When Computers Were Human

by David Alan Grier

  12. Coast Survey Office. Computing rooms are on the top floor

  After the war began, some of the survey computers volunteered for advance survey teams. These teams prepared maps for the Union army and charts for the navy, but this work could not be called organized computation.6 The majority of these field calculations were quite simple, a quick determination of longitude or adjustment of a triangulation point. The computers usually handled the numbers simply, writing the results in a dirty notebook or on a piece of scrap paper.7 As the historian Hunter Dupree observed about most science of the Civil War, “experimentation was largely done on the spot, when the men were actually confronted with specific and pressing problems.”8

  For those who remained behind, the Coast Survey computing floor became a haven for scientists unwilling to take up arms. The head of the office, a German immigrant named Charles Schott (1826–1901), “did not find military life congenial” and paid a substitute to take his place in a Northern regiment.9 The son of Benjamin Peirce, Charles Saunders Peirce (1839–1914), joined the office as a way of avoiding the draft. “I perfectly dread [military service],” he wrote the director of the survey; “I should feel that I was ended and thrown away for nothing.”10 In the computing office, he joined Schott in doing a mix of civilian and military computations. They did engineering calculations for the fortifications of Washington and analyzed the trajectories of new army cannons.11

  In general, the Civil War cannons did not inspire much organized computation. Though the armies and navies of Europe were starting to deploy steel guns with rifled barrels, the forces of both the Northern and Southern states used less accurate, smoothbore cannons. Shells fired by these guns rarely followed the trajectories predicted by mathematical ballistics, and so most ordnance officers needed only figures that showed the approximate range of their shots. The only real mathematical ballistics of the war was done at the experimental battery in the Washington Navy Yard, the same place where the Naval Observatory computers had learned the basic gunnery drill. The battery’s gun placement was a long, low building that faced the river. A flag on a small cupola warned nearby ships when the gunners were preparing to shoot. Its staff fired thirty to fifty shots a week across the water, rattling the windows of local houses and reminding the nearby members of Congress that the country was at war. From these shots, the officers estimated the trajectories of the weapons. Most of the computations were done by a single officer, John Dahlgren (1809–1870), who was more interested in estimating the stress on gun barrels than in predicting the flight of shells.12

  The war produced three new scientific institutions within the federal government, though none of the three had any immediate influence upon the practice of organized computing. The first organization, the Naval Inventions Board, included both Dahlgren and Charles Henry Davis. It reviewed proposals for new weapons that had been submitted to the Union navy, but it conducted no research of its own and produced no computation beyond the simple figures that the board members scratched on proposals. The second institution, the National Academy of Sciences, also listed Davis and Dahlgren among its first members. This group, modeled on the Académie des Sciences in France and the Royal Society in England, was intended to organize the nation’s scientists and advise the government “upon any subject of science or art.”13 The French Académie had been an active supporter of organized computation, as it published the Connaissance des Temps, but its American counterpart undertook no such activity during the Civil War. The early years of the institution were disrupted by controversies over the selection of the first members. The National Academy of Sciences undertook only a little research during the war, none involving large calculations.14

  The last new scientific institution of the war was a civilian organization, the Department of Agriculture. Congress, no longer checked by Southern members fearing the expansion of government institutions, formed the new department to promote agricultural production. The department’s leader, who shared the name Isaac Newton (1837–1884) with the seventeenth-century mathematician, wanted the new agency to start a new program of scientific computation, one that would gather and analyze statistics of agricultural output and consumption. “Too much cannot be said in favor of agricultural statistics,” he wrote. They reveal “the great laws of supply and demand, of tillage and barter, thus enabling both [farmer and merchant] to work out a safe and healthy prosperity.”15 In spite of his enthusiasm, the department was unable to start a comprehensive statistical program during the conflict.16

  Though the war may have produced no immediate changes in the division of computational labor, its relics slowly transformed scientific calculation. At the war’s end, the landscape was littered with thousands of miles of new telegraph wires. These wires followed the armies, Northern and Southern, as they marched across the countryside. They allowed Abraham Lincoln to second-guess his generals in the field and Jefferson Davis to follow the Confederate incursions into Pennsylvania; commanders used them to coordinate the divisions of their armies. After the war’s end, these lines were available for a national effort to collect and process weather statistics.

