1940 Computer Programming

The authors. In this 1994 photograph Arthur and Alice Burks are holding one of the ENIAC's accumulator decade modules, used to process one digit of a decimal number. Its 28 vacuum tubes are seen at the top.

As people grow older, their history grows longer. But if their memories do not fail too badly, the highlights of their history need not grow dimmer. Indeed, memories may become more vivid just because they have been revisited so many times. Such is the case with your present authors, who have shared a keen interest in the history of computers for over fifty years and who not only have lived some of that history but also have committed their observations to writing on numerous occasions during that half-century.

Whereas Arthur's involvement in the early days of the electronic computer revolution has been substantial, Alice's role has been almost entirely as a writer, not a direct participant. Yet her minor involvement predated Arthur's by a year and was, in fact, what brought them together. The place was the University of Pennsylvania's Moore School of Electrical Engineering, in Philadelphia, and the time was July of 1942. Arthur was already working at the Moore School, teaching and engaging in war research projects, when Alice arrived to work as what was then called a "computer," a human computer calculating artillery firing tables for the U.S. Army with the aid of a small desk machine.

We were married on February 27, 1943. Alice soon entered the University's College for Women to finish her undergraduate degree in Mathematics, and Arthur joined the new ENIAC project at the Moore School. He was to make significant contributions to the design of that pioneering electronic computer from its inception in May, 1943, until its dedication in early 1946, at which point he left the Moore School to join John von Neumann in his computer project at the Institute for Advanced Study, in Princeton, New Jersey. Then, that fall, he moved on to the University of Michigan, first in philosophy but later in computer science as well.

Arthur had gone to the Moore School in June of 1941, prior to U.S. entry into World War II, to take an intensive summer course sponsored by the government to train electrical engineers for the defense—and then the war—effort. At age twenty-five, he had just finished his Doctor of Philosophy degree, at the University of Michigan; but, with little chance of landing a teaching job in his chosen field at that time, he had decided on the defense course. Upon its completion, he was made an Instructor at the Moore School.

We are often asked about Arthur's dual identity as a philosopher and an engineer. The two pursuits seem worlds apart: philosophy, all theory; engineering, all practice. Clearly, the war was the immediate impetus, but the subjects are actually not so disparate as is usually imagined. Philosophy is a very broad discipline, with many branches, including Arthur's special interests, logic and the philosophy of science. Engineering, too, has a number of branches, now including what was to become another special interest, computer science, and, within that field, the logic of computers. Logic, then—or, more strictly, mathematical logic—was Arthur's bridge from philosophy to computer engineering.

As it happened, he had already undergone an earlier major shift of interest, from his undergraduate studies in mathematics to his graduate studies in philosophy. At DePauw University, with the help of a philosophy professor, he had been awakened to the existence of this field of mathematical logic, which seemed to him more exciting than pure mathematics. He had continued to pursue his high school interest in science, as well, with a minor in physics. He chose Michigan because it was the only school to which he applied that regarded his mathematics and science as relevant to philosophy. Two of the department's six faculty were in logic, and they understood the connection very well.

The further connection, from logic to computer design, is not so mysterious as is often imagined, either. Computer circuits perform logical functions: their designers interconnect basic components so as to perform logical operations (and, not, or, nor, at the simplest level) on incoming streams of electrical pulses representing numbers. For arithmetic applications, these numbers are direct representations; for other applications, such as word processing, they are codes for letters of the alphabet. There are many computer programming "languages," but each is a precise logical system, and each detailed computer design is a logical structure, each computer a sophisticated, powerful logic machine.

The ENIAC (Electronic Numerical Integrator and Computer) was the joint brainchild of an engineer, J. Presper Eckert, and a physicist, John W. Mauchly, with Mauchly providing more of the initial stimulus, Eckert more of the electronic expertise and ingenuity. However, they had a crucial assist from a third party, physicist John V. Atanasoff, who had built a prior electronic computer at Iowa State University and whose contribution they failed to acknowledge. Arthur was not to hear of Atanasoff's vital role until the late 1960s, a few years after the ENIAC patent was finally issued to Eckert and Mauchly and the matter of royalty payments began to haunt all the major computer manufacturers.

At the Moore School, Arthur soon made the acquaintance of John Mauchly, who at age thirty-three had left his post as Associate Professor of Physics in nearby Ursinus College to enroll in that same summer course; and of Pres Eckert, already something of an electronics expert at age twenty-two, who was the Laboratory Assistant for the course. As it happened, Mauchly had just returned from a visit to Ames, Iowa, where he had been a house guest of John Atanasoff for four days and had spent most of that time learning about the electronic digital computer Atanasoff was building at the University. Indeed, the Atanasoff-Berry Computer, or ABC, named for Atanasoff and his graduate assistant, Clifford E. Berry, was by then well along physically—and virtually complete conceptually.

John Mauchly, the only Ph.D. in the defense course besides Arthur, was also appointed an Instructor of Electrical Engineering in the fall of 1941. Meanwhile, they had become good enough friends that, with Mauchly's family unable to join him in Philadelphia because of a housing shortage, they decided to take a room together for the coming school year. Eckert now became a Master's degree student, but all three were soon caught up in military research: first, in underwater mine detonation, then in radar, and, lastly, in electronic computer design.

