Beatrice Helen Worsley: Canada's Female Computer Pioneer

Beatrice Helen Worsley: Canada’s Female Computer Pioneer

Scott M. Campbell University of Toronto When Beatrice Worsley died in 1972, Canada lost a computer pioneer and a witness to several great moments in computing history. This biography aims to provide insight into Worsley’s obscure, but remarkable and all too short, career.

In the past decade, several authors have begun Oaxaca, which her grandparents had estab- to explore the history of computing in Canada. lished in the 1850s. Around 1917, the mill was A special 1994 issue (vol. 16, no. 2, April–June) destroyed by rebels, and the family moved to of the IEEE Annals of the History of Computing El Salto where Joel rose to general manager of was dedicated to the theme, and John the Rio Grande group’s CIMSA mills. Charles Vardalas’s book, The Computer Revolution in and Beatrice were home schooled by their Canada, (MIT Press, 2001) examined the devel- mother, and the children spent many years iso- opment of a Canadian computer industry. lated and physically cut off from the surround- When discussions turn to early developments ing community for general safety reasons. They at the University of Toronto, one name that fre- learned Spanish, although not fluently. quently appears is Beatrice Worsley. Although many other computer pioneers still recall her Early days quiet personality, complete accounts of her In 1929, Joel and his wife left Mexico for career and life are more disjoint. Toronto, Canada, primarily for their children’s In fact, her experiences and accomplishments education. In Toronto, Charles attended Upper were remarkable. During her short career before a Canada College, one of the oldest and more fatal heart attack at age 50 in 1972, Worsley built prestigious private schools in the country. a differential analyzer; was a member of the Beatrice, aged 7, attended Brown Public School Cambridge, UK, Mathematical Laboratory, during until 1934, then attended the private Bishop the first EDSAC run; worked with well-known fig- Strachan School until 1939. Both Worsley chil- ures such as Douglas Hartree and Alan Turing for dren tested well enough to be put in advanced her PhD; published 17 technical papers; cowrote classes.2 a compiler for the Ferranti Mark I; and taught var- The Bishop Strachan School was founded in ious computer courses for 20 years at the 1867, under the Anglican Church of England, as University of Toronto and Queen’s University in an alternative to the many Catholic orphanage Kingston, Ontario. and finishing schools in Toronto. Originally, Worsley was a pioneering computer scientist students were typically daughters or relatives of and the first female in Canada to make signifi- clergy, or from the middle-class families of the cant contributions to the field. This biography region. In the 1920s, however, tuition was will not fully explore the role of gender in the increased, and the school began catering to the history of computing, but as we shall see, there well-to-do. When Worsley was a student in the is little doubt that Worsley’s gender had a sig- 1930s, the school’s reputation was academics nificant effect on her career.1 and religion and featured a heavy dose of Beatrice Helen Worsley was born in extracurricular sports, drama, and music. Most Queretaro, Mexico, on 18 October 1921, the teachers were British, hired specifically for a pre- second child of Joel and Beatrice Marie ferred English teaching style.3 Worsley. Their first was Charles Robert Worsley, Two educational tracks were available. The a few years older, born in Atemajac. Joel first emphasized religious and domestic train- Worsley had been born in Ashton-Under-Lynn, ing; the second offered a university track pro- Manchester, UK, in 1887 to a working-class gram including algebra, geometry, chemistry, family. In 1908, he moved to Mexico to work and physics. Worsley excelled in the university at his wife’s family’s textile mill in Xia near track, despite the fact that she was the youngest

