CHAPTER FOURTEEN

Tools of the Trade: Machinery 1937

WHEN MATILDA PERSILY came to work at the Mathematical Tables Project, she followed a little ritual to prepare her tools for the day. She was part of the special computing group, one of the few who regularly used an adding machine. Her machine was a old Sunstrand, an inexpensive device that had ten keys on the top and a crank on the right side. She would first add a few meaningless numbers while sharing a moment of conversation with other members of the staff. As she talked, she would listen for the sound of grit in the gears and try to feel any catch or slip in the mechanism. She could apply some lubricating oil from a can that sat on a small shelf, but too much oil would attract the very dust she was trying to avoid. She had found that the machine was best cleaned with an orangewood stick; such sticks were sold at drugstores to groom fingernails and cuticles. Once convinced that the machine was ready, she would sit at her desk, take out her worksheets, and begin to compute.

Most of the computing machinery acquired by the Mathematical Tables Project during its first years of operation was scavenged from terminated WPA offices and other government agencies.¹ Ida Rhodes described this equipment as “the most broken down, the most ancient of contraptions.”² The majority were inexpensive, hand-cranked Sunstrand calculators. They were not designed for the repetitious work of the computing floor, nor was their crank mechanism easy for project computers to pull. Rhodes would complain, “Oh, how my arm ached by the end of the day.”³ Few of these machines were in operating condition when they arrived at the project’s office. A couple of the senior computers had learned how to transplant gears and levers from one machine to another and were able to produce two or three working machines for every four that came to their door. Through 1938 and 1939, the group had no other way to acquire calculators. A new mechanical calculator cost $400, almost as much as the $560 annual salary paid to a WPA worker.

During the Great Depression, most computing laboratories had at least one staff member with enough mechanical expertise to repair or modify their desk calculators. Even the smallest laboratory had to follow a maintenance regimen for its machines, oiling gears and adjusting levers. Observatories relied on the same technicians that kept telescopes in working order. The larger organizations, such as the Iowa State Statistical Laboratory, kept a mechanic on their staffs. The Iowa State laboratory, formerly known as George Snedecor’s Mathematical and Statistical Service, had to maintain an unusually large number of machines, as the laboratory acted as a general computing facility for the campus. It not only did calculations for any college faculty member but also provided adding machines and calculators to campus laboratories and offices. Laboratory staff kept the machines in good repair and trained the people who would use them.⁴ With such a responsibility, it is perhaps not surprising that the laboratory director in 1938, Alva E. Brandt (1898–1975?), was a trained mechanical engineer and had once served as a professor of farm machinery at Oregon State University.⁵

The computing office at Bell Telephone Laboratories probably had the greatest access to trained engineers and inventors. Regularly, laboratory scientists turned their gaze on the computing staff of Clara Froelich and saw them as a model for some new calculating device. “In these laboratories,” observed the staff scientist George Stibitz (1904–1995), “we have 10 or more girls, including at least one Hunter graduate, who spend most of their time dealing with [complex numbers].” Stibitz, a physicist, was a new addition to the mathematical division. In spite of his patronizing language toward the women, who were often a decade or more older than himself, he was one of those rare scientists who treated the computers as individuals. He seems to have been friendly with them, pausing over their desks to share a bit of conversation, comment on the upcoming weekend, and learn how they handled their calculations. He rarely described his early research without mentioning them.

By 1937, Stibitz had begun to think about building a machine that could perform complex arithmetic. Unlike L. J. Comrie in England, he was not inclined to adopt existing computing machines. “Although there are well-known rules for the use of ordinary computing machines to handle complex numbers,” he observed, “the work is tedious and likely to lead to errors on the part of the operator.”⁶ Others at Bell Telephone Laboratories had already designed calculators for complex arithmetic, but they had met with mixed success. One laboratory engineer had created a special slide rule that could multiply complex numbers. Unlike an ordinary rule, this device could account for the two parts of a complex number, the value identified as the real part and the value identified as the imaginary part. In spite of its ingenuity, the new slide rule was never adopted by the human computers. In all likelihood, they found the device too cumbersome and judged that it was faster to do the work by hand.⁷

