modest ways, Chapanis invented the world we live in. He figured out which controls should work which burners on a kitchen range. His durable keypad design is still used on smartphone touch screens. Chapanis did not believe in telling clients or consumers what they should want. His concern was to find out what worked. He tested every reasonable design for range controls or telephone keypads to see which produced the fewest errors. His approach was experimental, using the methods and statistical tools of a psychology lab. An important part of that was randomized trials. Testers would be assigned to designs at random, to eliminate various sources of confusion or error.

You can’t randomize without a random sequence, and Chapanis noticed that it was not so easy to invent such sequences. In 1952 he performed an innovative experiment. He asked twelve volunteers at Johns Hopkins University to write long sequences of random digits. They were given four sheets ruled into squares and told to write a number in each square.

Use all the numbers 0, 1, 2, 3, 4, 5, 6, 7, 8, and 9 in a random way. By random, I mean that each number should be written down about an equal number of times, but there should be no regularity or order in the way in which the numbers are written down. A random series of numbers is a completely scrambled one with no system to it.

The subjects each produced 2,520 digits, a demanding task that took over an hour. As Chapanis expected, the volunteers weren’t very good at counterfeiting randomness.

Some digits were chosen more often than others, despite the instructions. The least chosen digit, for just about everyone, was 0. Otherwise, digit preferences varied. One volunteer consistently overused 3s; another liked 8s.

When Chapanis looked at consecutive pairs and triplets of digits, patterns were pronounced and often consistent across subjects. The ten least popular pairs of digits were (in order of descending popularity):

66 99 00 11 33 44 88 22 77 55

All are same-digit pairs.

Here are the ten most popular digit pairs:

32 43 21 76 65 10 31 87 86 54

See the pattern? All but two are decreasing pairs in which the second digit is one less than the first.

There were similar patterns for triplets. Same-digit triplets (like 888) were rarely used. This meant that streaks of the same digit were fewer and shorter in the volunteers’ worksheets than in truly random sequences. Decreasing triplets (like 987) were popular. Increasing sequences like 34 or 234 were popular, too, though to a lesser degree. The volunteers may have felt that decreasing sequences looked more random than increasing ones. The pattern in 321 does not jump out at you so quickly as the one in 123.

These made-up digit sequences were unrandom enough to be predictable. Chapanis computed that just by knowing a person’s previous digit choice, he could guess the next digit 17 percent of the time. That’s much better than the 10 percent chance you’d expect for random guessing. By using the two last digits, he could guess correctly 28 percent of the time. That’s almost three times better than could be expected. Were you able to guess the numbers in roulette with this accuracy, you could make a quick fortune (well… you could get quickly banned from the casino).

Chapanis divided his volunteers into “sophisticated” and “relatively unsophisticated” groups. The sophisticated group, with strong math backgrounds, was somewhat better at counterfeiting randomness but subject to the same types of errors as the others.

The most striking finding was that strings as long as eight digits might be repeated exactly, thousands of digits later in the sequence. One volunteer wrote the sequence 21531 four times and 21924 three times. Another repeated 43876538 and four other eight-digit sequences. These coincidences can’t be explained by chance. They have the quality of amnesia or sleepwalking. The subjects fell into mental ruts and repeated themselves without realizing it, like a dotty grandparent who tells the same joke every Thanksgiving.

Chapanis’s study was a “randomness experiment.” That term is now applied to one in which volunteers are asked to make up random sequences. The point is usually to examine how humans fail to be random.

Of the many such experiments done in psychology labs, Chapanis’s had particular relevance. It involved decimal digits, the kind of numbers that make the world go round. When a fraud artist fabricates financial data, he must make up a series of numbers that look normal and unsuspicious and random. We now know that fraudulent numbers usually have fingerprints of just the sort that Chapanis described. In recent years the idiosyncrasies of invented numbers have become a valuable clue in authenticating expenses, sales figures, tax returns, election results, and other important data.

Chapanis did not envision this application, however, and for complicated reasons his randomness experiment did not get the attention it deserved. He described his experiment in a slide lecture at an academic conference; an eight-paragraph abstract of that talk appeared in a 1953 issue of the American Psychologist. Then Chapanis got very busy with ergometrics, and with playing James Bond.

He had a double life as an American spy, traveling to Soviet bloc nations for design conferences and reporting back to his handlers. In this he was aided by his Russian-speaking wife. Chapanis never got back to publishing his random-number work until his retirement. The full report on his 1952 experiment appeared in 1995, in Perceptual and Motor Skills. By then there was a substantial, still-growing interest in the randomness experiment and its applications.

