[CS Home Page] [Top Level - All Titles] [Top Level - This Title] [Expand This View] [Collapse This View] [Previous Hit] [Next Hit] [Clear Search] [Show Frames]

Expand Search


Feature

IEEE ANNALS OF THE HISTORY OF COMPUTING 1058-6180/98/$10.00 © 1998 IEEE
Vol. 20, No. 4: OCTOBER-DECEMBER 1998, pp. 5-9


The author can be contacted at:
Rm. 5128, MRC636
National Museum of American History
Smithsonian Institution; Washington, D.C. 20560-0636, U.S.A.
e-mail: kidwellp@nmah.si.edu

Stalking the Elusive Computer Bug

[PDF ICON]

Peggy Aldrich Kidwell

From at least the time of Thomas Edison, U.S. engineers have used the word "bug" to refer to flaws in the systems they developed. This short word conveniently covered a multitude of possible problems. It also suggested that difficulties were small and could be easily corrected. IBM engineers who installed the ASSC Mark I at Harvard University in 1944 taught the phrase to the staff there. Grace Murray Hopper used the word with particular enthusiasm in documents relating to her work. In 1947, when technicians building the Mark II computer at Harvard discovered a moth in one of the relays, they saved it as the first actual case of a bug being found. In the early 1950s, the terms "bug" and "debug," as applied to computers and computer programs, began to appear not only in computer documentation but even in the popular press.

INTRODUCTION

Stalking computer bugs[mdash]that is to say, finding errors in computer hardware and software[mdash]occupies and has occupied much of the time and ingenuity of the people who design, build, program, and use computers.1 Early programmers realized this with some distress. Maurice Wilkes recalls that in about June of 1949:

I was trying to get working my first non-trivial program, which was one for the numerical integration of Airy's differential equation. It was on one of my journeys between the EDSAC room and the punching equipment that "hesitating at the angles of the stairs" the realization came over me with full force that a good part of the remainder of my life was going to be spent in finding errors in my own programs.2

U.S. inventors had long known that calling problems "bugs" suggests that they are small and manageable. How did those working with computers adopt and modify this usage; and how did the words "bug," "debug," and "debugging" spread among computer users and the general public? This small story in the history of slang has occupied several columns of the Annals over the years, and the 20th anniversary of the magazine seems an appropriate time to review the topic. Examination of published and unpublished sources reveals a somewhat more complex tale than what is often presented and suggests questions for future research.

Dictionaries and anecdotes offer clues to the origin of the term "computer bug." The second supplement to the foremost dictionary of the English language, the Oxford English Dictionary (OED), quotes a 1945 citation of the word "debugging."3 This might lead one to suspect that the use of the word "bug" to mean a flaw in the components or programming of a machine dates roughly from the time of the introduction of the computer. Indeed, as was pointed out in the first issue of the Annals of the History of Computing, there is an object associated with this etymology. According to that account, when Grace Murray Hopper and her associates were testing the Mark I computer at Harvard University, the machine suddenly stopped. On inspection, they found a dead moth that had shorted out one of the circuits. Removing the insect, they taped it in the logbook of the machine4 (see Fig. 1). Hopper prepared a slightly longer report of the incident that was published in the Annals in 1981. Here the time is given as the summer of 1945, but the computer is the Mark II, not the Mark I. A photograph shows the moth taped in the logbook, labeled "first actual case of bug being found." Small problems with computers have been called bugs ever since.5


Figure 1. The logbook of the Mark II for 9 September 1947, with moth. (Photograph courtesy of Smithsonian Institution)

Many parts of this story are true. Howard Aiken, Hopper, and their associates did build and test the Mark II computer at Harvard. A moth did get caught in one of the relays, stopping the machine. It was removed, taped in a logbook, and labeled "first actual case of bug being found." However, as John H. Palmer pointed out in 1991, full-scale construction of the Mark II did not begin until about the spring of 1946: There was no Mark II to test in 1945.6 The logbook in question was transferred to the Smithsonian Institution from the Naval Surface Weapons Center at Dahlgren, Virginia, in the early 1990s. Jon Eklund, the curator charged with responsibility for the object, noted then that the insect was found in 1947, not 1945.7

The Grace Hopper Papers in the National Museum of American History (NMAH, Smithsonian Institution) Archives Center and documents elsewhere show that by 1947, Harvard personnel had been using the word "bug" to describe problems with their computers for several years. Indeed, they apparently adopted the terms "bug," "debug," and "debugging" from contemporary U.S. engineering slang.