A second computing machine for complex arithmetic had been created by Thornton Fry, but this machine was a specialized instrument. Fry’s machine, which he called an isograph, could be used to find the zeros of a polynomial. The zero of a polynomial is a value that makes the expression equal zero. These values are often called “roots,” a term that suggests the part of a plant that is underground; hence roots are the values for which an expression vanishes to nothing. For the polynomial x² − 5x + 6, one of the roots is the value 3, because for x = 3, the value of the polynomial x² − 5x + 6 equals 3² − 5 × 3 + 6 or 9 − 15 + 6, which is 0. All polynomials have roots, but they can be complex numbers and are often quite hard to find.

The isograph is best understood as an oracle for polynomials. It was not a true calculator but a device that could provide computers with information about their work. The idea had come from an analysis of gears that had been done jointly by Fry and Stibitz. The two had shown that the action of a certain combination of gears was best modeled by complex numbers. Fry took a similar set of gears and used it as the basis for the isograph. The device was big, about twelve feet in length, and was driven by an electric motor. A wheel on the left side of the machine could be used to advance the machine should the motor jam. Froelich, or one of the other computers, operated the isograph by setting a row of knobs that protruded from the base of the machine. They started their work by guessing a value that might make their polynomial equal zero. The isograph would process this guess, grinding its gears and producing a graph of circles on a plotting table. The computer would have to interpret these circles and determine whether the guess was too little or too big. The computer would then adjust the guess and ask for a second response from the machine. Continuing in this manner, with a little bit of strategy and a fair amount of patience, the computer could eventually find the values that made a polynomial equal zero.⁸

The isograph was clever, but in practice it proved no more useful than the complex slide rule. It offered assistance for only a narrow class of problems and required the computers to master a fairly esoteric set of controls. From his observations of the computing staff, Stibitz wanted to build a more intuitive machine, a general-purpose complex calculator that could add, subtract, multiply, and divide. Instead of using gears for this machine, he employed electrical circuits and binary arithmetic. Binary arithmetic was novel to computing machinery, even though it had been explored by the Harvard mathematician Benjamin Peirce some seventy years before. Peirce had noted that binary arithmetic could simplify calculation because the only symbols involved were 0s and 1s. Stibitz built upon this idea and developed an electrical circuit that could perform additions. Representing 1 as a positive voltage and 0 as no voltage, he demonstrated these ideas with a simple prototype. Borrowing a few “relays from a junk pile that Bell Labs maintained,” he assembled a circuit that would add two single-digit binary numbers. “With a scrap of board, some snips of metal from a tobacco can, two relays, two flashlight bulbs, and a couple of dry cells,” he recalled, “I assembled an adder on the kitchen table at our home.”⁹

His prototype, which he labeled “Model K” for “kitchen,” has since become a staple of elementary computer science classes. It could complete three ordinary summations: 0 + 0, 0 + 1, and 1 + 1. By pressing on the scraps of metal, he would complete a circuit, which would activate a relay and light the combination of bulbs that represented the sum. By the time he completed Model K in November 1937, he had already moved ahead to more sophisticated circuits that could deal with larger numbers and all four arithmetic operations. He worked on several ideas that fall, including a binary version of Fry’s isograph for finding polynomial roots.¹⁰ These machines were rough drafts, attempts to master the problems of binary design. None of them was built. The design for his first operational binary machine required six months of careful effort. It was a partial calculator, a machine that could multiply and divide complex numbers. Stibitz added circuits for addition and subtraction only after the original units were operational.¹¹