  Before the construction of the telegraph, the study of the weather had been a frustrating affair. Though an individual scientist could collect weather data, such numbers were of limited use unless they could be put in the context of a region or a continent. Without a full picture, a researcher might mistake a small squall for a major front, a local wind pattern for the great movement of the atmosphere. In the 1840s, the U.S. Navy and the new Smithsonian Institution had attempted to collect weather data along the northern length of the Atlantic coast. They had recruited four hundred volunteers, including Maria Mitchell’s father, given them printed data forms, and created a standard ritual for collecting data. The observers had recorded wind, temperature, and precipitation at fixed times during the day and placed the completed forms in the U.S. mail. Workers at the Smithsonian had sorted the forms and forwarded the data to Lafayette College in Pennsylvania, where a staff of computers summarized the results.17

  The joint navy/Smithsonian meteorological project was a modest success at best. As weeks passed before the data reached Washington, the results could only be used for historical study and not for forecasting. Perhaps more troubling was a lack of discipline in the system. The volunteers were told to gather data at specific times, but in fact they recorded the numbers when it was convenient for them or perhaps did not record the data at all. The scientists in Washington could not be sure that their summaries represented an accurate picture of the weather at any specific time and day. During the 1850s, the director of the Smithsonian proposed using the telegraph as a means of collecting data more quickly and imposing more discipline on the volunteers. He wrote to several telegraph companies, requesting that they “allow [the Smithsonian], at a certain period of the day, the use of their wires for the transmission of meteorological intelligence.”18 Getting the telegraph companies to cooperate on a such a large project proved to be almost as difficult as collecting the weather data. Most of the telegraph companies owned only a single line or perhaps a small collection of lines that fanned outward from a central office. Generally, these companies were only willing to transmit the data at night, a restriction that still allowed the Smithsonian to gather and transmit their data in a day’s time. By the late 1850s, the telegraph lines were sending so much data to Washington that the Smithsonian hired a staff of “expert computers” to process it all.19

  The war disrupted the Smithsonian weather project. When the collection of weather data was resumed in 1870, it was under the control of the Army Signal Corps, which controlled many of the telegraph lines that had been built during the conflict. The Signal Corps devised a system to collect weather data three times a day; at midnight, 11:00 AM, and 4:00 PM. The observers recorded their data in a coded form and telegraphed their results to Washington.20 Data moved toward the capital as if it were water coursing through the tributaries of a river. It flowed from farms and vil
lages, joining other data en route and pausing at regional offices to wait for an open moment on the lines that stretched toward Washington. In general, the rising flood of data took about three hours to reach the central office of the Signal Corps.21

  The Signal Corps created a small computing staff that processed all of the weather data in intensive two-hour shifts. Computers worked at stand-up desks, where they recorded the information on preprinted forms and blank weather maps. Each computer was responsible for only part of the data. One handled temperature, another wind, a third precipitation. They reported to their stations just as the data began to emerge from the telegraph room. Young boys carried the paper from the telegraph to a single clerk, who read each telegram aloud, translating the symbols into placenames, amounts of precipitation, and percentages of humidity. As they heard their numbers being read, the computers would add the figures to their summaries and mark their maps. After the clerk had read the last telegram, the computers required only a few minutes to complete their calculations. The “work is done as fast as the translator can dictate,” reported the director of the office, “so that within an hour all the telegraphic reports are received and within the same hour all the translating and map making is done.”22

  At first appearance, the Signal Corps computers had little in common with observatory or almanac staffs. They worked in short bursts of effort, in contrast to the sustained efforts of the astronomical computing offices, and handled calculations far simpler than the process of reducing data or preparing an ephemeris. Yet the staff of the Signal Corps was building upon ideas that had been first seen in observatories. Astronomers had seen the telegraph as a means of coordinating scientific effort across large distances. The principles of electric telegraphy had been first demonstrated by the German astronomer Carl Friedrich Gauss in the 1830s. The staff of the U.S. Naval Observatory had witnessed the first test of an American telegraph in 1844.23 The second message to travel from Washington to Baltimore in that demonstration of the telegraph, the message that followed the famous invocation from the Hebrew scriptures, “What hath God wrought!” was a question concerning astronomical research: “What is your time?”24 A difference in time corresponded to a difference in longitude, which was often expressed in values of hours rather than in the modern unit of degrees. (In some sense, we retain this old scale of longitude when we talk about Los Angeles as three hours off New York or Iowa as an hour ahead of Colorado.) The telegraph promised the ability to make measurements of longitude that were accurate to within one hundred feet. A Nautical Almanac computer devised a way to make such measurements with equipment no more sophisticated than a pair of standard pendulum clocks.25 At the Harvard Observatory, the staff attempted to extract the possible accuracy for its telegraphic measurements by installing a telegraph switch panel in its basement and placing telegraph keys on its observing chair.26

  At the Computing Division of the Coast Survey, the end of the war brought new rigor to the adjustment calculations. For many years, the surveyors had known that the adjustments could be handled mathematically, without any appeal to the judgment of computers, by a method called least squares. Using least squares, a computer could find the minimal number of adjustments that would make all the angles meet and all the sides be the proper length. The method of least squares had been developed by the same Carl Friedrich Gauss who had demonstrated the telegraph. Though many scholars have concluded that Gauss did not invent the method, they do acknowledge that most English-speaking astronomers learned the method of least squares from Gauss’s book Theoria Motus Corporum Coelestium in Sectionibus Conicis Solem Ambientium.27 The fact that this book was written in Latin was a hindrance to many American computers. Increasingly, the classical languages of Latin, Hebrew, and Greek were being dropped by the nation’s colleges and replaced with modern languages, notably German. During the 1850s, Davis had concluded that his computers needed an English edition of the book and had undertaken the translation himself. He had finished the manuscript just before he departed the almanac to take command of a ship in the Caribbean.28