In seeking to discover the origin of technical terms, one must be cautious when consulting standard sources such as the OED. The full title of the OED is A New English Dictionary on Historical Principles: Founded Mainly on the Materials Collected by the Philological Society. The first edition was published in 125 serial parts between 1884 and 1925 (entries starting with B came out in 1888). In stating that the dictionary was based on "historical principles," the editors meant that definitions were made by examining several short texts in which a word was used. The view that a dictionary should be based on the way words were used, rather than dictating correct usage, had developed in Germany in the early 19th century, then spread to France and England, and eventually would reach the United States.8

The editors of the OED were not interested in just anybody's English. They insisted that the texts they consulted had been written down and, whenever possible, published. They also believed firmly that common and literary words should be at the core of the dictionary. Scientific and technical terms, as well as slang, were of interest primarily if they had entered everyday speech or been used in nontechnical literature. Americanisms were included if they were used by both American and English authors. When the first edition of the OED came out, the word "bug," meaning a flaw in a machine, was technical slang, apparently used principally in the United States. As such, it had no place in the OED. John Willinsky has pointed out that the OED also included few citations from the working-class press or from legal documents. Some of the first uses of "bug" meaning flaw appeared in precisely this form. Here again, the sense of the word lost out.9

By the time the supplement to the second edition of the OED appeared in 1972, the editors had broadened their criteria to include technical words and even technical slang. In addition to the word "debug," which I have mentioned, the OED editors included the definition of "bug": "a defect or fault in a machine, plan or the like." This usage, they reported, originally came from the United States. The earliest published example of the meaning they knew about was from 1 March 1889 in the English serial, the Pall Mall Gazette. The citation reads: "Mr. [Thomas] Edison, I was informed, had been up the two previous nights discovering 'a bug' in his phonograph[mdash]an expression for solving a difficulty and implying that some imaginary insect has secreted itself inside and is causing all the trouble."10 Fred R. Shapiro drew attention to this early use of the word "bug" in a 1984 article in the Annals.11

U.S. sources reveal a bit more about the history of "bug." The English-speaking inhabitants of the United States have long shown a tendency to expand and multiply the meanings they give to the word. This first occurred with the entomological sense. In 19th- and early 20th-century England, "bug" was usually used narrowly for what Americans now call bedbugs. As early as 1744, English settlers in Massachusetts used the word more often to refer to insects generally. Noah Webster noted this usage in his dictionary of 1828. It also appears in the first edition of the OED.12

Another sense of the word "bug" that is more relevant to computer bugs appeared in the two-volume dictionary published by Funk and Wagnalls in the mid-1890s. Consultants for this dictionary included Ralph W. Pope, who was secretary of the American Institute of Electrical Engineers and editor of the serial Electric Power. Perhaps because of Pope's influence, the dictionary included over 4,000 definitions of terms relating to electricity. One of these was the fifth meaning given to the word "bug," namely, "a fault in the working of a quadruplex system or in any electrical apparatus."13 This sense of bug had not been included in Webster's First International Dictionary of 1858, since quadruplex telegraphy had not yet been invented. Webster's New International Dictionary, published in 1938, would give as one definition of bug "a defect in apparatus or its operation."14

Quadruplex telegraphy, the transmission of four telegraph messages over one wire, became common in the mid-1870s. The 15 August 1875 issue of a magazine for telegraphers called The Operator describes goings-on at various telegraph offices in Upstate New York. Interest in sporting events in Saratoga, New York, had led to the introduction of a quadruplex system to handle expected heavy traffic on telegraph lines. As has occasionally happened in other technical systems, the "quad," as it was called, stopped functioning when it was most needed. Looking for the cause of this "bug," the chief operator found that one of the rheostats had been adjusted improperly.15

Beginning in 1873, the inventor and former telegrapher Edison had devoted considerable attention to improving quadruplex telegraph systems. When two messages were sent in one direction, one message could mutilate the other. Edison found a way to isolate the disturbance electromechanically, preventing interference. Cross-examining Edison about this invention in April 1877, E.R. Dickerson, a lawyer for Western Union, described circuits in which "there would be made a false signal, and they call that a 'bug' for short." Edison's invention was called a "bug trap."16