in a class of about 30 students. The head- possible in the Women’s Royal Canadian Naval mistress during her stay, a Miss E.M. Lowe, Service (WRCNS), also known as the Wrens. noted she was one of the most brilliant pupils The decision to serve was not unusual, and her ever at the school.4 Worsley finished middle brother, who graduated that year with a bache- school in 1937, with awards in math, physics, lor of architecture, joined the Royal Canadian divinity, English, and general proficiency. Engineers at the same time. In 1938, she completed junior matriculation, and in 1939 graduated, with honors, senior Military service: Wrens matriculation; with awards in math, science, An impression of her personality emerges and proficiency, and the Governor General’s from the pages of her military application. In a Award for the highest overall grade.5 Worsley letter to the WRCNS Recruiting Offices in enrolled in the honors course in science at February 1944, seeking an opening in the Trinity College, University of Toronto, in organization, Worsley expressed a preference September 1939. Her marks earned her a gener- for laboratory work or research duties.9 More al proficiency entrance award and the Burnside details are found in the notes from the recruit- Scholarship in Science from Trinity. She had the ment interview later that month. For example, number-one rank in most classes that year and Worsley expressed a fondness for music and was awarded the first Alexander T. Fulton piano playing, and an interest in photography. Scholarship in Science. In October 1940, she And as the interviewer noted and her refer- transferred to the Mathematics and Physics ences agreed, she had a quiet, pleasant, and division, specializing in applied mathematics. composed personality. Worsley continued to excel academically Probationary Sub-Lieutenant Beatrice Wors- and to impress the lecturers in courses such as ley started service on 5 April 1944 at the HMCS pure and applied mathematics, classical Conestoga base in Galt (now Cambridge), physics, thermodynamics, optics, quantum Ontario. Galt was the basic training center for theory, relativity, and electricity. Because her Wrens from across Canada, and it included college years coincided with World War II, her about four weeks of physical training, drills, studies included German and ballistics. She fin- and lectures on naval customs. For Worsley, ished First Class each year and won the James this was followed by another four weeks of offi- Scott Scholarship in Mathematics and Physics cer training. After several months of unspeci- in her third year. In 1944, she graduated with a fied general duty, she was commissioned and bachelor of arts in mathematics and physics. directed to the Special Branch and on 9 Sep- It’s possible that Worsley’s interest in com- tember transferred to the Naval Research Estab- puting machines was first sparked during these lishment (NRE) at Her Majesty’s Canadian Ship formative years. Between June and August of (HMCS) Stadacona in Halifax, Nova Scotia, for 1942, she had worked with mechanical calcula- harbor defense research. tors in the Actuarial Department of Manufac- Although the NRE was a recent creation of turers Life Insurance Company in Toronto. The the Royal Canadian Navy (RCN), the origins clerical position would have had little in com- went back to 1940, when the RCN accepted mon with numerical analysis, but the early responsibility for degaussing ships, sailing from exposure could have caught her attention in Canada to Europe, to reduce their magnetic sig- some way.6 Regardless, by graduation the foun- nature and vulnerability to German magnetic dation of Worsley’s career was set as an out- mines. Although in principle a straightforward standing student, with obvious aptitudes in operation, the research, testing, and fine-tun- mathematics and science. Professors later ing of existing and experimental techniques recalled she was one of their best students: gift- fell to a small group of about 50 officers, scien- ed, dependable, conscientious, and painstaking tists, and support staff stationed in Halifax. At in all her work.7 the peak of this activity, Worsley was one of six But what of Worsley’s personality and ambi- Wrens at the NRE. The research and testing of tions? No records at Bishop Strachan exist that degaussing techniques was the beginning of suggest she participated in team sports or scientific naval research in Canada, and the drama. Later, in the 1944 Torontonensis year- NRE’s work was crucial to keeping the East book, the only activity she listed after five years Coast harbors clear of enemy mines.10 at the University of Toronto was the Math and By the time Worsley arrived in Halifax in Physics Society. She did, however, leave no 1944, many of the degaussing operations had doubt as to what she foresaw in her future: become routine, and the scientists were reduced “War service, then advanced study.”8 Indeed, to handling new or special cases only. Worsley Worsley enlisted as soon after graduation as was employed on generalities at first, likely