  Prior to the war, the method of least squares was not widely used in the survey office because the calculations were long and arduous. It was much simpler to adjust the values by instinct and trust the judgment of the computers. The calculations for least squares required two steps. First, the computer had to prepare what were called “condition equations” or “normal equations.” A typical problem would adjust twenty-five values in a survey spread over two or three square miles. In all, there might be two or three hundred measurements that had been made by the surveyors that had to be reduced into twenty-five condition equations, one for each of the values being adjusted. The second step disassembles the condition equations in a messy series of calculations. The process can be compared to the job of disassembling a steel-frame building, moving it to a new site, and reconstructing it in a slightly different shape. As the building is reassembled in its new form, each girder needs to be adjusted with a cut or an extension. Like the process of moving a building, least squares computations required detailed record keeping. Any mistake in either the records or the arithmetic required the computer to begin the process again.29 The English translation of Theoria Motus Corporum did much to promote the method of least squares, but it described a plan for the computations that was slow and redundant. A computer knowing only the plans described by Gauss would resist any large problem.30 In the United States, the method became widely used only after a Coast Survey computer named Myrrick Doolittle (1830–1911) developed a more efficient means of doing the calculations.

  Doolittle was yet another student of Benjamin Peirce, though he had come to Peirce later in life. He had been born in Vermont and lacked the time or the money or the inclination to attend college. Proving to be a natural teacher, he had begun his career by instructing the children of friends and neighbors. At the age of 26, he accepted a position at the New Jersey Normal College and remained there until the onset of the war.31 Deciding that it was time for him to get a degree, he enrolled at Antioch College in Ohio. Antioch was a small religious college that embraced the abolitionist cause with all the fervor of the age. He studied at the college for two years and was awarded a bachelor’s degree in 1862. That summer Doolittle volunteered for an Ohio regiment in the Union army, only to find “his physical condition preventing enlistment.”32 He remained at the college for one more year as a mathematics instructor and then left to join Peirce at Harvard.

  We have no record of Doolittle’s life in Cambridge. Given the nature of his mathematical interests, it seems likely that he spent some time in the Nautical Almanac Office with the computers, but he stayed less than twelve months with Peirce. In 1864, he left Massachusetts without completing a degree in order to join his wife, Lucy Doolittle, who was working in Washington, D.C. She was a volunteer with the Sanitary Commission, a precursor of the Red Cross.33 With the help of Benjamin Peirce, Doolittle took a job at the Naval Observatory as an assistant observer. He found the observatory work “wearying,” as the job required him to record the positions of stars by night and to reduce observations by day. Deciding that he wanted to live a less taxing life, he resigned his position and took a job at the Patent Office examining the applications for patents on steam boilers.34

  During Doolittle’s time at the Patent Office, Benjamin Peirce came to Washington as the director of the Coast Survey. Peirce surprised both his friends and his detractors by proving to be a competent manager. He demonstrated that he could schedule survey crews, approve new projects, draft budgets, and work with members of Congress. He also found time among the demands of his official duties to conduct mathematical research. He returned to the classical problem of mathematical astronomy, the three-body problem, and developed a new way to compute the motion of the moon. He requested that the Computing Division prepare a table of the moon’s position from his equations, but this work had to be suspended “for want of a sufficient and adequate force of computers.” He was able to comple
te the tables only when he convinced the secretary of the navy to allocate an extra $2,000 to pay the salaries of three new computers.35

  Peirce also experimented with a new form of arithmetic which used only two symbols, 0 and 1, instead of the ten digits of decimal arithmetic. Eventually named “binary arithmetic,” it would be the basis of electronic computing machines, though such machines were seventy-five years in the future and beyond the vision of Benjamin Peirce. “I have no such extravagant thought as that of a substitute for our decimal system,” he wrote, though he believed that it might be interesting to compute “some of the fundamental numbers of science by a new arithmetic for the purpose of comparison and verification.”36

  13. Benjamin Peirce, director of Coast Survey Office

  In 1873, Peirce offered Doolittle an appointment as a computer. According to his family, this appointment brought Doolittle “into his life’s work.”37 He quickly came to dominate the Computing Division, though he never became its formal leader. He was not a mathematician after the manner of Benjamin Peirce or even Charles Henry Davis, but he had an innate grasp of calculation and was able to produce concise, simple computing plans. In 1874, he turned his attention to the calculation of least squares. He did nothing to the method itself, but he found a simpler procedure for handling the computations. He reduced the redundancies in Gauss’s procedure, dropping calculations that were handled in other ways. The result was a plan that was more orderly and more efficient. In effect, he found a way of taking down the girders and reassembling them with the fewest steps. “The results reached appear very satisfactory,” reported the director of the computing floor.38 The technique became the standard method of computation at the Coast Survey. It removed the old judgments from survey adjustment and the old “gentlemen in private life” from the computing floor.