Edison also used the word "bug" to describe the numerous small problems an inventor must overcome to proceed from a flash of insight to a product. I have already quoted an 1889 article in which he used the word this way. John Lord has pointed out that in a letter of November 1878, written as Edison began work on electric lighting, he wrote: "This thing gives out and then that[mdash]"Bugs"[mdash]as such little faults and difficulties are called[mdash]show themselves and months of anxious watching, study and labor are requisite before commercial success[mdash]or failure[mdash]is certainly reached."17 I said earlier that by calling flaws in systems "bugs," engineers and, later, programmers suggested that they had only minor problems that would be resolved with a bit of patient effort. Here, Edison was trying to minimize difficulties in development of the electric light and to reassure an anxious business associate. The little faults he had yet to overcome included such basic problems as finding a suitable material for the filament of the light bulb and designing and producing an evacuated glass bulb to hold it.18

The term "bug," then, was slang used by U.S. telegraphers and inventors to refer to problems in their machines. By 1940, this usage had established itself as part of the vocabulary of U.S. engineers. Also by this time, Aiken of the Physics Department at Harvard University had persuaded IBM to build a room-size programmable computing machine known as the Automatic Sequence Controlled Calculator (ASCC), or Mark I. Both the Mark I and its successor, the Mark II, used thousands of relays to control the flow of current through the machines. Instructions and tabular information were entered on punched paper tape, with data supplied on punched cards or set on a bank of switches. Malfunctions could and did result from faulty components, poor connections, flawed paper tapes and cards, and switches set incorrectly. The ASCC Mark I arrived at Harvard in February of 1944 and was installed with the assistance of IBM engineers (see Fig. 2). I.B. Cohen has examined a photocopy of the logbook kept by Robert Campbell, a young physicist at the Computation Laboratory. It tersely summarizes hours spent finding and correcting errors. On 17 April 1944, Campbell wrote, "Ran test problem. Mr. Durfee from I.B.M. was here to help us find 'bugs.'"19


Figure 2. U.S. Navy personnel who operated the ASCC Mark I at Harvard during World War II. Robert Campbell is on the far right in the front row, sitting next to Grace Murray Hopper. Howard Aiken, who envisioned the machine, is on Hopper's other side. Grace Murray Hopper Papers, NMAH Archives Center. (Photograph courtesy of Smithsonian Institution)

The Mark I, unlike earlier computing devices, could be programmed. Programs were coded on paper and then punched onto paper tapes. One of the people most concerned with programming the Mark I was Hopper, who had left her Vassar mathematics professorship to enlist in the Waves (the Women's Reserve of the U.S. Navy). Hopper arrived at Harvard in early July of 1944. She soon was acutely aware of the bugs that could appear in coding and punching the paper tapes. Instructions she prepared for running one problem on the machine included "a debugging note to operators."20 This problem ran from November 1945 to March 1946. In a contemporary list of people who worked on solving the problem, Hopper reported "errors in mathematics and tape bugs pursued and captured by Ensign Bloch and Ensign Campbell."21

Hopper also drew a cartoon showing bugs she encountered working on the Mark I (Fig. 3). The "table worm," which arose from errors in paper tapes, is in the shape of a Mark I paper tape. The "kitchie boo boo bug" was associated with loose relays. The Naval Research Laboratory's "NRL bug" sent the wrong data for Mark I calculations, throwing off the results. Hopper did draw a face with a big smile for "he who brings good data." The cartoons were drawn over several days in July. The year is not given, but it seems unlikely that she would have been so deeply involved in programming the month she arrived at Harvard in 1944. A 1945 date seems probable.22


Figure 3. Grace Hopper's pencil drawing of bugs in the Mark I computer. Hopper Papers, NMAH Archives Center. (Photograph courtesy of Smithsonian Institution)

Impressed by work done on the Mark I, the U.S. Navy ordered a modified version of the machine for the Dahlgren Proving Ground in Dahlgren, Virginia. This machine was designed, built, and tested at Harvard beginning in 1945 and moved to Dahlgren in 1948. The Mark II was even larger than the Mark I, with a control console and control panels in front and six large cabinets of relays and other electrical components in the back.23 On the afternoon of 9 September 1947, computer operators at Harvard were testing the arithmetic units of the Mark II when the machine's action stopped. Checking the circuits, they found the moth already mentioned and taped it in their logbook.