The complex calculator was built by laboratory technicians using standard telephone parts.¹² The machine cost $20,000 to build, about thirteen times the $1,500 salary of Clara Froelich, thirty-five times the pay of a Mathematical Tables Project computer, and fifty times the price of a traditional mechanical calculator. The Bell Laboratories computing staff seems to have kept the machine fairly busy, at least during the working day. Much of the demand on the device came from three divisions of the laboratory that dealt with telephone circuit design. The calculator was kept out of sight in a central equipment closet. The computers dealt only with a special keyboard that was connected to the calculator with ordinary telephone lines. This keyboard resembled an ordinary desk calculator. The laboratory built three of these keyboards, though only one could be used at a time, and placed them in the offices that made heavy use of complex numbers.¹³ On special occasions, the laboratory gave outside researchers access to the calculator through long-distance phone lines. Stibitz demonstrated the machine at a meeting of the American Mathematical Society in New Hampshire. The mathematicians were invited to test the calculator themselves. One of the Aberdeen veterans, Norbert Wiener, spent several hours typing at the keyboard and watching the results appear on a roll of paper.¹⁴

33. The Bell Telephone Laboratories complex calculator

The complex calculator of George Stibitz was the first of three machines that had its origins in a computing office of 1937. Taken as a whole, the three machines show how inventors adapted new technologies and new devices to the operations of computing laboratories. Stibitz developed binary arithmetic and relay circuits in order to give Clara Froelich a simple calculator for complex numbers. The second computing machine, which had its origins in the Iowa State Statistical Laboratory, attempted to meet a similar goal for a different problem, the calculations of least squares. By 1937, the laboratory had become a substantial contractor to the United States government. It received $35,000 a year to do tabulations and analyses for the WPA and for the U.S. Department of Agriculture, where the laboratory’s former patron, Henry A. Wallace, served as secretary. Its staff processed data on farm production, analyzed crop experiments, and identified the trends in agricultural markets. In spite of their best efforts, the computers had been unable to mechanize the central calculations of least squares computation. They had acquired an IBM 601 multiplying punch, the same model that could be found at the Columbia University Astronomical Computing Bureau, but even this machine could assist only with the first and last steps of least squares computation. The remaining work was done with adding machines by a staff of seven.¹⁵

The Iowa State calculating machine was built by an outsider to the computing lab, a professor named John Vincent Atanasoff (1904–1995).¹⁶ Atanasoff had been at Iowa State College for a little more than a decade. He was originally trained as an electrical engineer but had become interested in physics and had taken two years at the University of Wisconsin in order to complete a doctorate in the subject. Sometime in the early 1930s, he drifted into the statistical laboratory, intrigued with the stories of the mechanical tabulators. He did not seem to have a clear idea of how he might use the machines, for he later reported that he went “looking for a problem in theoretical physics that could be solved by IBM equipment.”¹⁷ The problem that he chose was a broader version of the least squares calculations that were regularly handled by laboratory computers, a problem called a simultaneous equation calculation.

In a simultaneous equation problem, a researcher has a certain number of unknown values that are defined by an equal number of equations. They are often taught with word problems. “When Caroline was born, Rose was twice as old as Ginny. Last year Rose was 10 percent older than Ginny. If Caroline is 28 this year, how old are Rose and Ginny?” This problem has three unknown values, the ages of the three sisters. These three values are defined by three equations, one for each sentence in the problem. Computers would find the three values by manipulating the three equations. Like complex arithmetic, such manipulations are detailed and time-consuming. The work becomes more difficult as the scale of the problem increases. A calculation with six unknown values is only twice the size of a problem with three unknown values, but it requires eight times the effort. A problem with twenty-four unknown values, the size of the least squares calculation that Alfred Cowles had proposed to Harold T. Davis, requires five hundred times the effort of a problem with three unknown values.¹⁸

John Atanasoff began his research into computing machines by modifying the laboratory’s IBM tabulator. This work paralleled the development of the isograph at Bell Telephone Laboratories. Atanasoff experimented with the punched card technology to see what he might do with it. His results were no more successful than the isograph but for slightly different reasons. In this experiment, he worked closely with laboratory director A. E. Brandt. The two of them found a common bond in the work and seemed to enjoy tinkering with machinery. Before tackling the problem of simultaneous equations, they addressed a simpler calculation, one that came from the study of the light spectrum. This calculation was difficult, if not impossible, to perform on an unmodified tabulator. Atanasoff and Brandt found a way to handle the calculation with the tabulator, but their solution involved a new circuit and a special set of punched cards. The circuit plugged into the tabulator control board and took charge of the machine whenever it encountered one of the special cards. The modifications gave “trouble-free operation over long periods of time,” according to Atanasoff and Brandt, but they handled only the intermediate problem, not the more difficult problem of simultaneous equations.¹⁹