A proper history of the randomness experiment would begin with Hans Reichenbach’s The Theory of Probability, a 1934 textbook published in English in 1949. Reichenbach, a distinguished philosopher of science, was apparently the first to articulate two key points. One was that “persons not acquainted with mathematics… are astonished at the clustering that occurs” in a true random sequence. In tosses of a fair coin, strings of consecutive heads are longer and more common than most expect. Reichenbach’s second point was that people “asked to construct artificially a series of events that seems… well-shuffled” create too many alternations. When inventing a sequence of coin tosses, people tend to switch back and forth between heads and tails, neglecting to include enough streaks of the same choice. This was amply demonstrated in the Zenith broadcasts and in Chapanis’s study.

By 1972 W. A. Wagenaar of the Netherlands’ Institute for Perception could review fifteen publications on the randomness experiment. Wagenaar complained, “There is no way of combining details of the results… into one coherent theory.” The researchers came to this peculiar subject with diverse agendas. Their volunteers were asked to simulate random sequences of coin tosses, die throws, digits, letters of the alphabet, or nonsense syllables. They wrote their answers down, or said them aloud, or pushed buttons. In some experiments subjects could consult a running list of their previous choices; in others they couldn’t. Though all the subjects were explicitly told to be “random,” the instructions did not always define that word. (A mathematical or philosophical discussion of the meaning of randomness could fill a book much bigger than this one.) The experimenters used different and incompatible methods to evaluate their results.

Despite this Tower of Babel, there were areas of consensus. Nearly all the papers supported both prongs of Reichenbach’s claim. The randomizing was poor with just two choices (heads and tails, say) and worse with many choices (decimal digits or letters of the alphabet). When attempting to write strings of random letters, subjects overused those letters that occur most often in words (the vowels and M, N, R, S, and T). Consecutive occurrences of the same letter (FFF) were avoided, but volunteers favored pairs of letters that are adjacent in the alphabet (AB or FE). This parallels Chapanis’s finding about digits in ascending or descending order.

To be random is to be unpredictable. Turn that around: Any human action that fails to be random is, in some measure, predictable. For years, mathematician Theodore P. Hill performed a classroom randomness experiment. He assigned the homework of flipping a coin 200 times and recording the results. About half the class (those whose mothers’ maiden names began with the letters M through Z) was told to skip the coin flipping and merely fake the data. Either way, the data had to be submitted at the next meeting.

Hill astonished his students by glancing at the reports and suavely sorting them into two stacks. His accuracy in distinguishing real from faked coin flips was close to perfect.

Hill’s main tip-off was that a run of six consecutive heads or tails in 200 tosses is almost certain to occur with genuine random data. Few fakers dare to invent a sequence of that length, though. Hill could spot streaks of six at a glance, so he didn’t need to count. The faked data just looked different.

There are now Web apps that perform the same stunt. You enter real coin toss results in one box and a made-up sequence in another. The app “reads your mind” and tells which is which. Because it can run several mathematical tests at once, it requires only about fifteen tosses in each box to achieve high accuracy. It’s surprisingly difficult to fool the computer—even for those who know the giveaways and try to correct for them.

Outguessing is founded on the semipredictability of human choices. The semi- is the hardest part of that idea to accept. We have been led to think that predictions must be certain in order to be useful or even legitimate. The truth is that we make predictions of human thoughts and actions every day, and even a small statistical advantage can be valuable.

Mentalists were among the first to make a living from not-so-certain predictions of arbitrary choices. A mentalist is an entertainer who pretends to read minds. Much of mentalism is fake. Like regular magic, it uses sleight of hand and outright deception. There’s a part of mentalism that isn’t fake, though, and it’s the more interesting part. When a mentalist tells you to think of a number or a color, you reflexively try to make it hard for him to guess your choice. That is, you try to pick randomly. In fact, you make a choice that fits your mental stereotype of randomness. This is more predictable than you think. Mentalists play the odds and usually have ways of sneaking in a few more guesses, should the first one not hit. They accept a chance of failure. J. B. Rhine unwittingly helped out there by promoting the credo that telepathy is never 100 percent. Audiences, who are often uncertain whether a mentalist’s powers are “real,” take the occasional miss as a badge of authenticity.

I’ll describe a mentalism effect I witnessed that illustrates some of the tactics that we’ll be using throughout this book. It goes by the name of Terasabos. The performer calls an audience volunteer onstage and positions him at one end of a table with five upside-down teacups in a neat row. He asks for a personal object, like a watch, and the volunteer supplies one. The mentalist says he is going to turn his back and let the volunteer hide the watch under any one of the cups, ranging from one to five. He demonstrates the action—lift up a cup, put the object under it, and replace the cup.

