When Computers Were Human

by David Alan Grier

  The armistice released a tremendous mathematical energy on both sides of the Atlantic. Less than twenty-four hours after the firing stopped, the American army began to terminate ballistics experiments and demobilize the computers at Aberdeen. “On November 12, the telegraph wires fairly hummed with cancellation orders emanating from Washington,” wrote historian David Kennedy. “Within a month, the [ordnance] department had unburdened itself of $2.5 billion of [weapons contracts].”31 The Aberdeen computers felt the impact of these cancellations in less than three weeks. The number of test firings peaked on November 26, the busiest day of the war, and quickly began to decline.32

  The proving ground released the first computers in early December as it began to conclude ballistics experiments and range table production. When the computers looked for jobs as civilians, they looked for positions as mathematicians rather than in fields related to ballistics or computation. “For many years after the First World War,” recalled mathematician Norbert Wiener (1894–1964), “the overwhelming majority of significant American mathematicians was to be found among those who had gone through the discipline of the Proving Ground.”33 Wiener’s observation probably exaggerated the influence of the forty-two mathematicians who served with Veblen and Moulton, but it suggests the reputation that this group had acquired in the weeks and months after the end of the fighting. Most of the proving ground computers, including Wiener, were able to use their war experience to advance their careers. Shortly after leaving Aberdeen, Wiener was offered a position at the Massachusetts Institute of Technology.34

  The female computers, the women of the Experimental Ballistics Office in Washington, had fewer opportunities than their male counterparts. None of them held advanced degrees in advanced mathematics, and hence they were not qualified for the few positions that were open to women at the nation’s universities and colleges. Some hoped to find positions as computers, but they soon found that computing jobs declined in times of peace, and again, these jobs went to men first. Elizabeth Webb Wilson, perhaps the most ambitious of the group, tried to find a computing job in Washington, D.C. One of the ballistics officers described her as looking for “employment in which her somewhat exceptional preparation can be made useful in the national service.”35 She was no less aggressive in attempting to use her war record than the men, but she could not insist upon a job that made full use of her mathematical talents, as she had in March 1918. After a year of unemployment, she became a high school mathematics teacher in Washington.36

  22. Final picture of army ballistics computers in Washington, D.C.

  The task of closing the army computing offices fell to Oswald Veblen. At the exact moment of the armistice, he was with the American Expeditionary Force in France.37 Anticipating the end of the war, the army had sent him to inspect the ballistics facilities of Britain, Italy, and France before they were disbanded. He packed his bags with the latest calculations from the proving ground to give to the artillery command in France. He also took a new tuxedo, in case he was invited to any formal parties when abroad.38 During his trip, he took every opportunity to meet with European mathematicians. He visited Cambridge in England, the École Polytechnique in France, and the University of Rome in Italy.39 He returned to the United States in March and relieved Forest Ray Moulton at the Washington Ballistics Office. For the next six weeks, he prepared reports, summarized experiments, secured office records, and demobilized the few remaining computers.40 With the experimental program coming to a close, he had time to attend the opera, take long walks through the city, and make plans for the future of American mathematics. “Range tables are not being worked on to any extent nowadays,” was the final word from Aberdeen.41

  The armistice allowed Karl Pearson to reclaim the leadership of organized computation. He had withdrawn from ballistics computations six months before the end of the war and a full year before Oswald Veblen resigned his commission. He spent the spring and summer retrenching, a metaphor borrowed from the front lines in France. He hired new computers, evaluated the state of his laboratory, and started on a new plan of research. In many ways, the war clung to him longer than it touched the lives of the computers at Aberdeen or at Washington. On a visit to his country house, he wrote a long, elegiac memoir of his time during the conflict. He confessed to having become sensitive to the sound of thunder and associating the smell of pumpernickel bread with the odor of explosives. “I want instinctively to whinny like the dogs, if there be a sudden clap of thunder, and will-power has still to be exercised to avoid it.”42