Atanasoff designed a similar modification that would allow the tabulators to solve simultaneous equation problems, but before he could implement his idea, he lost his access to the IBM equipment. Later in life, he would suggest that IBM itself had barred him from the machines, but the reason was probably nothing more than the departure of his partner, A. E. Brandt.²⁰ In the spring of 1937, Brandt left the laboratory and took a job at the Iowa Agriculture and Home Economics Experiment Station. Brandt’s successor was more interested in statistical work than in computing machinery and may have felt that Atanasoff’s experiments interfered with the obligations of the laboratory.²¹ The reason that removed Atanasoff from the statistical laboratory is less important than the consequences of that removal. No longer able to modify the IBM equipment, he turned to the idea of building an entirely new computing machine. This machine was more technologically daring, and yet it was a better match for the computers of the statistical laboratory.

As Atanasoff would tell his story, the basic principles of his new machine came in a late-night epiphany. Through the summer and fall of 1937, he considered several different ways of building a computing machine, but none of them would accomplish what he wanted to do. “I had outlined my objectives,” he later recalled, “but nothing was happening and as the winter deepened, my despair grew.” One evening, he left for his office, hoping “to resolve some of these questions.” Instead of working at his desk, he got into his car and “started driving over the good highways of Iowa at a high rate of speed.”²² Moving away from the Iowa State campus and its statistical laboratory, he crossed half the state before reaching a roadside bar in Illinois. There, relaxed by the drive and perhaps by a drink, he identified the key elements for a computing machine that would solve least squares and simultaneous equation problems. This machine would be an entirely new device instead of a modified tabulator. It would use the binary number system, like Stibitz’s machine, so that the arithmetic could be handled by electrical circuits.²³

John Atanasoff would admit that he was “somewhat off the beaten track of computing machine gossip,”²⁴ though there was little such gossip to be found outside of Bell Telephone Laboratories or International Business Machines. With no access to the statistical laboratory and lacking an organization to support him, Atanasoff had to find a place to build his machine, as well as funds to pay for supplies and assistants. He spent about eighteen months building a simple demonstration model, his own “Model K.” This machine was more sophisticated than the machine that Stibitz had demonstrated, but it did approximately the same thing: it added two binary numbers together. It gained him a grant of $650 from Iowa State College, enough to hire an assistant and start work on the full machine.²⁵

To raise more money, Atanasoff went to the Rockefeller Foundation in New York City. The foundation was one of the larger financers of scientific research, and one of the foundation officers, Warren Weaver (1898–1978), had taught at the University of Wisconsin when Atanasoff was studying for his doctorate. The meeting did not proceed quite as Atanasoff might have hoped. Weaver had recently been hospitalized and had to be propped up with pillows.²⁶ He was not predisposed to Atanasoff, as he remembered the former Wisconsin graduate student as “rather bright but queer and opinionated.”²⁷ He patiently listened to a description of the proposed machine and firmly stated that the Rockefeller Foundation did not support such research.²⁸ Yet, in the exchange, Weaver must have seen something of value, for he mentioned that a private foundation, the Research Corporation, supported engineering projects and might be willing to provide some money for Atanasoff’s machine. He offered to let Atanasoff use his name in correspondence to the foundation and agreed to review Atanasoff’s proposal.²⁹ The dismissal may have been disheartening, but it proved to be good advice, for the Research Corporation gave Atanasoff $5,330 for his computing project. This was a substantial grant for the time, even though it was about one quarter of the money American Telephone and Telegraph had spent on the complex calculator.³⁰

Atanasoff constructed his machine in the basement of the Iowa State physics building. The device was about the size of a large desk. It had a reader for IBM punched cards and two rotating drums that held the numbers. When it was running, the two drums made a clicking noise, like a piece of cardboard slapping against the spokes of a bicycle. The operator stood in front of the machine and loaded the simultaneous equations onto these drums, one value at a time. Once an entire equation had been given to the machine, a special card punch would place all of the values on a card. Atanasoff had designed this punch so that it created holes with high-voltage sparks, rather than with a mechanical die. It recorded the numbers with a flash of blue light and a puff of smoke, an operation that would regularly singe the cards and occasionally set one alight. Each equation required one of these special cards, and each step of the calculation required that all special cards be repunched. The operator would stand in front of the machine, shuffling the cards back and forth, while the drums turned and snapped. Fresh cards would be drawn from a pile, and old cards would be discarded on a table or dropped to the floor.