52 IEEE Annals of the History of Computing assisting with the onshore data analysis, proba- the famed MIT Radiation bly not much different than the lab work she Laboratory and the super- had requested when enlisting. Although exper- visor of Worsley’s master’s iments were under way at the NRE to develop thesis.12 countermeasures against acoustic mines and That thesis, A torpedoes, she was not a member of that par- Mathematical Survey of ticular group. Computing Devices with an With the end of hostilities in 1945, the imme- Appendix on Error Analysis diate relevance of the NRE research declined, and of Differential Analyzers, by early 1946, most of the staff had left, typical- provides a fascinating ly taking advantage of educational opportunities snapshot of contemporary available to veterans. Worsley was one of just a computing technology. handful who chose to stay on, and the only With the eye of a mathe- Wren to do so. In Septem-ber 1945, she was matician, Worsley had promoted to lieutenant at the same time as she combed the literature and was reassigned to a new project: hull corrosion compared the features and research. Hull corrosion is a nontrivial problem characteristics of many that affected metal-hulled ships at the time and continuous and discrete required regular, expensive dry-dock repairs. computing devices com- The underwater electrochemical effects were pleted, under construction, not well understood, but scheduling difficulties or planned. In the discrete during the war had prevented most investiga- category, her tables Figure 1. Worsley at sea on the Quinte tions beyond limited visual inspections. With described the Harvard in midwinter, conducting an the freeing of ships, earnest research began in Mark I and II, several IBM experiment under decidedly chilly late 1945. calculating machines, a conditions. (Photo courtesy of Alva As part of the fundamental research section Bell Labs electromechani- Worsley.) of this project, Worsley performed experiments cal computer, the ENIAC, and consolidated findings on board various test the EDVAC, the Institute ships and the NRE research minesweeper for Advanced Study computer at Princeton HMCS Quinte (see Figure 1). The project University, the Whirlwind I and II at MIT, the demanded more than 150 days at sea—a record UK’s National Physical Laboratory (or Pilot for Wrens’ at-sea service—which were often Automatic Computing Engine) computer, and spent during adverse midwinter weather con- the University of Cambridge EDSAC. The ditions. Worsley’s endurance earned her much appendix provided a practical and theoretical consideration and respect from the crews, as analysis of the sources and areas for error reduc- did her knowledgeable onshore interactions tion in differential analyzers. The research was with hull surveyors and specialists. Known to suggested by Samuel H. Caldwell, of MIT’s elec- all as “Trixie,” she was well able to hold her trical engineering department, who had helped own. Working at what she herself acknowl- Vannevar Bush design recent analyzers. In her edged was a man’s job, by her actions and man- conclusion, Worsley stated that her mathemat- ner Worsley well deserved the recognition she ical results corresponded well with and con- earned.11 In August 1946, she was demobilized. firmed the experiences of analyzer users.13 With her advanced degree in hand, Worsley Graduate studies returned to Canada near the end of summer In September 1946, Worsley entered gradu- 1947. Never one to share her personal ate studies at the Massachusetts Institute of thoughts, she now confided to her family that Technology (MIT) in a one-year master’s pro- her future was in computers. She was quickly gram of mathematics and physics. This would becoming an expert on the new technology. It’s be the preliminary launching point for her unlikely she was exposed directly to the ongo- computing career. The challenges she met at the ing Whirlwind research at MIT during her short NRE may have prepared her for research, but stay in Boston, but the proximity was enough: MIT provided the first exposure to computers. The seed had been planted. Her coursework included solid-state physics Unfortunately, although MIT was a leader in taught by Laszlo Tisza; a course based on John computing technology, Canada had yet to L. Synge’s text, Principles in Mechanics (McGraw- leave the starting blocks. With no immediate Hill, 1942); an applied electronics course on employment opportunities in the computer feedback amplifier design; and a function the- field, Worsley found work in Ottawa at the ory course with Henry Wallman, a member of National Research Council (NRC) of Canada.