  Armistice Day found Pearson sitting in a hospital recovery room next to the bed of Leslie John Comrie (1893–1950). Between the two of them was a Brunsviga calculator. Pearson was explaining the finer points of machine calculation, while Comrie was asking how certain problems might be handled by the device. L. J. Comrie, as he preferred to be called, had been a late recruit for the war. He was part of a New Zealand regiment that had been assembled to replace troops from the home island. He had studied chemistry at the University of Aukland before joining the army, but he had a deep love of astronomy and a special affection for the classical problems of positional astronomy. As his troop ship steamed across the Indian Ocean, he had occupied himself by tracking the ship’s course. In ordinary circumstances, it would have been a harmless diversion, but in time of war, when troop movements were secret, it defied military discipline and could have earned him a court-martial. He arrived in France, either undiscovered or forgiven, only to meet with one of the many meaningless events of the war. A munitions accident badly wounded him and forced the army surgeons to amputate one of his legs. While he convalesced in London, volunteer nurses visited him and asked if he would like to be trained for some trade or occupation that might be suited for the handicapped. Comrie replied that he would much prefer to continue his university education and become an astronomer. This conversation made its way to Pearson, who was always looking for potential computers. Brunsviga in hand, he found his way to Comrie’s hospital ward, where the two began a friendship over computation.43

  Pearson and Comrie had little in common beyond their mutual ambitions and their love of numbers. Pearson was an imperious man, a scientist who could speak from the mountaintop of his grand visions and his mathematical methods of proof. His biographer wrote of Pearson’s “fierce intellectuality and disposition to theorize about everything from religious faith to sexual love.”44 Comrie was a scrapper, always impatient to show that he was no one’s inferior. Once his health had recovered sufficiently, he started working in the Galton Laboratory, now the formal name for the office that Pearson had started as the Biometrics Laboratory. His heart was not in the study of mathematical statistics, and he certainly did not share Pearson’s infatuation with the Brunsviga calculator, but the laboratory gave a focus to his life while he prepared for the future. In all, he spent nine months with Pearson before a scholarship for New Zealand veterans allowed him to depart for Cambridge University and the study of astronomy.

  23. L. J. Comrie at calculator

  During Comrie’s term at the Galton Laboratory, Pearson brought his computing staff back to full strength and began a new round of statistical research. Either through Comrie’s influence or from his observations of the scientific world, Pearson realized that he had become one of the world’s experts on scientific computation. As he labored to train new workers, Pearson was “struck by the absence of any simple text-book for the use of computers and still more by the absence of obviously necessary auxiliary tables.”45 Before the First World War computing had been a craft skill, a loosely organized body of techniques that were passed from generation to generation like the skills of a carpenter or the knowledge of a butcher. One generation of computers had learned their techniques from Nevil Maskelyne. Another from Benjamin Peirce. A third from Myrrick Doolittle. At that juncture, Pearson realized that a new generation was learning their methods from him.

  Pearson proposed to codify the methods of computation in a series of pamphlets
, entitled Tracts for Computers, which would provide solutions for most “practical difficulties of the computer.” The name may have been inspired by the Edinburgh Mathematics Laboratory, which had published a series of tracts on the theory of numerical methods. If Pearson borrowed the title, he did not borrow the goal of the Edinburgh series. He intended that his tracts would present practical lessons, such lessons “as we have met with [in] our own experience.”46 With these lessons, a computer could develop a computing plan for any kind of numerical problem. Of all the computers of the First World War, the staff of the Galton Laboratory had handled the largest variety of problems. They had reduced data and computed ephemerides for the University of London astronomical laboratory. They had tabulated census data for the government and handled statistical correlations for Pearson. For the Munitions Ministry, they had computed trajectories and adjusted surveys. This expertise had been scattered during the last months of the war, but Pearson remained in contact with many of the computers who had served with him and could recruit a substantial pool of talent to prepare the Tracts.