34. Computing machine of John Atanasoff with operator

Atanasoff’s calculator was never a finished production machine like the complex calculator at Bell Telephone Laboratories, but by the most generous accounts, it did what it was intended to do. “It was good enough so that we were able to solve small systems of equations,” wrote Atanasoff, though he acknowledged that the high-voltage card punch was troublesome. “We made substantial efforts to solve this flaw, including changes in the card material and careful changes in the voltage use for each material.”³¹ After working on the machine for two years, he left Iowa State College in order to take a job at the Naval Ordnance Laboratory in Washington, D.C. He apparently left no one at the college who was interested in preserving his machine or in keeping it operational. When the school’s physics department decided to reclaim Atanasoff’s office space, they disassembled the device, salvaged what they could for scrap, and disposed of the rest.³²

Some thirty years later, Atanasoff’s machine would acquire notoriety as a central exhibit in a court case that contested the patent on the electronic digital computer. In the trial, it would be named the Atanasoff-Berry Computer, and its builder would be identified as the inventor of the modern computer. The controversy that followed the verdict would last for two decades more and would create partisans who claimed “that the first electronic digital computer was constructed at Iowa State by J. V. Atanasoff and [his assistant] Cliff Berry”³³ and opponents who believed with equal fervor that the machine was not a computer, as it was “premature in its engineering conception and limited in its logical one.”³⁴ This debate, with its implications for the reputations and fortunes of the participants, cannot be easily dismissed. Yet, in commanding both scholarly and public attention, it has obscured the position of the human computer in the late 1930s. Both Atanasoff’s computing machine, be it a computer or no, and George Stibitz’s complex calculator would have fit nicely into existing computing laboratories, just as the adding machines of the 1890s moved easily into the Coast Survey Office, the Nautical Almanac Office, and the Harvard Observatory.

The last computing machine of 1937 moves one step further from the offices of human computers, though it remained tied to the kinds of calculations that were being done by human computers. It was conceived at Harvard University by a graduate student in the school’s electrical engineering program. The student, Howard Aiken (1900–1973), was studying the actions of electrons in vacuum tubes. The mathematical expressions that described the electrical forces inside a vacuum tube were a messy set of differential equations. Like all other problems driving the development of computation, they could not be solved in a simple, symbolic fashion. In common with the equations of Richardson’s weather model, they described a phenomenon in three dimensions and would have required a substantial computing staff to calculate the solution. Harvard had access to funds from the National Youth Administration to pay the salaries of human computers, but such assistance was not sufficient for Aiken. “At the present time,” he wrote, “there exist problems beyond our ability to solve, not because of theoretical difficulties but because of insufficient means of mechanical computation.”³⁵

Like Atanasoff, Aiken turned from the problems of physics to the problems of calculation. He designed a machine that used gears and wheels but had a special control mechanism that provided “automatic sequencing.”³⁶ This mechanism would read a series of instructions from a paper tape and would direct the machine to perform those instructions. These instructions were almost a program, as we now use the term. By changing instruction tapes, the operator could make the machine perform complex arithmetic, solve simultaneous equations, compute orbits and trajectories, and reduce data.³⁷ In many ways, Aiken’s idea was similar to Charles Babbage’s second computing machine, the one he had called the Analytical Engine. Aiken discovered the work of Babbage while he was preparing the basic outline of his machine. He was even able to inspect a partial adding mechanism that had been built according to Babbage’s specifications by his son. This connection between the nineteenth-century mathematician and the emerging computing machines is superficial, according to Aiken’s biographer, I. Bernard Cohen. “At that time [Aiken] did not have a detailed and accurate knowledge of the purposes and principles of operation of Babbage’s two proposed machines.”³⁸

