Abbate 2012
Abbate, Janet. Recoding Gender: Women's Changing Participation in Computing. MIT Press, 2012.
Why were women encouraged to take up skilled computing work in the 1940s and 1950s, a period when many other technical professions did not welcome them? And given this early success, why are there not more women in the computing professions today?
Page 12
The invisibility of women also reflects a more general bias: early histo- ries of computing largely equated computing with hardware. This narrow focus led to a relative neglect of software, theoretical computer science, and the practice of programming. Since women rarely had the chance to participate in building computers, the initial focus on hardware uninten- tionally biased the history of computing toward the activi…
Page 16
Before the 1950s, the tasks that we associate today with computers were done by an array of special- ized machines. Mechanical desk calculators were used to perform arith- metic, and the people who did these calculations—often women—were called computers; this term did not yet refer to a machine. 4For business and government record keeping, there were electromechanical tabulators that sorted and tallied information that was stored on punched paper cards. Punched-card machines had been invented by Hermann Hollerith to tabulate the 1890 U.S. Census; in 1911, Hollerith’s firm became the core of a new company called IBM, which quickly dominated the busi- ness. 5Scientists sometimes did complex calculations using analog com- puters, which used physical motion, such as the rotation of a cylinder, to represent mathematical functions or model physical processes. Analog computers had been introduced in the 1930s, most notably the Differen- tial Analyzer invented by Vannevar Bush at the Massachusetts Institute of Technology. A general-purpose computer that could handle all of these diverse …
Page 21
Colossus was a boon to the Allies, allowing 13,508 messages to be de- coded by war’s end. The second Colossus machine was built in a rush to have ready in time for D-Day.
Page 25
On the other side of the Atlantic, the U.S. Army was exploring the use of machines to tackle a very different problem—ballistics calculations.
Page 25
The goal was to compute firing tables, which showed gunners how to aim their weapons to hit a target at a particular range. The calculations needed for such a table were complex. To create a single firing table required a month of continuous work for either the Differen- tial Analyzer or a team of a hundre…
Page 25
Like Colossus, the ENIAC used
Page 25
vacuum tubes—18,000 of them—to perform calculations, making it the first fully electronic computer in the United States (figure 1.2). To program the ENIAC, the army recruited a team of six women. The original pro- grammers were Jean Jennings (Bartik), Betty Snyder (Holberton), Fran- ces Bilas (Spence), Kay McNulty (Mauchly Antonelli), Marlyn Wescoff (Meltzer), and Ruth Lichterman (Teitelbaum). 17Several additional wom- en worked on the ENIAC after it was transferred to Aberdee…
Page 26
One of its major tasks while still at the Moore School was a set of calculations that modeled thermonuclear bombs in 1946.
Page 26
How did women come to staff the world’s first electronic digital com- puters, a seemingly masculine engineering domain? The short answer is that they were actively recruited during a time of urgent need and scarce male labor.
Page 27
But invoking wartime labor scarcity begs the question because labor scarcity is not a given fact, it is a social construction that depends on assumptions about what types of people are suitable to fill which types of positions. Other solutions to the problem of supplying both fighters and workers were certainly possible. Why not allow women to serve in combat so that men could keep their existing jobs? Why not recruit black men for skilled wartime jobs instead of confining them to the lowest- paid work? These alternatives would have been unappealing, even un- thinkable, to most political and business leaders (and women themselves) because the same cultural attitudes, priorities, and power relations that had excluded women and minorities from certain occupations before the war were still in pla…
Page 27
It is worth asking what made computer work seem possible and desirable for these women. In other words, what defined this type of work as a good opportunity rather than something unimaginable or unappealing?
Page 28
Women’s only involvement in building the ENIAC hardware was performing factory-like assembly of components.
Page 29
24On the other hand, operating and programming c ers could more easily be constructed as women’s work. Since the jobs of computer operator and programmer did not exist before the war, they were not already stereotyped as masculine.
Page 29
In the scientific world, women had been employed in large numbers to do calculations by hand. These human computers generally, although not always, had some training in mathematics, and it was from the ranks of such mathematically trained women that the ENIAC project drew its first corps of programmers.
Page 29
In the separate world of administrative data processing, women typically held the low-level job of key-punch operator (the person who punched holes in the paper cards used for input) or the slightly more ad- vanced job of tabulating machine operator (the person who loaded decks of cards into the machine and initiated its operation). The existence of female tabulator operators may have provided a precedent for employing women as computer operators at Bletchley Par…
Page 29
Unlike prior women’s work, such as hand calculations, programming and operating computers were completely new tasks with unknown demands. No one had ever operated an electronic digital computer before Colossus.