The terms "bug" and "debug" spread gradually among practitioners of the evolving discipline of computer science. Engineers and mathematicians who built and programmed early computers made lists of the words they used while working with the new machines. The first of these glossaries, prepared by a committee of the Institute of Radio Engineers in 1951, made no reference to "bug" or "debug."24 As more people worked with computers, special magazines and organizations developed to serve their needs. Edmund C. Berkeley's journal, Computers and Automation, includes an early glossary of common terms in the field. "Debug" is listed as a colloquial phrase meaning "remove malfunctioning conditions from a computer." There is no listing of "bug."25

In the late 1940s and early 1950s, the Eckert-Mauchly Computer Company (later Remington Rand's Univac Division) in Philadelphia began building electronic computers for commercial use. The company hired Hopper and others to develop programs for the new machine and training materials for programmers. In 1953, Hopper prepared a glossary for Univac programmers who attended one of two Workshops on Automatic Coding held in Washington, D.C. This revised version of the glossary was published the next year. Here Hopper defined "debug" as a colloquial phrase meaning "to remove a malfunction from a computer or an error from a routine."26 In other words, she explicitly extended the colloquial use of the term "bug" to include errors in programming. Remington Rand, like other early computer manufacturers, distributed standard programs for its customers. Programs provided for the Univac I included an "Automonitor" routine. A January 1953 description of it began, "It frequently becomes necessary in the debugging of routines to follow step by step each operation involved in a specific computation." Documentation for another program, called "Codesearch," explained that it would be used if "a bug is encountered in a routine and on typing out of the memory, the programmer discovers that a memory location, 289 for instance, contains decimal zero instead of the expected instruction."27

The 1953 Hopper glossary mentioned above was the basis of a longer document prepared by the Committee on Nomenclature of one of the first organizations of computer scientists, the Association of Computing Machinery. Here "debug" is defined "to isolate and remove all malfunctions from a computer or all mistakes from a routine." The word is listed as one of the terms used by programmers, not as slang. There is no mention of "bug."28 In 1962, the ACM Sub-Committee on Programming Terminology, chaired by W. Barkley Fritz, prepared a revised glossary of programming terms. This included a definition of "bug" as well as "debug."29 Both terms were routinely included in later glossaries.

In the early 1950s, the phrase "computer bug" also began to appear in the popular press. For example, the "Grin and Bear It" cartoon (see Fig. 4) prepared for Sunday, 23 March 1952 by the cartoonist George Lichty (George Maurice Lichtenstein) shows a large greenish-yellow box with appropriate colored lights, switches, cables, and digit-shaped dials that is labeled "Eureka Electronic Brain." Several plump men stand in front of the machine in bow ties and suits; one wears a lab coat over his suit. The caption for the cartoon reads: "We've eliminated the last 'bug', gentlemen . . . the human element! . . . it'll now do an income tax return without blowing a fuse!" Hopper cut out the cartoon from a Philadelphia newspaper and saved it in her files.30


Figure 4. "Grin and Bear It" by George Lichty, 23 March 1952, Hopper Papers, NMAH Archives Center. (Photograph courtesy of Smithsonian Institution. Reprinted with special permission of North America Syndicate.)

In summary, the phrase "computer bug" has served several functions. Talking about bugs rather than flaws appeals to the wit of programmers and computer users. In the early days of computing, the phrase was used primarily by a relatively small group of experts, marking their special status. At the same time, having a short general term that covers a wide variety of flaws has been useful, particularly because the exact source of the error was often unknown. Calling such problems bugs rather than failures suggests that they are small faults that can be corrected[mdash]not a general failure in thinking or design. This usage had long been a refuge of U.S. inventors and served the optimistic nature of computer designers and programmers.31

In conclusion, I wish to mention several topics relating to computer bugs that merit further historical attention. The first is the large subject of the evolution of the vocabulary used to describe computers, both in English and in other languages. Where and when did terms arise, how were they adopted, and how did they spread from one language to another? Some German sources from the 1960s use rechner for computer and fehler for a bug (the latter is simply the German word for error). The English words "computer" and "computer bug" apparently are now widely used in German speech. How did the German terminology develop, and how was it affected by the English? In French, ordinateur, meaning computer, is still widely used. However, the term for computer bug is a cognate, bogue informatique (bogue does not mean a small insect in French).

Second, historians might also enjoy tracing humor surrounding the phrase "computer bug" as it evolved from Hopper's cartoons and the moth in the logbook for the Mark II. Recent examples abound. For example, in about 1990, the Boston computer maintenance firm Cosmic Enterprises distributed a plastic flyswatter to advertise its services. The object's slogan boasts, "We'll Get the Bugs Out." Images of bugs appear on decals, on books about computer bugs, in printed advertisements, in science fiction novels, and on jewelry. Recruiters trying to attract programmers distribute fluffy green and orange "computer bugs."