Mauchly, although he discussed Atanasoff's computer with Eckert and possibly with others at the Moore School that summer, did not tell Arthur about his trip to Iowa until mid-1943, as the ENIAC project was getting underway. Arthur recalls his mention of vacuum tubes and also of a rotating memory, but little else, certainly no connection to the ENIAC and no hint of the basic electronic concepts the ABC embodied. The overall impression was negative.

It was through a series of steps that Arthur learned, some twenty-five years later, that the two men who had become world-famous for the invention of the ENIAC, and with whom he had worked so closely on it, had claimed the ideas of another man as their own. He first heard of this possibility in a curious way. A New York City patent attorney named Seymour Yuter had the idea that Arthur Burks, T. Kite Sharpless, and Robert F. Shaw were actually co-inventors of the ENIAC, since they had all made essential design contributions. This idea came as a surprise to Arthur, who had believed that Eckert and Mauchly, as originators of the initial broad concept, were the sole inventors.

Yuter had learned of the ENIAC project from Shaw and Sharpless in the course of drawing up patents on their more recent inventions. Now he asked Arthur to join those two in a venture that, if successful, would give them the royalties and the credit to which they were entitled. He explained that the original designs the three of them had contributed to the ENIAC did constitute invention, for which they had been denied credit. Moreover, if their names could be attached to the ENIAC patent, they would be free to sell licenses for its use—in short, they could undercut the current royalty demands, which were universally deemed excessive.

It was a further complication that, years earlier and long before the ENIAC patent was actually granted, Sperry Rand Corporation had acquired all rights to it. So it was now this company, not Eckert and Mauchly, offering licenses to the various large corporations in the computer business.

Yuter had also learned that one of these corporations, Honeywell, was resisting the high royalty rates Sperry Rand was pressing on it. He decided to invite several lawyers from the Honeywell side to a meeting at his home, along with Arthur. Yuter explained his view that Mauchly and Eckert were not the sole inventors of the ENIAC. He then suggested that it would be good strategy for Honeywell to take a license from the three legitimate co-inventors and be prepared to defend their status in court if Sperry Rand fought back. In this way, Honeywell would have a fair royalty rate, but would carry the burden of any lawsuit.

The Honeywell attorneys, however, had a different strategy in mind. They said they had learned that a man named John Atanasoff had invented an earlier electronic computer, and because of its priority, certain of its features, and Mauchly's knowledge of it, they stood a good chance of "breaking" the ENIAC patent. Arthur immediately remembered Mauchly's negative remarks about Atanasoff's computer, all those years ago, and he was now shocked to learn that Mauchly and Eckert may have misappropriated some of Atanasoff's ideas.

From his own perspective, Arthur knew that Yuter's plan would require many hours of work on his part: reviewing old records, studying the claims of the ENIAC patent, recalling just who had done what, writing reports. And while the rewards for success would be great, he felt the chances were rather slim. He is by nature—or perhaps by training from his father—somewhat skeptical and certainly reserved in his expectations for the future.

Now he was skeptical on two fronts. Could the Honeywell attorneys establish co-invention in a court of law, if they chose that route? Or could they, on their alternative option, have the ENIAC patent invalidated on the basis of Atanasoff's prior work? If the patent was thrown out, he realized further, its value to the three co-inventors would be eradicated.

Still, he decided to work with Yuter, and as he did he grew more and more intrigued by this newly-raised historical issue of who really did invent the electronic computer—not just with regard to himself and Sharpless and Shaw, but with regard to this other figure, Atanasoff. He was to pursue this issue, off and on with his other research, for many many years. Alice, too, was gradually drawn in, first as a "sounding board" for Arthur's growing concerns and, ultimately, as an equal partner in what was to be a long struggle to establish the true history.

For a heated controversy would arise between the Eckert-Mauchly advocates, on the one hand, and the Atanasoff advocates, on the other, with each side claiming its machine was the world's first electronic computer.

By the mid-1970s, we had settled in firmly on the Atanasoff side. Long before that, Honeywell had chosen to fight the ENIAC patent, rather than try to add more inventors to it. There were years of preparation by both sides for the Honeywell vs. Sperry Rand trial, which ran from June 1, 1971, until March 13, 1972, and ended with the October 19, 1973, decision of Judge Earl R. Larson, U.S. District Court, Minneapolis, Minnesota[1]. Honeywell had prevailed. The ENIAC patent was declared invalid on grounds not only of a prior electronic digital computer, the ABC, but also of the ENIAC's derivation from it.

The outcome of this trial has been discounted, even ridiculed, by some historians who argue that lawsuits cannot produce true history because of the biases of the two sides. We have argued that, in this case anyway, a vast reservoir of material—documents, artifacts, and testimony—was accrued, often by subpoena: material far beyond what these historians could have collected through scholarly research; material subjected to the closest scrutiny by highly skilled attorneys and an astute, impartial judge; material these historians may study, too, if they will. We contend that the ENIAC patent trial did indeed disclose the true story of the invention of the electronic computer.