Page 33
A review of the Colossus and ENIAC projects reveals that in both cases, project leaders underestimated the level of training needed and skill exer- cised by their female staffs. But even though gender bias played a large role in shaping women’s responsibilities and rewards in the workplace, this role was mediated by the open-endedness of the technology itself. Computers were new and complex enough that the work of programming them could be organized in many possible ways and their personnel could extend the boundaries of their jobs to incorporate new ski…
Page 33
Despite all this preparation, the ENIAC women were not given any specific training in programming—a telling omission.
Page 35
Rather than having a clear image of programming as merely clerical, as some historians have argued, the ENIAC designers seem to have had only the vaguest notion of what programming might involve.
Page 35
…orry about those things later.” 44For the ENIAC women, the undefined nature of their job was challenging, but it also gave them the freedom to define program- ming—through their actions and achievements—as a wide-ranging and intellectually fulfilling task.
Page 35
Keeping the decoding process on track therefore depended on the Wrens’ mechanical ingenuity, and as Ireland noted, they keenly felt this responsibility: “That was a tricky operation, getting the tape to the right tension.
Page 36
The operators also had to devise methods for keeping the tapes in good repair, as Dorothy Du Boisson recalled: “Each tape . . . had to be joined into a loop. It was difficult to get the join right, and if we didn’t the tape might not stand up to the speed of the machine. After many experiments, we found that special glue, a warm clamp, and French chalk produced a good joint—the art was using just the right amount of glue.”
Page 36
To check whether the decoding was successful, the operator set switches on the front of the machine to program it to count the frequency of certain letters. These counts were printed out on a nearby typewriter. If the cryp- tographer chose the wrong wheel pattern, the results had no particular pattern of letter usage, but if the correct wheel pattern had been found, the output had a nonrandom pattern that was characteristic of the lan- guage being used (in this case, German military lan…
Page 36
To be a successful Colossus operator therefore required physical endur- ance, mechanical aptitude, attention to detail, and the ability to memorize codes and do mental arithmetic. Female operators were also entrusted with a certain amount of initiative.
Page 36
Despite the various physical and mental abilities that they employed, Wrens were not treated as skilled workers or expected to understand the logic behind the machine’s operations—at least at first.
Page 37
As noted above, the ENIAC women were offered no instruction in pro- gramming (indeed, no one was qualified to teach it), and no manuals were available until Adele Goldstine produced one in June 1946, well after the original programmers had trained themselves. 55The women learned by trial and error, discussing ideas among themselves with little more than engineering drawings to guide th…
Page 39
Designing programs for the ENIAC took considerable creativity, since every programming technique had to be worked out for the first time. The women made a number of innovations in programming methods. Betty Snyder is credited with inventing break points, a debugging tech- nique that involves stopping the machine in the middle of a program to check intermediate resu…
Page 41
The women also developed a system of programming notation that allowed them to visualize the multiple simultaneous opera- tions that the computer performed at each step of executing a program.
Page 41
The programmers also developed new methods in numerical analysis. 63For example, a fundamental concern in applied mathematics is that for many problems, the computer calculates an approximate rather than exact solution, which can lead to significant errors. The ENIAC women responded to this challenge by creating ex- perimental programs to generate knowledge about the magnitude of these errors and ways to minimize the…
Page 41
In addition to the various programming methods that the women in- vented, Snyder and Jennings created an important software product—the program to calculate ballistics trajectories, which had been the original goal of the ENIAC and would be the centerpiece of its public debut in 194…
Page 41
One common experience for the Colossus and ENIAC women was that neither group received public recognition for their work, either dur- ing the war or for many decades afterward.
Page 45
…ongoing struggles to define the nature of programming skill allow us to observe some of the ways in which the culture of com- puting is based on and contributes to structures of power and meaning in the wider society. Skill is a social construct: neither the skills required to do a job nor the skill possessed by an individual can be defined in purely objective terms. Although it is possible to identify physical, intellectual, or social abilities that are relevant to particular jobs, there is never only one way to define these requirements. Other abilities can be substituted, work processes can be reorganized, different tools can be used, and varying criteria can be applied to job performance. 7Given this flexibility, social judgments often determine which job-related capabilities are seen as nec- essary and which become optional or even invi…
Page 49
Since technical skill conveys p cluding prestige, access to well-paid employment, and the opportunity to shape the tools used by a whole society—the dominant groups in society tend to assert their “natural” superiority in these fields. In particular, tech- nical expertise has been an important component of masculine identity in Western cult…
Page 49
Although no reliable national figures exist for the years before 1960, an estimated 1,200 to 1,400 programmers were employed in the United States in the mid- 1950s. 11In 1960, the U.S. Census Bureau counted 13,000 “professional/ technical computer specialists,” and by 1970, this had soared to 163,000 programmers and 108,000 computer systems analysts and scientists—a twentyfold increase in one decade. Women accounted for 24.2 percent of these programmers (39,000) and 13.6 percent of analysts (15,000). 12Esti- mates for the United Kingdom range from the Economist’s figure of “about 6,500” programmers in 1967 to an industry expert’s approximation, for the same year, of 16,330 programmers and 10,160 analy…
Page 50
Another confusing variable was the scope of the programming task. Practitioners often divided the work of producing a program into two conceptual categories—systems analysis and programming.