October–December 2003 53 Beatrice Worsley: Female Computer Pioneer

After only a few months as an aerodynamics more ambitious electronic engineering research research officer in the mechanical engineering project: the design and construction of division, she resigned. With the approval of her Canada’s first computer, the University of director, she had negotiated a position at the Toronto Electronic Computer Mark I (UTEC). It new, University of Toronto Computation was a parallel-processing machine designed by Centre, and by mid-January 1948 joined the a team of professors and graduate students. Centre as a project assistant. This was no toy built of spare parts but a prototype worth tens of thousands of dollars to the Computation Centre at Toronto Canadian Defense Research Board (DRB) and to Toronto was among the few universities in the NRC, which were sharing the development North America to have a computer research and costs. The project represented an attempt to development program in place at the end of the obtain a single large computing resource to be 1940s, and the only one in Canada for a number shared nationally among government, military, of years thereafter.14 Preliminary discussions and university research groups. The UTEC proj- regarding the Centre had taken place among fac- ect ran for a number of years as parts were ulty seeking support from government research designed and built locally by hardware special- agencies throughout 1946 and 1947 that culmi- ists at the Computation Centre, who operated nated with a relatively small grant from the NRC independently of the analog and mechanical in September 1947. Funds were provided for two computing side of the Centre at the time. IBM punch card calculators and salaries for two By October 1951, the UTEC was nominally project assistants: Worsley and J. Perham Stanley, operational but plagued by tube failure and a graduate student and former colleague of mechanical reliability issues. Unfortunately for Worsley through the Math and Physics Society.15 the team, Wilfrid B. Lewis of AEC had grown Day-to-day management of the Computa- impatient for a dependable, full-scale computation Centre was handled by Calvin C. “Kelly” tional machine for Canada. Instead of finishing Gotlieb, a physics PhD graduate from the Uni- the UTEC, the influential Lewis urged that the versity of Toronto a few years earlier and a mem- DRB and NRC money would be better spent by ber of the Physics Department. After training on purchasing a complete system, specifically a the new tabulators, Gotlieb, Worsley, and Stan- Ferranti Mark I. (He knew that the UK govern- ley tabulated a function for Atomic Energy ment had canceled an order earlier that year, Canada (AEC) at Chalk River, Ontario. This and leaving Ferranti without a buyer.) A decision was similar computing problems simultaneously forced quickly, against the wishes of the hard- raised the profile of the fledgling center and ware development team, and the University of helped stimulate interest in relay and electron- Toronto ordered the Mark I early in 1952. ic computing research in Canada.16 Worsley had little to do with the UTEC pro- Over six weeks during summer 1948, totype development. Her career always leaned Worsley constructed a differential analyzer toward software rather than hardware, and the using Meccano (a metal construction system UTEC was built during a period when she was designed for building models), based on absent from the Centre because she was on Douglas Hartree and Arthur Porter’s 1935 arti- assignment in the UK. However, she did play a cle.17 Constructed from about CAD$75 worth unique role when the Ferranti Mark I arrived, by of Meccano, the analyzer was minimally mod- naming it Ferut. (Ferut is an initialism standing ified from the original design but offered slight for “Ferranti computer at the University of improvements to the electrical power distribu- Toronto.”) Ferut was installed and gradually tion system, the design of the torque ampli- made operational over the spring and summer fiers, and the output pen support.18 of 1952, and it quickly replaced the UTEC.21 Unfortunately, there is no information regarding what use, if any, the analyzer was put to or EDSAC and Cambridge why Worsley built it. Possibly the Centre was Meanwhile, after finishing the Meccano looking for an inexpensive computational project in 1948, Worsley and Stanley had been device in addition to the punch card machin- sent overseas to learn what they could of the ery; if constructed properly, a Meccano analyz- EDSAC at the Cambridge Mathematical er could be put to serious use as Porter explains Laboratory. (It was not uncommon for in a recent article.19 Work on a second analyz- researchers to join the lab for a time to help er, or perhaps the original model, was restarted build and test the computer.) The two project by another student in 1951 at the behest of assistants arrived to find the EDSAC in a fairly Gotlieb for teaching purposes.20 advanced state of construction, and although Under way at the same time was a much neither had an engineering background, both