Let us look at that story briefly. As we have indicated, there were two issues before the court with regard to Atanasoff: the priority of his computer and its influence on the ENIAC. Now some have said, in hindsight, that priority should never have been an issue, because there was no question that the ABC was completed in 1942, the ENIAC in 1946. But the Sperry Rand defense team, relying on Mauchly's account of Atanasoff's machine and also on Eckert's derisive view of it, failed to see it as a threat. Thus the Sperry team, in order to maximize the ENIAC's value in the broad computer market, took the position that the ENIAC patent disclosed the first "automatic electronic digital computer." At trial, however, Mauchly's own testimony, as well as Atanasoff's, left no doubt that the ABC was also an "automatic electronic digital computer," and of course it was earlier.

As to the second issue, derivation of the ENIAC from the ABC, Mauchly testified that he learned nothing from his visit to Atanasoff's laboratory that he had not already conceived in his own efforts to design an electronic computer. Indeed, he portrayed those efforts as far more grandiose than Atanasoff's meager design. His problem here was that he could produce absolutely no evidence of any progress toward the conception of any novel part of any electronic computer, let alone of the ENIAC, as he was claiming.

Moreover, his own correspondence, both with Atanasoff and with others, pointed in the opposite direction. He was impressed, even excited, by what he had seen and read and heard as both Atanasoff and graduate student Berry revealed the ABC's secrets to him. He also saw the possible application of Atanasoff's electronic computing principles to a machine that he might build at the Moore School if he could get support for it.

The Mauchly-Eckert advocates, who have persisted in great force even after this unappealed decision of Judge Larson, always cite Mauchly's research at Ursinus College as evidence that he was well into designing the ENIAC before he learned of Atanasoff's work or saw his computer. And it is true that he had been interested in calculating devices for several years before he met Atanasoff, at first because he was simply attracted to them through a keen natural curiosity, but later because his research as a physicist involved great quantities of weather data that had to be analyzed to establish theories he held about physical phenomena.

At Ursinus College, where he taught from 1933 until he entered the Moore School in 1941, Mauchly had embraced an old theory that the intermittent occurrences of flares on the sun, or sunspots, coincide with rainfall patterns on the earth. He collected large stacks of data from the U.S. Weather Bureau and, with the assistance of students working under a relief program of the government in those Depression years, performed statistical analyses on these data.

But the task was enormous, even with the help of the desk calculators of that period, and Mauchly cast about for faster methods and devices. In 1939 or 1940, guided by journal articles about machines called harmonic analyzers, he built a small desktop model designed specifically for his particular problem. Although this new machine enabled him and his students to speed up their calculations significantly, the task remained enormous. As for any connection between it and the ENIAC, the harmonic analyzer was not electronic and it was not digital—the two defining features of both the ABC and the ENIAC.

But then, in late 1940 while still at Ursinus College, Mauchly did begin to entertain the idea of a desk machine that would calculate with vacuum tubes. After his death in 1980, his widow, Kathleen R. Mauchly, discovered two letters that were not submitted as evidence in the ENIAC trial, but that seemed to her to prove substantial progress toward the ENIAC several weeks before Mauchly met Atanasoff. Actually, these two short letters, to a former student and to a meteorologist friend, spoke only of a very limited calculator on which he might start work in a year or so. Moreover, his own courtroom testimony made clear that he never did build, or even design, the calculator mentioned in them.

His ambitions in that direction, however, were spurred on by his first encounter with Atanasoff, in Philadelphia on December 28, 1940. He had given a paper on weather analysis at a session of the American Association for the Advancement of Science, and when Atanasoff, a member of the audience, approached him afterward their conversation quickly turned to computers. Atanasoff, pleased to find such a knowledgeable and interested fellow-physicist, not only told Mauchly quite a bit about the electronic computer he was building but invited him out to Iowa to see it.

So it was that, after several exchanges of letters between the two men, Mauchly made the long drive to Ames in mid-June of 1941. There Atanasoff and his assistant Berry explained and demonstrated the ABC to him, shared with him Atanasoff's detailed write-up of it, and—the final touch—discussed with him a suggestion Atanasoff had made in one of his letters as to how to apply his same principles of electronic digital computation to a much more powerful machine, one that Atanasoff himself hoped to develop when the ABC was finished. It was this suggestion that, over the next two years, evolved into the Moore School's ENIAC project.

There were, of course, great differences between the two machines. Atanasoff's was small, about the size of an office desk, whereas the ENIAC was eighty feet long, its thirty units stretched in a "U" along the walls of a large room. More importantly, the ABC was a special-purpose computer, designed to perform a series of operations for the solution of large sets of simultaneous linear equations—equations of the sort that we all learn to solve in elementary algebra classes, but in sets far too large for human solution, even with the aid of desk calculators.

The procedure executed by Atanasoff's computer was extremely complex, calling for many novel features that are found in computers to this day. For example, it had a separate memory and arithmetic unit: it did its calculations in an array of adding and subtracting mechanisms and stored the results on a rotating drum memory. And while the ENIAC followed the pattern of the desk machines, with counters that added on or subtracted off numbers and, accordingly, always showed (stored) the final results, electronic computers thereafter used the concept of separation of the two functions. We are all familiar with the home computer's need for a large "memory," and also with rotating hard disk and floppy disk storage, separate from but interacting with the "processing"unit.