Page 51
In the context of computing, however, systems analysis meant analyzing the problem that was to be tackled—whether a missile trajectory or a payroll—and specifying a general method of solution, including what types of information would be needed as input and output from the computer. Programming then meant the detailed design, writing, testing, and debugging of the actual computer code that would turn the input provided into the required output. Systems analysis was usually seen as having greater status than programming, and jobs for systems analysts usually offered higher salaries.
Page 51
Many installations employed programmer-analysts who handled the entire spectrum of tasks.
Page 52
…the need to find proxy measures for programming ability. It was simply not practical to observe a job applicant directly as he or she spent months creating a complex piece of software, so employers searched for other indicators that they hoped would correlate with coding skill.
Page 54
Thus the characterization of computers as an engineering tool, a mathematical device, or a business machine mattered greatly in constructing females as possessing the skills necessary for programming.
Page 63
The association of programming with math proved helpful to many women. Women were more likely to major in math than any other sci- ence, earning 27 percent of U.S. math B.A.s in 1960, 33 percent in 1966, and 37 percent in 1970. 66Women were also more likely than men to have entered programming with a math backgrou…
Page 64
To help make women feel welcome, female programmers and those urging their employment used metaphors that associated programming aptitude with more “feminine” pursuits such as knitting, music, and cooking, thus casting women’s participation in computing as natural and
Page 74
desirable. Some of these appeals to feminine ingenuity may have been attempts by employers to tap women’s labor potential without seriously challenging the dominant gender system that constructed technical com- petence as masculine.
Page 75
…m- puting Laboratory, where she began programming in the 1940s, also used needlework as a reference point for skill: “Nobody, of course, was trained in computers, so they were looking to hire people with certain character- istics, like if you played chess.
Page 76
As these comments illustrate, some experts asserted that women were even more suited to programming work than men. They attributed this to supposedly feminine personality traits or skills, such as patience and attention to detail.
Page 76
…a double ambiguity: (male) managers were unsure which intellectual capacities might be linked to gender, but they were also uncertain about which skills were actually required for program- ming. There was a temptation to reason backward: if women did well at programming, it must be because programming utilized stereotypically feminine skills such as patience and meticulo…
Page 77
A 1964 survey of “ tional Interests of Computer Programmers” noted “the prevalent belief in a relationship between programming and musical activity” and found modest statistical support for it.
Page 77
Laboratory argued in 1959 that employers should seek a greater diversity of talents among programmers: “It is undesirable to overemphasize the sciences and neglect the humanities and letters. Yet too often systems analysts and pro- grammers are regarded as recruitable exclusively from mathematicians, engineers, and the like. This is myopic. The computer world does and will require many skills among its peopl…
Page 78
…ents have been described in other histories of computing, this chapter pays particular attention to the multiple, contest- ed meanings of such terms as automation, crisis, and engineering to show how the ongoing debate over programming methods was also a s…
Page 82
to define the programmer’s professional and social identity.
Page 83
Programming innovations were not just technical milestones. They were also social commentary, because each new method reflected ideas about labor and gender.
Page 83
Grace Murray Hopper laid out one of the earliest descriptions of the new approach in her widely cited article “The Education of a Computer,” which she first presented at the May 1952 ACM national meeting. There Hopper boldly proclaimed, “It is the current aim to replace, as far as possible, the human brain by an electronic digital computer.”
Page 84
Wheeler, and Stanley Gill confessed as much in The P aration of Programs for an Electronic Digital Computer (1951), which is widely considered to be the first programming textbook.
Page 85
The first programming languages, called machine languages, specified only the most basic hardware operations, such as copying the contents of a memory location to the machine’s adder or multiplier.
Page 85
One of the trickiest chores was handling floating- point operations.
Page 85
…the early improvements in programming methods—subroutines, compilers, and high-level languages—constituted a largely bottom-up movement by programmers themselves who wanted to achieve better results with less drudgery. Labor, rather than management, led the way.