54 IEEE Annals of the History of Computing helped prepare it for the first run on 6 May 1949.22 The EDSAC (see Figure 2) was working just in time for a conference on high-speed computing held at Cambridge that June. On the first day, it was demonstrated to the atten- dees by printing tables of squares and primes. Worsley prepared a report that included a reproduction of the results and a description of the observable physical events and order code routines. This report went into the conference proceedings and subsequently was reprinted in the famed resource for historians of computing, Randall’s The Origins of Computers.23 There was no particular glory to this task, and it must Figure 2. A posed photograph (Worsley is on the have seemed minor at the time, but the report right) in front of the EDSAC in Cambridge, circa earned Worsley a place in history. 1949–1951. (Photo courtesy of Alva Worsley.) Stanley earned his historical footnote soon after the EDSAC was running when he cowrote fessor of mathematics at the University of a routine with Maurice Wilkes to tabulate the Toronto and a member of the Computation reciprocal of the complex Gamma Function.24 Centre. Worsley wrapped up the writing quick- The extensive tables were published the next ly and was rehired as a staff mathematician at year after he returned to Toronto to complete the Centre in July 1951. Hartree approved the his PhD.25 The routine possibly represents the dissertation in Cambridge, and Worsley’s doc- first time a machine produced values never torate was awarded in 1952 after Hartree made before computed by any other means. The a visit to Toronto.29 event was made historically unique when John von Neumann surprised the lab by visiting Worsley’s dissertation: A review unannounced while the routine was running.26 Worsley’s dissertation, Serial Programming for Worsley did not return to Canada until 1951 Real and Idealized Digital Calculating Machines, because she had resumed her graduate studies is believed to be the first PhD dissertation in mathematical physics at Newnham College, involving modern computers.30,31 It was divid- University of Cambridge. While helping in the ed into four logical sections, each tackling an lab, she attended or at least audited a number academically interesting aspect of digital com- of courses and lectures. These included several puting, and offers some early examples of rou- on quantum theory, taught by a collection of tines used to solve nontrivial scientific well-known figures: Paul Dirac, Sir Lennard- problems numerically. Jones, and Nicholas Kemmer. Perhaps more sig- The first section outlined and discussed nificant to her career were courses she took on Class I machines, which were defined as digital, summable series and number theory, with automatic, serial, and universal. These criteria Albert Ingham, and on numerical analysis, led were then used in the next few sections of the by Douglas Hartree.27 dissertation to explore three specific machines: Hartree was an important figure at the EDSAC, the Manchester Automatic Digital Cambridge with great interest in the EDSAC’s Machine (MADM), and the UTEC. She was potential in solving numerical problems. Many clearly familiar with all three and knowledge- have recalled Hartree’s helpfulness and gen- able of various idiosyncratic programming erosity toward others; he certainly collaborat- practices, but because she included only EDSAC ed with an unusual number of women for the routines in latter sections, the full extent of her time, and supervised at least one other exposure to the MADM is unknown. It is also Canadian, Charlotte Fischer, before his death likely that she added the UTEC section after she in 1958.28 He likely would have been impressed returned to Toronto. that Worsley constructed her own Meccano In the second section of her dissertation, analyzer, and perhaps this is why he chose to “Idealized Machines,” Worsley discussed the supervise her dissertation. potential of universal Turing machines and Worsley started writing her dissertation at what sorts of problems could properly be han- Cambridge, but in 1951 for some unknown rea- dled. She treated the subject formally and made son she returned to Canada with the text clear the distinctive limitations and differences unfinished. She arranged to complete the work between abstract and physical computers. with the supervision of Byron A. Griffith, a pro- For the third section, Worsley wrote her

October–December 2003 55 Beatrice Worsley: Female Computer Pioneer

own EDSAC routines to solve three scientific Routines could also be developed and placed in problems numerically. The first problem was a libraries that any programmers could reuse, determination of the second-order correction regardless of the machine they were using. to be applied to the value of gravity from pen- Worsley broke down potential Class I com- dulum measurements at sea. The problem was puter usage into four scenarios or applications: considered well suited for the EDSAC because logical and truth calculations, applied mathe- of the magnitude and requirement of many matical and engineering problems, pure math- solutions of similar basic forms, and she cred- ematical problems, and payroll or statistical its an R.I.B. Cooper for proposing the research. problems. Each was evaluated in terms of order This chapter was later published separately as code complexity to determine the minimum her first scientific article.32 number of instructions per case. Her analysis The second problem was an automatic ren- suggested that although the scenarios over- dition of the Hartree self-consistent field lapped, a substantially different set of instruc- method for the solution of atomic wave equations was necessary in each case. tions with exchange. This was suggested by This invalidated the premise of a universal Hartree himself and was of course based on the machine code, which she acknowledged. technique that he had developed for desk cal- Although highly specialized computers could culators. The third question was selected to be built, general-purpose models represented demonstrate a way in which automatic com- the foreseeable future, given the cost and putations could be used to explore the merits unique nature of existing machines. of a method together with variations. Worsley Consequently, computers were destined to concluded that a Lieberman one-dimensional occupy a middle ground between the two process compared unfavorably with step-by- extremes of scientific computation and data step and adjustment methods. processing. Furthermore, machine and order In the fourth and final section of her disser- code design could and would result in infinite tation, Worsley’s experience and command of variations and not an optimal, universal solu- computing science and technology is unam- tion. Worsley returned to this concept later in biguous. On the basis of the first three descrip- her career, and she held a long interest in com- tive, theoretical, and applied sections, Worsley puter libraries and subroutines. presents an empirical and systematic theory of “axiomatic programming,” which could “guide Unsolved mysteries the design of order codes and libraries, based By the time Worsley finished her assign- on the proposal of an optimum basic automat- ment at Cambridge, she was one of the most ic universal digital machine.”33 Although her computer-literate women in the world, with theory was partially derived from ideas appar- practical and theoretical expertise that few ently presented by Turing at the 1950 could have matched. She was one of the first Cambridge summer school on program design female academic computer programmers who of automatic digital computing machines, wrote all her own programs, a point she strong- Worsley did not specify his exact contribution. ly emphasized in her dissertation. Not only did Possibly, she was elaborating on Turing’s pref- her time at Cambridge prove that real scientif- erence for simplicity of hardware instructions.34 ic work could be accomplished with computers The general problem that Worsley present- (as opposed to Meccano), it undoubtedly solid- ed was not new. As she explained in her disser- ified her career choice. tation, Claude Shannon provided a guaranteed However, several mysteries remain from this way to set up relays to solve logic problems, but period. First, it is unknown why she returned to faster or more intuitive arrangements were Canada in 1951 without having completed her often possible, and work was under way to dissertation. At the time, PhD students regis- minimize or otherwise economize the number tered in Cambridge for three years and then had of relays.35 This notion could be transferred to two years to finish writing; it was not unusual electronic computers, but instead of optimiz- to do this at another location.36 But why did she ing relays, the problem was to determine which leave Cambridge? There were no family-related hardware instructions were fundamental and obligations or known romantic attachments in which were supplementary. Her intention was Toronto encouraging her to return. to narrow down a universal set of fundamental Moreover, the Computation Centre in order codes to be used by all Class I machines. Toronto did not rehire her right away, although Designers and engineers could then selectively her knowledge of UTEC suggests she visited the optimize the execution of the machine code in facility regularly, so her return was probably question to provide better overall throughput. not related to employment. Money was not