Despite the ABC's complexity—and the fact that the automatic solution of large sets of simultaneous equations would serve many fields of scientific inquiry—it is still properly called a special-purpose computer because of its single, hard-wired, non-programmable procedure.

The ENIAC, on the other hand, was a general-purpose computer, capable of being programmed to solve any of a wide variety of problems in science, engineering, and mathematics. We have termed the ENIAC the first general-purpose electronic computer, to honor it for that great achievement and to distinguish its priority from that of the ABC, the first special-purpose electronic computer but also, therefore, the first electronic computer.

Judge Larson's ruling of derivation of the ENIAC from the ABC rested broadly on Atanasoff's very achievement of electronic computation: he was the first to adapt the vacuum tubes of radio technology to arithmetic operations. Moreover, although the two machines had such different arithmetic and memory circuitry, the ENIAC made use of Atanasoff's basic concept of electronic switching throughout. Larson's ruling also rested on several particular claims of the ENIAC patent that came directly from the ABC.

The ENIAC project had been sponsored by the U.S. Army, to compute those same artillery shell trajectories that Alice and many other women had been hired to compute on desk calculators. The University of Pennsylvania's Moore School of Electrical Engineering had become a center for the computation of these trajectories, which were also being computed in the basement of the school by a large machine known as a differential analyzer.

Indeed, the differential analyzer, a solver of differential equations, was the missing link between the ABC and the ENIAC. For Mauchly had envisioned the ENIAC as an electronic digital version of the mechanical analog differential analyzer.

The mechanical aspect of the differential analyzer was that it was literally a machine, in the sense of an assemblage of interacting gears mounted on shafts. Its analog aspect was that it computed through the rotation of these gears, so interconnected as to simulate (be analogous to) the conditions of the problem it was solving. One can liken this action to that of a very simple odometer, which translates the rotations of an axle through a system of gears to a mileage readout on a dial. Its operation is continuous, or flowing, rather than discrete: it "counts" rotations only in the sense that the gear sizes, the numbers of their teeth, and their interconnections have been so chosen as to yield the correct translation from axle to dial.

The electronic aspect of the ENIAC was that it was an assemblage of interacting vacuum tubes. Its digital aspect was that these vacuum tubes computed by processing groups of electrical pulses that represented numbers. Its action was discrete, or step-by-step, rather than continuous.

Thus the Moore School's differential analyzer computed artillery trajectories in analog fashion (by simulation of the trajectories of specific guns under varying conditions) and mechanically (by rotation of gears). The ENIAC was designed primarily to compute these same trajectories digitally (arithmetically) and electronically (with vacuum tubes).

As we have seen, Atanasoff had designed his computer to solve simultaneous linear equations, also digitally and electronically. And it was Atanasoff, not Mauchly, who took the difficult and seminal step of adapting the vacuum tube of radio, itself an analog device that emitted its signals in continuous waves, not discrete pulses, to digital computation. We have also seen that it was Atanasoff who suggested to Mauchly that the electronic computer he was building at Iowa State could be extended to do the work of the differential analyzer many, many times faster.

After considerable consultation with Eckert, Mauchly wrote a short preliminary proposal in August of 1942, a little over a year after his visit to Atanasoff.[2] It was March of 1943 before this paper stirred enough interest to prompt a more detailed proposal, now feverishly written by Eckert and Mauchly together and approved for Army Ordnance support in early April.

Events had moved swiftly, and they continued to do so as the Moore School became a site of tremendous activity, attended by great energy, and enthusiasm. A group of engineers, mathematicians, physicists—even a philosopher whose expertise in logical relationships proved valuable—worked together in a common conviction that they were breaking new ground in science and mathematics. Although the ENIAC was not finished in time for the war effort, it saw a decade of valuable service before the new stored-program computers rendered it obsolete.

Arthur has always been proud of his work on the ENIAC and proud of the ENIAC itself. Because of his later support for the Atanasoff side of the dispute, he has been accused of disloyalty, or even disgruntlement over not having received his full share of credit for his own contributions. He felt that he had no choice, however, but to recognize Atanasoff's work once it became known to him. Moreover, he has always felt privileged to have spent the war years in such interesting and challenging work. Alice has been of the same mind.

Arthur did play a significant role in the ENIAC project. He participated in the general planning sessions for the entire machine; he conducted design experiments and raised critical issues with Eckert and Mauchly; and he carried out three major assignments. Chief among these assignments was the design of the electronic circuitry for the High-Speed Multiplier, the first such multiplier ever invented. This task was a mix of logical on-off considerations and the engineering of timed voltage swings and current strengths, with every detail subject to the constraints of an intricate set of rules laid down by Pres Eckert to assure that a computer with 18,000 vacuum tubes would be reliable.

Arthur's second major contribution was the fundamental organization of the Master Programmer, the central control unit of the ENIAC, which was then carried out by Robert Shaw after a modification or two from Eckert. His third was the checking of the circuit diagrams of all the units to see that they were logically correct and in accord with the prescribed electronic principles, both individually and in relation to other units. This was a painstaking task, in which he had the assistance, first of Kite Sharpless, then of Shaw; as might be expected in such a large, groundbreaking enterprise, there were numerous design errors to be chased down.