Page 86
The earliest attempts to improve programming practice sought to re- duce coding time and errors by reusing portions of working programs, so that fewer pieces of code would have to be written and debugged from scratch. These reusable sections of code became known as subroutine…
Page 86
the Harvard computation lab during World War II, for example, Hopper and her colleagues kept a notebook of completed programs and copied sections of these programs for use in new programs.
Page 87
A series of incremental advances during the 1950s took programmers from using machine code plus subroutines to high-level languages that allow code to be written using English words or mathematical notation and that provide simple ways to specify the logical flow of a program. The earliest systems that provided programmers with a language other than machine code were pseudocodes, which resembled machine code but specified more complex actions than the machine’s hardware could actu- ally perform. A program written in pseudocode had to be interpreted by another program that would translate the code line by line into machine instructions, executing each instruction before moving on to the ne…
Page 87
In 1951, Hopper, then head of programming for UNIVAC, took a decisive step toward high-level languages with her most significant invention—the compiler. With the interpreter system, the pseudocode had to be reinter- preted each time that it was run—a painfully slow process on contem- porary machines. In contrast, Hopper’s new program, called A-0, would read through the pseudocode and compile the specified subroutines into a complete program in …
Page 87
For computer manufacturers and marketers, there was also a symbolic appeal. Unlike interpreters, the compiler created a program: the machine was “programming itself.” Hopper continued to refine her invention, and by 1953 her third compiler, A-2, was in tri…
Page 87
14A though FORTRAN made it possible to write programs much faster, it was really the quality of IBM’s FORTRAN compiler—which produced code as efficient as that written by hand—that won over programmers.
Page 88
Another important algebraic language, called ALGOL (Algorithmic Language), was conceived by a committee of European and American computer scientists in 1955, with the first specifications pub- lished in 1958. 16The first business programming language was created by Hopper in 1955. With her usual terseness, she called the new language B-0, but it was renamed FLOW-MATIC when Remington Rand released it to customers in 19…
Page 88
COBOL (Common Business-Oriented Language), which was developed in 1959 by a committee representing the U.S. government and the com- puter industry. It quickly became the standard for commercial data pro- cessing—partly because COBOL programs could run on any type of com- puter and partly because the U.S. Department of Defense (which wanted the ease of using a single language) insisted that all of its computer sup- pliers provide suppor…
Page 89
Women were unusually prominent in this movement, given their small percentage of the computing workforce.
Page 89
Snyder had created an important type of subroutine called a sort-merge genera- tor, which allowed the programmer to specify a set of input files and the type of sort and merge operations to perform and then generated machine code for performing the specified ac…
Page 89
Other female innovators included Adele Mildred Koss, also at Reming- ton Rand, who created the first editing generator. This produced code to format data for output to tape or printers, reducing the time that was needed for this task from weeks to minutes.
Page 89
Would programmers be deskilled, perhaps even replaced by their own users? Or would they be empowered? Gender was also part of this contested identity, since certain roles in the computing workplace were tied to the sex of the worker.
Page 90
Pen
Page 91
Pen
Page 91
Hopper’s vision had evolved by the time she spoke at the May 1954 Office of Navel Research Symposium on Automatic Programming for Digital Computers, an event that she helped organize. Here she decisively rejected the notion that the programmer would be replaced by a machine.
Page 92
36Such rhetoric portrayed programmers as the masters, not the victims, of automation, and the female-friendly field of software as every bit the equal of the masculine domain of engineering.
Page 93
…ms.” 41Of all the discourses surrounding automatic programming, these marketing efforts aimed at managers come the closest to invoking a classic labor-management con- flict, with automation leading to deskilling or unemployment of formerly skilled workers.
Page 94
Neither p nents nor opponents were primarily concerned with deskilling. In fact, some managers worried that compilers would require programmers to become more skilled. 46Rather than a serious goal, deskilling functioned more as a handy cultural trope that marketers could use to appeal to potential customers.
Page 94
Software production in the 1960s was challenging on many fronts. Pro- grammers struggled to make full use of a steady stream of hardware im- provements, such as a tenfold increase in memory size and speed between 1960 and 1965. 55The bulk of user applications shifted from the elegant simplicity of mathematical equations to the real-world messiness of busi- ness processes. Computer manufacturers were trying to develop a new type of software, operating systems, which were forbiddingly complex since they had to provide an interface between the computer and all of the other programs that were running on it. Larger projects often meant larger programming groups, creating coordination and communication issues on top of the technical ch…
Page 98
In the sections that follow, I argue that the word “engineering” should be seen a metaphor, not a simple description, and as only one of many possible models for programming.