Aiken accomplished what Babbage could not: he built a working relationship between a commercial business and a scientific computing laboratory. In 1937, Aiken was older than most graduate students. Cohen has characterized him as “tall, intelligent, somewhat arrogant [and] assertive.” Aiken had supported his family since the age of fourteen, when he and his mother were abandoned by his father. As a high school student, he had taken night jobs while attending classes during the day. When he was an undergraduate at the University of Wisconsin, he had worked from four to midnight at the local electric and gas utility.³⁹ His position at Harvard freed him from the need to seek outside employment and allowed him to search for someone who might be able to sponsor his computing research.

He first presented his ideas to an engineer at the Monroe Calculator Company, “a very, very scholarly gentleman,” Aiken recalled. The engineer quickly recognized what Aiken was attempting to do and “foresaw what I did not … the application to accounting.” The engineer gave a favorable review of the machine, but the management of Monroe decided that they were not interested in the project.⁴⁰ Following this rejection, Aiken then turned to the computing staff of the Harvard Observatory. The observatory computing room operated much as it had in 1880 under Edward Pickering. A staff of computers and assistant astronomers, many of them women, measured photographs, interpreted data, and reduced the values recorded by the telescopes and sensors. The office had at least a few touches of modernity, such as mechanical adding machines, but it was more concerned with astronomy than with general methods of computation.⁴¹

35. Mark I mechanical computer at Harvard

The observatory director, through an indirect path, helped Aiken gain the attention of the senior managers at IBM. On a trip to New York, Aiken presented the IBM managers with a plan for a machine that would “be fully automatic in its operation once a process [was] established.”⁴² He visited the Columbia University Astronomical Computing Bureau, met Wallace Eckert, and studied the Orange Book. By the time his visit with IBM ended, Aiken had gained the attention of company president Thomas J. Watson. Watson was impressed with the proposal and offered to finance the project and build the machine in an IBM factory. Aiken would provide the general design and work with IBM engineers to develop the appropriate technology. Harvard would provide the computer center and operate the device.⁴³ In a move that suggested that the two groups would not long cooperate on the project, IBM decided to call the machine the Automatic Sequence Controlled Calculator, while Harvard would name it the Mark I. By the time the project was finished, IBM had invested $100,000 in the construction of the machine and donated another $100,000 to cover operational costs, a combined sum that approached the annual budget for the Mathematical Tables Project.

Viewing the new computing machines, George Stibitz prophesied that “Human agents will [soon] be referred to as ‘operators’ to distinguish them from ‘computers’ (Machines).”⁴⁴ Neither his machine nor that of John Atanasoff would take the title “computer” from human beings. The computers at Bell Telephone Laboratories may have operated the complex calculator, but they were more concerned with mathematics than with machinery. Atanasoff’s machine handled only one modestly complex step of a large process. Both inventions were intermediate devices that did not quite reach the era of stored programs and still looked back at the age of oil cans and orangewood sticks. Even Howard Aiken’s Mark I, the most sophisticated of the three machines, looked over its shoulder toward older technologies. One of Aiken’s assistants captured the traditional nature of the Harvard computing laboratory when he described the Mark I as emitting “a distinct sound, not unlike the clatter of steel-shod horse’s hooves clanging along a paved street.”⁴⁵ Aiken generally employed his computing machine in work that could have been handled by the computing floor of the Mathematical Tables Project or the First World War computers of Aberdeen or even the Nautical Almanac computers of Charles Henry Davis. Shortly after the machine began operations, Aiken produced a set of mathematical tables. His volumes covered a different set of expressions from those being prepared by the computers of the Mathematical Tables Project, but the real difference between the two sets of computations was the difference between Harvard and the WPA, not the difference between machine calculation and handwork. The WPA reproduced its tables from mimeographed stencils. The Mark I tables were typeset and printed on fine paper. The WPA books were bound in a rough tan cloth and avoided references to work relief. Aiken used a fine blue cover and printed the university seal on the title page.⁴⁶