But it was a natural assignment for Arthur, who had been charged with taking notes at the design meetings and helping with the progress reports. It provided an understanding, too, of the overall system of the ENIAC and of the details of its many components, an understanding that proved invaluable to his later recording and teaching of this history.[3]

Arthur was to have a role, as well, in the immediate aftermath of the invention of the ENIAC—again, both in the actual developments and in the preservation of the history. We indicated above that, revolutionary as it was, the ENIAC was soon outmoded by the stored-program computer. Indeed, the stored-program concept was the third and final critical step toward the creation of today's computers, the first having been Atanasoff's very concept of an electronic digital computer, the second the Eckert-Mauchly concept of programmability in such a computer.

For the stored-program concept encompassed the basic characteristics of the modern computer. To be sure, there have been countless truly ingenious improvements in computing technology since that third step was taken; yet the fact remains that these have been ever smaller, faster, cheaper, more powerful manifestations of that stunning series of inventive concepts.

The mammoth ENIAC was severely limited by its manual method of entering the program for a given problem: that is, entering the particular set of instructions for manipulating the arithmetic data being fed in from punched-card machines. For each problem, the programmers were faced with a giant, eighty-foot-long "plugboard" requiring two or three days of setting switches and plugging in cables. This was not a great drawback in computing artillery trajectories—the main rationale for building the ENIAC—because for each trajectory the same setup could be used over and over, for several weeks, with only the arithmetic data changing. It was a serious drawback, however, for the other types of problems the ENIAC's designers had hoped to include in its repertoire.

Long before the ENIAC was finished, Eckert and Mauchly were envisioning a computer that would avoid this programming bottleneck. And for it they would use the Atanasoff scheme of separate memory and arithmetic unit. In fact, they would use electronic adders that were faster versions of his original mechanisms. But instead of his necessarily slow rotating drums, they would use a new form of memory incorporating several novel features that were to carry the computer revolution forward dramatically. This memory would be large enough to store the program and the arithmetic data internally at the start of a problem run; it would be set up quickly via magnetic tape entry of both; and it would be readily erasable at the end of each run.

This next computer, built at the Moore School—also under Army Ordnance contract—was the EDVAC (Electronic Discrete Variable Computer). Its new mercury-delay-line memory was invented by Eckert, with crucial help from Sharpless, as an adaptation of an earlier acoustic-delay-line timing device invented by Bell Telephone Laboratories' William B. Shockley.

These advances, on the part of both Eckert and Mauchly, held tremendous promise for the ease with which an unprecedented range of mathematical and scientific problems could be solved. But a third figure, John von Neumann, was to enter the picture with further, equally ingenious advances that would propel the stored-program concept over the threshold of modern technology. They would also mark the conclusion of the wartime efforts of these two Moore School figures with a dispute that eclipsed, for sheer bitterness, the dispute with Atanasoff that marks its beginning.

The world-renowned mathematician von Neumann had come to the Moore School as a consultant in the fall of 1944. Thrilled to learn of the ENIAC and of plans for the EDVAC, he joined in a series of discussions of those plans in March and April of 1945. He then produced what was to become a famous—or infamous, the Eckert-Mauchly side would say—document, "First Draft of a Report on the EDVAC," setting forth in logical symbolism the entire structure of the proposed computer, together with a set of rules for solving problems on it.[4]

This approach, starting with the logical structure from which the electronic design would follow, was not only completely novel; it was to be adopted as standard procedure. The set of rules, or programming language, was also completely novel; variations on it also became standard. On top of these advances, von Neumann suggested that a cathode ray tube, then being introduced for television, would be a better form of memory than the mercury delay line because of its much greater storage capacity. He went on to use this form of memory in his Institute for Advanced Study Computer, while the Moore School stayed with the Eckert form.

The ensuing dispute between Eckert and Mauchly, on the one hand, and von Neumann, on the other, is too complicated and convoluted to relate here. The crux of the matter, as we have analyzed it, is that the stored-program concept at stake in this quarrel had two distinct parts: the first contributed by Eckert, the second by von Neumann.

Eckert clearly had the original inspiration for a large erasable memory, together with an original version of it. Both he and Mauchly, however, thought of this only in terms of its (tremendous) improvement over the ENIAC's hand-entered, unchangeable form of storage. It was von Neumann who saw that this erasable feature of the mercury delay line—and likewise of the cathode ray tube—would allow the program to change its own instructions in the course of a problem run, depending on that problem's needs at any given juncture. This advance, incorporated in the programming language worked out by von Neumann in his 1945 EDVAC report and adopted for both the EDVAC and the IAS machine, completed the concept of a stored program as required—taken for granted—in all modern computers.

Finally, for that Institute computer, von Neumann carried his cathode-ray-tube concept into a computer memory that permitted random access to a large grid of information on the inner surface of the tube's screen. Whereas the EDVAC accessed a number in its mercury-delay-line memory serially, digit by digit, the IAS machine accessed digits in parallel, an entire number at a time. This random-access type of memory (RAM) was one more advance that became standard, though it would ultimately take the form of the integrated circuit, or chip, that we know today. It was in this stage of von Neumann's work in electronic computing that Arthur was privileged to play a role.