Then he turns his back. There doesn’t appear to be any possibility of his peeking, not with everyone in the audience watching him. The volunteer chooses a cup (say, #4) and hides the watch under it.

The performer turns around and instructs the volunteer to concentrate on the location of the watch. He says he has to eliminate the four cups where the watch isn’t, and to zero in on the one where it is. For a few moments he stares intently at the cups.

Then ultimately the performer lifts cup #4, revealing the hidden watch.

The name Terasabos and its brilliantly optimized form were invented by mentalist Rick Maue, who drew on a number of traditions of mental magic. The most amazing thing about this effect is how real it is. The performer is guessing where the hidden object is, using the psychology of choice. He’s usually right. Properly done, the accuracy is said to be about 90 percent. Understand that, and this simple trick takes on cosmic dimensions. It is about free will, here revealed as the greatest illusion of all.

Start with the basics. The performer stands a 1 in 5 chance of being right. Right?

That’s what you’re supposed to think. Having read this far, you should know better. Whenever there is a set of options, you should expect that some will be more frequently chosen, even when none offers a clear advantage. This is true of ESP card symbols, lottery numbers, passwords, and just about everything else.

When presented with five objects in a neat row and asked to pick one “at random,” most people avoid the choices at either end (#1 or #5). Those prominent positions, at the beginning or end of the line, seem less typical and less random. That is despite the fact that a truly random chooser would pick one or the other end position 40 percent of the time.

Eliminating those two leaves three of the five options as credible choices. But the one in the center is also atypical because it’s in the center. That makes it a choice to avoid, too. Eliminate it, and we’re left with options #2 and #4.

There are other mentalism stunts that make use of this preference for positions 2 and 4 out of five. Terasabos enhances the basic 2-4 effect with further psychology. Anything that singles out one choice to the volunteer’s attention makes it less “random” and, thus, less likely to be chosen.

You’ll recall that the volunteer is made to stand next to cup #1, and the performer stands next to #5. The mentalist casually mentions that the cups “range from one to five,” touching those two named cups in the process. He takes the offered object itself and places it under cup #1, supposedly to demonstrate how to do it. All of this foregrounding of #1 and #5 decreases the chance that they will be chosen. Touching is especially potent. Touched options are like baby birds returned to the nest, doomed to be rejected. In fact, cup #1, which had the object under it in the demonstration, is so radioactive that it motivates the chooser to pick a cup as far from #1 as feasible. As there are only two or three likely choices, this boosts the chance that cup #4 will be chosen. In practice, about 40 percent do choose cup #4. Less than 10 percent pick #1.

No performer would want a trick that works only 40 percent of the time. That’s where the safety nets come in. Before picking up cup #4, the performer says, “Concentrate on the cup you’ve chosen… I am going to eliminate the empty ones… and zero in on the right one.” He waves his hand over the cups and then, with great deliberation, lifts #4.

About 40 percent of the time, the object is there. The audience goes home talking about a feat that utterly defies logical explanation. (So it appears!) Otherwise, #4 is empty and the performer announces without missing a beat: “I have eliminated the first empty cup!”

He goes on to lift up the other cups in order of least likely to most likely to hold the object. The next cup to be revealed is the very unlikely #1. Before lifting it he announces, “I believe this cup is empty, too.”

There is a small risk of disaster here. Despite all the precautions, the object may be under #1. In that case the performer mentions, for the benefit of any who haven’t heard, that true mentalism is never certain. He segues to a more foolproof trick.

Usually the object isn’t under #1, and the performer continues by announcing that #5 is empty, too. He lifts up #5. Provided it’s empty, he’s home free.

It’s now down to two cups, #2 and #3, with #2 being more likely to hold the object. “This is the moment of truth,” the performer says. “It is time to reveal the location of the object.”

He reaches for cup #3. As he begins to lift it he says, “Here is…” If cup #3 is empty, the sentence ends “… the last empty cup.” The performer immediately lifts up cup #2 to reveal the object.

Should the object be under #3, the sentence becomes “Here is your object.” In that case, the performer doesn’t lift up #2. The audience is left to conclude that, when it got down to the last two cups, it would have been anticlimactic to show the empty one, so naturally the performer cut to the reveal.

The Terasabos effect is an interactive drama in which words, actions, and not-so-random choices form a garden of forking paths. The goal is to convince the audience that whatever plotline they witness was precisely the one preordained. Rick Maue explains the effect’s name (but never in his act), and it could apply to some of the stratagems in the chapters that follow. Terasabos stands for “This Effect Requires Acting Skills And Balls Of Steel.”