  In all, the friends and staff of the Galton Laboratory completed twenty-six pamphlets. L. J. Comrie wrote one of them, and Pearson prepared two of the Tracts. Pearson’s contributions dealt with the techniques of interpolation, the process of filling in the points between two existing values. Pearson had hoped that most of the tracts might deal with similar methods, but he was only able to publish four booklets on such subjects, including the two that he contributed. The other two methodological pamphlets dealt with mechanical quadratures, or the method of small arcs, and the technique of smoothing, the mathematical means for drawing a simple curve through clouds of data.47

  In one pamphlet Pearson tried to catalog the available literature of computation. Tables and notes on computation could be found in the books and journals of at least a half dozen fields, far more than an ordinary computer could follow. He asked a colleague to prepare a bibliography of logarithm tables by reviewing the literature of physics, astronomy, optics, surveying, and engineering. Pearson claimed that scientists regularly asked him for a bibliography of tables, but he did not seem fully committed to this kind of research. In the preface to the bibliography he asked, “Has [the author] adequately supplied an admitted want?” His reply was not especially confident. “I hope it may be so,” he wrote, “but only the critics, present and future, can provide a satisfactory reply.”48

  Comrie’s tract was a table of tangents and logarithms. In all, twenty-one of the twenty-six pamphlets were mathematical tables, far more than Pearson probably intended. He claimed that these tables had “special value to the practical computer,”49 but they were an odd collection of special functions, sampling numbers, and probabilities. Many of these were originally computed during the war “because the required tables [had] not yet been published to the necessary numbers of figures, or because we did not know, or still do not know, if such tables were ever computed.”50 The Galton Laboratory computers prepared these tables for publication by checking the original text for errors, proofreading the typeset table, and preparing an introduction. The introduction often proved to be the most valuable part of the tract, for it described the mathematics behind the table and showed how the values might be employed.

  The largest table in the series filled eight volumes. It possessed the grand title Logarithmetica Britannica and embodied the nationalism that had contributed to the start of the Great War. “When it came to my knowledge that the French proposed to issue a fourteen figure table and the Germans a fifteen figure table,” Pearson wrote, “it seemed to me that it was fitting that the land wherein logarithms were cradled should rise to the occasion and issue a standard table … to twenty figures.” Through most of the nineteenth century, computers had used logarithm tables to simplify calculations by turning multiplication into addition. When the Observatory Pinafore computers sang of using the tables of Crelle, they were referring to the use of logarithms for astronomical calculation. By 1919, logarithms had only a limited role in scientific calculation, as they had been replaced by calculating machines. Pearson claimed that people who used ordinary logarithm tables for calculation “are either ignorant of the existence of slide-rules and mechanical calculators or else unfortunately cannot afford them.” The one use he saw for logarithms was in high-precision calculations, and it was for that reason that he agreed to publish the twenty place values of the Logarithmetica Britannica. It was not, he said, “an enterprise of profit.”51

  Pearson hoped to publish at least one tract describing the features of calculating machines and the techniques of machine operation, but he never found the time to write such a pamphlet or identified anyone else to do the job. The first part of this work, the description of the machines, was already covered in a German book, Die Rechenmaschinen (The Calculating Machines).52 The second part, far more difficult to write, required contributions from many individuals, as no one could claim to be an expert operator of all calculating machines.53

  The Tracts for Computers probably achieved the goals that Pearson set for them. Judging from the worn condition of most library copies, we can conclude that at least some of those computers who came of age between 1920 and 1939 learned their lessons from the wartime staff of the Galton Laboratory. At the same time, these little booklets received no critical response from the scientific community. Few scientific journals printed notices of their publication, and only one or two offered reviews of the pamphlets. Even the mathematicians most qualified to pass judgment, such as those who had served at Aberdeen, expressed no opinion on the series.54 They were simply part of the war production, part of the contribution that computers had offered to the conflict.