As to the dispute over historical credit, it is ironic that what von Neumann did, essentially, was take the Eckert concept and run with it—into further new and uncharted territory, just as Mauchly and Eckert had taken Atanasoff's earlier concept and run with it. But Eckert and Mauchly then claimed not just Atanasoff's basic ideas, in their ENIAC patent, but particular features of his computer, in other patents; and but for blockage by both the Moore School and Army Ordnance, they would have applied for a patent on the entire EDVAC, including von Neumann's contributions. In contrast, von Neumann was determined to protect his ideas so far as his intellectual legacy was concerned, but he never tried to patent his own contributions to the Moore School or the Institute computers, let alone anyone else's.

The special irony, for us as historians of this era, lies in the fact that Eckert and Mauchly felt that their advances over Atanasoff's work were so great as to entitle them to claim his ideas along with their own, but deeply resented von Neumann's claim to his very considerable advances over their work.

Eckert and Mauchly left the Moore School in early 1946, soon after the official unveiling of the ENIAC to the public. They left to start their own business when they could not reach agreement with the University of Pennsylvania over patent rights for employees. Eckert's leadership in the design and construction of the EDVAC fell to Sharpless, and the EDVAC team shifted many times before that earliest conception of the stored-program computer was finished in 1952.

"EDVAC-type" computers were built by Mauchly and Eckert (the BINAC and the famous UNIVAC, the world's first commercially marketed electronic computer) and by many other institutions around the world; "IAS-type" computers were also built by various institutions (including IBM, with its commercial version, the IBM 701).

When Eckert and Mauchly left the Moore School, they urged Arthur to join them, but he chose instead to accept von Neumann's invitation to work on the Institute computer. He made it clear, however, that he wished to return to philosophy in the near future. In fact, he had already, in the fall of 1945, begun to teach at nearby Swarthmore College, evenings and early mornings, while still working full-time at the Moore School. Now he commuted by train to Princeton three days a week as he finished out the spring term at Swarthmore, then five days a week for the summer of 1946.

At Princeton, Arthur worked with both von Neumann and mathematician Herman Goldstine, the Army officer who had served as liaison between the Moore School and Army Ordnance during the war. With von Neumann in the lead, the three of them co-authored a monograph on the general design of the Institute for Advanced Study Computer, "Preliminary Discussion of the Logical Design of an Electronic Computing Instrument"; this widely read work provided the paradigmatic form of what became known as the "von Neumann architecture" for electronic computers.[5] Arthur also wrote for the IAS project a draft of a report on a "library of subroutines" to be used in programming.

Finally, he wrote two articles on the ENIAC in the spring of 1946. One was a "popular" description called "Super Electronic Computing Machine" requested by ElectronicIndustries.[6] The other was a technical description, "Electronic Computing Circuits of the ENIAC," first given as a talk to the Institute of Radio Engineers and later published in its Proceedings.[7]

As we have said, Arthur became a member of the University of Michigan Department of Philosophy that fall. But he never left computing. He returned to the Institute for Advanced Study for the summers of 1947 and 1948, and he consulted for the Burroughs Corporation in Detroit one day a week until 1955. Then he founded his Logic of Computers Group at the University that continued until his retirement—a group whose very name joins philosophy and computer science. He and a colleague, Gordon Peterson, also founded a program to train doctoral students in computer science. In 1967, this became the Department of Computer and Communication Sciences in the College of Literature, Science, and the Arts; Arthur was finally officially half in philosophy, half in computer science.

Arthur's return to writing about the early history began a few years after the ENIAC patent trial ended. He gave a paper in 1976 at an international conference on the history of computing, held at Los Alamos Scientific Laboratory in New Mexico.[8] This paper was the first to address in any depth the Eckert and Mauchly debt both to Atanasoff for the ENIAC and to von Neumann for the EDVAC. Mauchly read two papers expressing opposing opinions, one written by himself and one written by Eckert.[9]

Arthur's position on the debt to Atanasoff was based on a pair of documents, copies of which came into his hands in early 1974 at the conclusion of the ENIAC patent trial. The first was Judge Larson's decision. This 100-page ruling required careful study, not so much because of its legal terminology as because of what might be called its cumulativestructure. That is, every section presumed and built upon all the previous sections. Such tight reasoning was challenging to the lay reader, but was also increasingly satisfying for its logical structure and the intricacy of its argument.

Arthur paid particular attention to Judge Larson's argument that Eckert and Mauchly had derived the "subject matter" of the ENIAC "from one Dr. John Vincent Atanasoff," and he found himself agreeing with the judge. He was pleased, too, with the ruling that the ENIAC project had been a team effort. Larson went so far as to cite the "inventive contributions" of Sharpless, Burks, Shaw, and others, and to cite in particular the "major contributions" of Burks and Sharpless.

The second document that influenced the views Arthur expressed at the Los Alamos conference was a manuscript Atanasoff had written in 1940, now published for the first time in a collection of papers on the origins of digital computers.[10] This paper, which took the form of a proposal for funding, had been a key exhibit in the trial, not only because it was a detailed disclosure of the ABC but also because Atanasoff had permitted Mauchly to study it, even take notes on it, when he visited him in Iowa.