Third, and more substantially, what stories can we tell about the debugging of specific hardware and software? Surviving records are often fragmentary. Martin Campbell-Kelly's account of an EDSAC program that he found in both an early and a later form offers a model of the kind of tales that can be told.32 Surviving libraries of computer programs, perhaps in conjunction with oral histories and correspondence, may offer other clues.

Another large topic that merits attention is the history of program editors, debuggers, and other programs designed to assist in or carry out debugging. How, and to what extent, have these become standard commercial products? What inspired them, and to what extent have they been proprietary information?

Finally, early computer users could not always tell why a computer was malfunctioning, but they could usually clearly distinguish between errors in what would now be called hardware and software. The components of early computers were not designed using computers. This is no longer the case. One of the most publicized failures in computer hardware of the 1990s, an error in the divide unit of early versions of the Pentium microprocessor, was attributed by its manufacturer to an error in programming by the designers of the chip. An error in software led to a hardware bug.33 Such a blurring of categories suggests that the meaning of phrases such as "computer bug" will offer challenges to both historians and lexicographers well into the future.

Acknowledgments

This paper builds both on earlier scholarship and on the patient work of numerous people, particularly colleagues at the National Museum of American History and the Smithsonian Institution Libraries. Jon Eklund, Nance Briscoe, Joan Nichols, Jim Roan, and Uta C. Merzbach deserve special thanks.