  CHAPTER ELEVEN

  Fruits of the Conflict: Machinery 1922

  Can a man sit at a desk in a skyscraper in Chicago and be a harnessmaker in a corn town in Iowa ….?

  Carl Sandburg, “Accomplished Facts,”

  Smoke and Steel (1922)

  THE ARMISTICE LEFT the United States with a vast pool of equipment, energy, and vision. Beginning in the winter of 1919, train after train arrived at the Aberdeen Proving Ground with field artillery pieces that had been built for a final offensive into Germany. The proving ground staff unloaded the weapons, one by one, and towed them to the large fields where Oswald Veblen had conducted his first range tests. They placed the guns in long, straight lines to await the next war to end all wars. As the army was starting a period of slow decline, they sat in winter snows and summer heat, leaving the base only when some veterans’ lodge requested a gun to use as a lawn decoration or to serve as a memorial to fallen comrades.

  The trains carried surplus punched card tabulators and sorters and punches north from the government office buildings to the warehouses of the Computing, Tabulating and Recording Company in New York. “Rising inventories became a problem in 1919 and 1920,” wrote historian James Cortada, “before commercial enterprises could sift production back to civilian levels.”1 Unlike the field artillery in storage at Aberdeen, some of this equipment would form the basis for a new class of scientific computing laboratory, a type of laboratory that would combine the tabulators with the expertise of human computers. Two veterans of the Food Administration would shape this new kind of computing facility, Henry A. Wallace and Howard Tolley (1889–1958).

  Howard Tolley was one of the last links to Myrrick Doolittle and, through him, to Benjamin Peirce. He had come to Washington in 1910 and become a computer for the Coast and Geodetic Survey office. “[The job] consisted of sitting in an office up … on Capitol Hill and running a computing machine,” he later wrote, “computing such things as the latitude and longitude of particular triangulation stations in different parts of the United States and computing the altitude of different hilltops and mountaintops.”2 Doolittle was only a few years from retirement, but he was still a guiding force in the office and taught the new computers his method of computing least squares adjustments to surveys. Tolley was initially intrigued
with this work, thinking that it was “the only part that required any knowledge of real mathematics,” but before long he recognized that the calculations “follow[ed] a regular fixed routine, requiring no judgement.”3 After a few months of adjusting surveys, Tolley tired of the work. “What is there to this?” he complained. “I [would] come over to the office every morning at nine o’clock and I [would] work on computing these things, adding, multiplying, running these computing machines, deciphering what’s in the books of these surveyors.” The job paid $100 a month, a sum that had been unchanged for nearly twenty-five years. “In effect it’s all spent before I draw it, and I [had] a pretty hard time keeping good clothes on my back.”4

  24. Howard Tolley (back row right) and Henry A. Wallace (front row center) at U.S. Department of Agriculture

  Tolley’s frustration was compounded by the knowledge that the major surveys of the North American continent were complete and that he was only handling refinements and detailed adjustments. “Just being a computer in the Coast and Geodetic Survey was completely futile,” he later remembered; “it wasn’t helping the world any.” He considered returning to college for graduate study or even joining a survey team for the Alaska railroad, but he concluded that graduate study was expensive and that the surveyors “didn’t want a desk mathematician.”5 One morning, when Tolley was chatting with his supervisor in the Coast and Geodetic Survey offices, he learned that the Department of Agriculture was seeking a general-purpose mathematician to work on some “problems that were related to genetics—Mendelianism,—and on some that were related to the capacity of farm silos.”6 When he went to interview for the job, Tolley discovered that the Department of Agriculture was most interested in what he had learned from Doolittle, the “knowledge of least squares and the adjustment of observations.”7 He accepted a position with the department and worked on a number of issues that were fundamentally economic in nature. During the war, he assisted Raymond Pearl with the statistical work of the Food Administration and, in 1921, became one of the first members of the department’s new Bureau of Agricultural Economics.8