Arthur set about studying this manuscript, as well, another hard task because it was at least as concise as Judge Larson's decision. Not a word was wasted. And, as Arthur also discovered, not an element in the computer itself was wasted, either. Atanasoff clearly had an economical turn of mind. He continued to take satisfaction, in his later years when we came to know him personally, in the efficiency of his add-subtract mechanisms—only seven triodes in each of the thirty. Today, as scientists at Iowa State University reconstruct the ABC, they are repeatedly amazed at the ingenuity of its design details. We and they, of course, attribute a portion of that ingenuity to Clifford Berry; Atanasoff was wise in his choice of a graduate assistant.

Arthur, because of his own familiarity with vacuum-tube technology, was able to work through the design features and the electronics revealed in the text and in the drawings, charts, and photographs. Again, he was satisfied that Judge Larson had not erred in his decision.

When, upon publication of this paper, Arthur was asked to write a much longer article on the ENIAC project, he began a series of contacts with Atanasoff. By that time, he was teaching a course in the history of computers at the University of Michigan, and he had secured stacks and stacks of copies of trial documents from a variety of sources. He wanted to learn as much as he could, firsthand, about the ABC, its inventor, and the linkage of both to the ENIAC and its inventors.

By that time, Alice had joined him in his efforts to tell the full story of the invention of the electronic computer. She had studied both Larson's decision and Atanasoff's manuscript, and she was delving into the trial records. Over the next few years, the two of us interviewed Atanasoff extensively, in person at his home near Frederick, Maryland, and by telephone, and we came to know both him and his wife Alice very well. Our immediate and lasting impression was of a man who, while wanting full credit for his own achievements, eschewed credit for anyone else's—and could tell the difference.

Our 60,000-word article, "The ENIAC: First General-Purpose Electronic Computer," was published in 1981 in Annals of the History of Computing.[11] At the end were comments, invited by us, from all mentioned parties who cared to respond, positively or negatively. This article generated substantial further response, in the Annals and in many other print formats, usually expressing appreciation for our exposition of the ENIAC itself but then focusing on our presentations of the Mauchly-Eckert disputes with Atanasoff and von Neumann.

After it appeared, we began work on a book to be devoted entirely to Atanasoff: his computer, his influence, his place in the history. This book, The First Electronic Computer: The Atanasoff Story, was published by the University of Michigan Press in 1988[12], a few days before the Iowa State University Press published the late journalist/lawyer Clark R. Mollenhoff's Atanasoff: Forgotten Father of the Computer [13], and a few months before ScientificAmerican carried the late physicist Allan R. Mackintosh's article, "Dr. Atanasoff's Computer".[14] We believe these three works will, sooner or later, carry the day for Atanasoff as inventor of the electronic computer.

Arthur wrote and taught in both philosophy and computer science until his retirement in 1986. Intriguing questions have arisen from this combination, some of them ancient philosophical questions in technological form. The Greek atomists, for instance, maintained that atoms moving in space are the basic entities of the universe and that humans are compounds of atoms: our skeletons are composed of interlocking atoms, our minds of smooth atoms moving rapidly through our bodies. If we replace this simple picture with the complexities of modern physics, chemistry, biology, and the social sciences, we have the conjecture of very complex finite automata, or robots, approximating if not equaling or surpassing humankind—a conjecture that is being addressed today, in a variety of guises, by both philosophers and computer scientists.

It is a curious circumstance that Arthur was drawn to philosophy in part by a play about robots that he saw as an undergraduate at DePauw University, years before the invention of the electronic computer made the theory and the actual manufacture of robots very serious subjects; and that one of his most enduring academic pursuits at Michigan has been the study of the logical structure of finite automata.

In Karel Capek's famous science-fiction play, R.U.R., written in 1920, the brilliant Dr. Rossum had succeeded in manufacturing reliable robots to replace humans at routine and tedious tasks in factories, homes, and offices; but when the element of self-interest was incorporated, as well, Rossum's Universal Robots organized and by the end of the play had revolted and killed nearly all of their human exploiters.

The basic issue raised by R.U.R. is whether automata can be designed to perform all human functions. For example, can they be designed to reproduce themselves? It was none other than John von Neumann who, in the early 1950s, used computer logic and technology to propose a model of a self-reproducing automaton: a robot that could, at least in theory, be programmed with its own description and instructions to duplicate itself any number of times. It would be composed of structural elements, basic electronic switches and memory units, sensing devices, and acting devices; and it would be placed in an environment of sufficient quantities of all its components to be used as needed in the assembly process.

When von Neumann had difficulty formulating this idea more precisely, he consulted his friend, Stan Ulam, of Los Alamos Scientific Laboratory, who suggested the concept of what is now called a cellular automaton. One was to imagine an indefinitely large chessboard with identical finite automata in the squares, one per square and each connected to its immediate neighbors. Before his death in 1957, von Neumann had worked out the fundamental features of the design of his automaton and its process of self-reproduction. Arthur was asked to take over the task of completing the design and editing and completing the partial manuscript into a book.[15]

Together with colleagues in a broad range of disciplines, he has continued to explore logical systems, particularly as to the possibility of creating machines that can learn from their own experiences. Needless to say, this topic is being explored throughout the world, both in theories about artificial intelligence and artificial life and in concrete applications such as chess-playing computers, factory robots, surgical instruments, and a host of other applications in science, industry, business, and finance.