Page 107
The computing literature of the 1960s is rife with competing metaphors, each of which was chosen to make a particular claim about the nature of programming. Some managers continued to view programming as a creative art or craft—and in a positive sense, rather than as a problem to be cured by imposing scientific rationality. 99In the academic community, computer science was often seen as a branch of mathematics, which had a higher intellectual status than engineering.
Page 107
Engineering had multiple connotations that could serve diverse agen- das. For people who were directly involved with software production, it could mean having a set of clearly specified, standard techniques that produced measurable results. An important early technique that was as- sociated with this idea of software engineering was structured program- ming, a concept articulated by Edsger W. Dijk…
Page 108
Douglas McIlroy of Bell Labs complained at Garmisch that the image of programming as an art or craft gave it a lower status than engineering: “We undoubtedly get the short end of the stick in confrontations with hardware people because they are the industrialists and we are the crofters.”
Page 109
120For academic researchers, the word engineering had a convenient ambiguity that could encompass both theoretical ab- stractions and practical tools.
Page 111
Women had not only helped create the compilers and high-level languages that provided a key tool for producing better code, they had also devoted their professional lives to me- diating between manufacturers, programmers, and users, a key goal of the conference. Their exclusion suggests that the new paradigm of program- ming as engineering brought with it unspoken ideas about which gender could best elevate the practice and status of pro…
Page 112
In this cultural context, the engineering metaphor could not be gender-neutral. Indeed, given the greater status assigned to masculine pursuits, it is conceivable that by choosing the engineering la- bel—and omitting women like Hopper—the Garmisch organizers and their followers were distancing themselves, consciously or not, from an earlier era when programming had been seen as “women’s work…
Page 112
Given this gendered status hierarchy, some practitioners—women as well as men—may have believed that programmers would rise in stature by adopting the title of engineer. An unintended consequence of this move may have been to make programming and computer science less inviting to women, helping to explain the historical puzzle of why women took a leading role in the first wave of software improvements but become much less visible in the software engineering era.
Page 113
Before the mid-1960s, there was no such thing as a software industry sell- ing standard, off-the-shelf applications. Instead, organizations that used computers had three main sources of software—their own programming staffs, outside contractors, and free software provided by computer man- ufac…
Page 127
But many companies, especially smaller businesses, could not afford a dedicated programming staff or simply preferred to outsource their software development. To meet this demand, a software contracting
Page 127
industry—also known as custom programming or programming servic- es—arose in the mid-1950s. These new companies contracted with gov- ernment or industry clients to produce one-of-a-kind programs that were tailored to the client’s specific needs.
Page 128
Elsie Shutt and Stephanie Shirley were early entrants into the industry. Shutt appears to have been the first woman to start a software business in the United States, and Shirley was probably the second in England, following Dina Vaughan.
Page 129
Shirley and Shutt did more than succeed as female entrepreneurs in a man’s world; they also radically rethought what a business could be. They challenged the business culture that refused to take mothers or part-time workers seriously, and they even questioned the fundamental purpose of running a commercial enterprise. The fact that these two women devel- oped such similar visions—even though they worked in different coun- tries and were unknown to each other until many years later—speaks to the common problems that were faced by women who tried to maintain professional careers as…
Page 136
The term computer science is attributed to Louis Fein, a consultant who wrote a widely read 1959 re- port on the state of computing education in U.S. universities.
Page 155
I mediately after World War II, the schools that built the first computers— the University of Pennsylvania, Harvard University, the Massachusetts Institute of Technology, and Cambridge University—began offering semi- nars and hands-on classes to share knowledge about the new technology with researchers and advanced students. 14With financial support from government and industry, other universities adopted the computer as a research tool and then recruited computing experts (or potential experts) to help their faculty use the machine. 15Theory raced to catch up with practice, and computer science as a discipline emerged as an afterthou…
Page 155
The first computer science departments arose in parallel with a series of feminist milestones, such as Betty Friedan’s The Feminine Mystique (1963), the U.S. Equal Pay Act (1963), Title VII of the U.S. Civil Rights Act (1964), the founding of the National Organization of Women (NOW) (1966), the UK Equal Pay Act (1970), the beginning of Ms. magazine (1971), the passage of the Equal Rights Amendment by both houses of the U.S. Congress (1972) (although the amendment failed to be ratified by the states), and the UK Sex Discrimination Act (1975).
Page 159
Feminism became a resource that was available to women who wanted to transform and not just survive the culture of academic computing.
Page 159