Notes and References


1. Fred Brooks, known for his contributions to the operating system for the IBM System/360, estimated over 20 years ago that testing and debugging took up roughly 50 percent of the time required for such a software project. He saw no reason to alter his estimate in a recent update to his book, The Mythical Man-Month. Even after programs are released, maintaining and debugging continue. See F.P. Brooks, The Mythical Man-Month. Reading, Mass.: Addison-Wesley, 1995, esp. pp. 19-20, 121-132, 231.
2. M. Wilkes, Memoirs of a Computer Pioneer. Cambridge, Mass.: MIT Press, 1985, p. 145.
3. R.W. Burchfield, ed., A Supplement to the Oxford English Dictionary, vol. 1. Oxford, England: Clarendon Press, 1972, p. 747.
4. J.J. Horning, "Additional Viewpoints," Annals of the History of Computing, vol. 1, no. 1, pp. 69-71, 1979.
5. G.M. Hopper, "The First Bug," Annals of the History of Computing, vol. 3, no. 3, pp. 285-286, 1981.
6. J.H. Palmer, "The First Bug[mdash]Discussion," Annals of the History of Computing, vol. 13, no. 4, pp. 360-361, 1991,
7. J. Eklund, "The Final Word on 'The Bug,'" Annals of the History of Computing, vol. 14, no. 3, pp. 6-7, 1992.
8. On the history of the Oxford English Dictionary, see K.M.E. Murray, Caught in the Web of Words: James A.H. Murray and the Oxford English Dictionary. New Haven, Conn.: Yale Univ. Press, 1977.
9. John Willinsky has done a computerized search of the citations in various editions of the OED to find out which sources were used and which were ignored. See J. Willinsky, Empire of Words: The Reign of the OED. Princeton, N.J.: Princeton Univ. Press, 1994.
10. R.W. Burchfield, ed., A Supplement to the Oxford English Dictionary, vol. 1. Oxford, England: Clarendon Press, 1972, p. 377.
11. F.R. Shapiro, "'The First Bug' Examined," Annals of the History of Computing, vol. 6, no. 2, p. 164, 1984.
12. S. Clapin, A New Dictionary of Americanisms. Detroit: Gale Research Co., 1968, p. 80. (This is a reprint of the 1902 edition.) W.A. Craigie and J.R. Hulbert, Dictionary of American English on Historical Principles, vol. 1. Chicago: Univ. of Chicago Press, 1938, p. 343-344.)
13. I.K. Funk et al., eds., A Standard Dictionary of the English Language. . ., vol. 1. New York: Funk & Wagnalls, 1895, pp. iii, 248.
14. W.A. Neilson et al., eds., Webster's New International Dictionary of the English Language, 2nd ed. Springfield, Mass.: G. & C. Merriam Co., 1938, p. 350.
15. R. Rattlepate, "The New Office[mdash]The Force and Its Aspirations[mdash]Chief Chase After a 'Bug,'" Operator, vol. 3, Aug. 15 1875, pp. 7-8.
16. On the history of Edison's "bug trap," see R.V. Jenkins et al., eds., The Papers of Thomas A. Edison, vol. 2. Baltimore: Johns Hopkins Univ. Press, 1991.
17. John Lord quoted this passage in a letter to Byte magazine of July 1984. His letter was republished, with discussion, by Henry S. Tropp, in a note entitled, "Whence the 'Bug,'" Annals of the History of Computing, vol. 10, no. 4, pp. 341-342, 1988.
18. For a discussion of early uses of the term "bug," see F.R. Shapiro, "Etymology of the Computer Bug: History and Folklore," Am. Speech, vol. 62, pp. 376-378, 1987. On Edison and the electric light, see R. Friedel and P. Israel with B. Finn, Edison's Electric Light: Biography of an Invention. New Brunswick, N.J.: Rutgers Univ. Press, 1985. Edison made no claim to have coined this use of the word "bug." Skimming prior volumes of a magazine for telegraphers called The Telegrapher, I found no earlier examples of use of the word in this sense. Any citations would be welcome.
19. I.B. Cohen, "The Use of 'Bug' in Computing," Annals of the History of Computing, vol. 16, pp. 54-55, 1994.
20. "Operating Instructions Problem L[mdash]Determination of Which Machine Function Is Causing Trouble: A Debugging Note to Operators," Box 2, Grace Murray Hopper Papers, Collection 324, Archives Center, National Museum of American History, Smithsonian Institution, Washington, D.C.
21. "Problem L," Box 6, Hopper Papers, Archives Center, National Museum of American History, Smithsonian Institution, Washington, D.C.
22. G.M. Hopper, Drawings of Computer Bugs, Box 6, Hopper Papers, Archives Center, National Museum of American History, Smithsonian Institution, Washington, D.C.
23. Harvard University, The Staff of the Computation Laboratory, "Description of a Relay Calculator," Annals of the Computation Laboratory of Harvard Univ., vol. 24, 1949. This volume describes the Mark II, with extensive illustrations.
24. R. Serrell et al., "Standards on Electronic Computers: Definitions of Terms," Proc. I.R.E., vol. 39, pp. 271-277, 1951.
25. E.C. Berkeley, "Glossary of Terms in the Field of Computing Machinery," Computers and Automation, vol. 2, no. 4, p. 18, May 1953.
26. G.M. Hopper, "A Glossary of Computer Terminology," Computers and Automation, vol. 3, no. 5, p. 16, May 1954.
27. "Service Routines. . .," 12 Jan. 1953, Box 3, G. Goldstein Papers, Collection 554, Archives Center, National Museum of American History, Smithsonian Institution, Washington, D.C.
28. Association for Computing Machinery, Committee on Nomenclature, Report to the Association for Computing Machinery: First Glossary of Computing Terminology. Washington, D.C.: Association for Computing Machinery, 1954, p. 7.
29. Association for Computing Machinery, Sub-Committee on Programming Terminology, ACM Glossary of Terms in the Computer and Information Processing Field. Washington, D.C.: Association for Computing Machinery, 1962, pp. 10, 25.
30. George Lichty Cartoon, Box 6, Hopper Papers, Archives Center, National Museum of American History, Smithsonian Institution, Washington, D.C.
31. For a general discussion of the purposes served by slang, see E. Partridge, Slang To-Day and Yesterday With a Short Historical Sketch; and Vocabularies of English, American, and Australian Slang, 3rd ed. London: Routledge & Kegan Paul Ltd., 1950, pp. 6-7.
32. M. Campbell-Kelly, "The Airy Tape: An Early Chapter in the History of Debugging," Annals of the History of Computing, vol. 14, no. 4, pp. 16-26, 1992.
33. Accounts of the error in the divide unit of early versions of the Pentium chip are still found largely on sites on the World Wide Web. For one discussion, see T.R. Halfhill, "The Truth Behind the Pentium Bug," Byte, Mar. 1995.

Peggy Aldrich Kidwell watches over the Mathematics Collections in the Smithsonian's National Museum of American History, where she has prepared exhibits on topics ranging from intelligence testing to the history of the metric system to computer bugs. She holds a PhD in the history of science from Yale University. Her research interests include 19th- and 20th-century computing devices, and she presently is working on studies of the adding and calculating machines at the National Museum of American History.