This intensified study of automata is but one of the many ways in which the invention of the electronic computer has made practical the traditional philosophical questions of, say, determinism, consciousness, goal-directedness, social responsibility, teaching and child-rearing, evolution, free-will, and ethics. Arthur has addressed many of these in his writings and public lectures as well as his courses in both philosophy and computer science. The University of Michigan's College of Literature, Science, and the Arts published a collection of these works in a book titled Robots and Free Minds.[16]

We are both continuing to write on the early history of the electronic computer, wanting particularly to round out the three-stage development from the special-purpose ABC to the general-purpose ENIAC to the stored-program EDVAC and IAS computer. We find this effort very demanding, but satisfying and often exciting, not least because of the human element. We view it as an important undertaking, in the spirit of clarifying the story of an invention that has impacted all of our lives in so many ways, and of seeing justice done to the original participants. And we view ourselves as in a unique position to ease the work of future historians and obliged to do so as well as we can.

Lastly, it would seem that apportioning credit where credit is due is a basic tenet not only of fairness but of academic integrity. It should also help to counter a prevailing cynicism in today's culture that tends to discourage creativity except in the interest of the narrowest short-term advantage.

NOTES

An earlier version of this essay was published in Japanese in the February and March 1996 issues of the journal, Bosei, by Tokai Educational Research Institute, in association with Tokai University.

ENIAC Trial Records. Pretrial depositions, affidavits, complaints, transcripts, exhibits, briefs, and decision. Honeywell, Inc. vs. Sperry Rand Corp. et al. No. 4-67 Civ. 138. D. Minn. Filed May 26, 1967, decided October 19, 1973. General Services Administration, Federal Records Center, Chicago. Decision published in U.S. PatentQuarterly 180:673-773.
John W. Mauchly, "The Use of High Speed Vacuum Tube Devices for Calculating," 1942, originally unpublished. The Origins of DigitalComputers: Selected Papers, edited by Brian Randell (Berlin: Springer-Verlag, 1973), 329-32.
There is an exhibit of a large part of the original ENIAC at the west end of the Atrium of the Electrical Engineering and Computer Science Building on the North Campus of the University of Michigan. There are also an explanatory display cabinet and three short videos of film taken in 1946, with voice explanations added by Arthur Burks.
John von Neumann, 1945. "First Draft of a Report on the EDVAC," privately distributed. Reprinted, Papers of John von Neumann on Computing and Computer Theory, edited by William F. Aspray and Arthur W. Burks (Cambridge, MA: MIT Press, 1987), 3-82.
Arthur W. Burks, Herman H. Goldstine, and John von Neumann, "Preliminary Discussion of the Logical Design of an Electronic Computing Instrument" (Princeton: Institute for Advanced Study, 1946). Reprinted Papers of John von Neumann on Computingand Computer Theory, 97-142.
Arthur W. Burks, "Super Electronic Computing Machine," ElectronicIndustries, 1946, 5:62-7, 96.
Arthur W. Burks, "Electronic Computing Circuits of the ENIAC," Proceedings ofthe Institute of Radio Engineers, 1947, 35:756-67.
Arthur W. Burks, "From ENIAC to the Stored Program Computer: Two Revolutions in Computers," A History of Computing in the Twentieth Century, edited by N. Metropolis, J. Howlett, and Gian-Carlo Rota (New York: Academic Press, 1980), 311-44.
John W. Mauchly, "The ENIAC," A History of Computing in the Twentieth Century, 541-50, and J. Presper Eckert, "The ENIAC," A History of Computing in the Twentieth Century, 537-39.
John V. Atanasoff, "Computing Machine for the Solution of Large Systems of Linear Algebraic Equations," 1940, originally unpublished. See The Origins of Digital Computers, 305-25.
Arthur W. Burks and Alice R. Burks, 1981. "The ENIAC: First General-Purpose Electronic Computer," Annals of the History of Computing 3:310-99. With comments by John V. Atanasoff, J. G. Brainerd, J. Presper Eckert and Kathleen R. Mauchly, Brian Randell, and Konrad Zuse, together with the authors' responses.
Alice R. Burks and Arthur W. Burks, The First Electronic Computer: TheAtanasoff Story (Ann Arbor, MI: University of Michigan Press, 1988).
Clark R. Mollenhoff, Atanasoff: Forgotten Father of the Computer (Ames, IA: Iowa State University Press, 1988).
Allan R. Mackintosh, "Dr. Atanasoff's Computer," Scientific American, 1988, 259, No.2:90-96.
John von Neumann, Theory of Self-Reproducing Automata, edited and completed by Arthur W. Burks (Urbana, IL: University of Illinois Press, 1966).
Arthur W. Burks, Robots and Free Minds (Ann Arbor, MI: College of Literature, Science, and the Arts, University of Michigan, 1986).