56 IEEE Annals of the History of Computing likely the issue, because in May 1950 she was awarded a CAD$1,500 Senior Fellowship by the Canadian Federation of University Women. With the Korean War under way, the RCN requested she consider service, but she turned down the opportunity. Unfortunately, because of her quiet personality and the decades that have since passed, the answer may never be known. No one from the Cambridge lab seems to recall Worsley much beyond that she was a member for awhile.37 What makes this issue intriguing is an event that occurred 20 years later. After her Figure 3. Beatrice Worsley sitting at the Ferut in death in 1972, Worsley’s entire estate was the University of Toronto’s Computation Centre, bequeathed to the University of Cambridge. A circa 1952–1958. (Photo courtesy of Alva Worsley.) total of £75,000 would form the Lundgren Fund, in honor of one Helge Lundgren.38 The Lundgren Research Award is presented to PhD the proper methods of programming, adapting students, not ordinarily residents of the UK, techniques for their installation along the way. who have completed at least four terms and are In 1953, they arranged two more formal cours- engaged in research in a scientific subject, es for a wide variety of computer novices, including mathematics. All things being equal, including actuaries, scientists, and gradate stu- preference is given to candidates “working in dents. The lessons contained both theoretical the Computer Laboratory, or whose research and practical components devoted to machine has been interrupted by national service or per- design, computer logic, and numerical tech- sonal misfortune.”39 The identity of Helge niques. The attrition rate seems high, however, Lundgren remains unknown, as does the moti- because just eight of 30 students completed vation for the large donation. one course.40 Few could master the Ferut; in the words of one user, it was appallingly difficult to Ferut and Transcode program.37 As the Ferut was installed at the Toronto In the fall of 1953, inspired by John Computation Centre over the spring and sum- Backus’s Speedcode system for the IBM 701, a mer of 1952, Worsley soon found herself oper- solution to the Ferut’s programming difficul- ating the console (see Figure 3). Gotlieb was ty was considered. Speedcode effectively sim- sent to Manchester to learn what he could from ulated the existence of a reasonably easy to the Ferranti Mark I installed there, but Worsley program computer; software ran slower, but would have been somewhat familiar with it programming and debugging times were because it was a copy of the Manchester com- reduced significantly. If something similar puter described in her dissertation. While the could be done with the Ferut, it would be engineers were getting the Ferut up and run- valuable to the Centre. As the only comput- ning, the Centre staff was faced with the prob- ing facility in Canada, simplifying the pro- lem of software. Fortunately, because of the gramming cycle for remote users was an contact with Manchester, they had access to important goal. Worsley and J.N. Patterson the original system software and libraries, and Hume, an assistant professor from the Physics staff members also picked the brains of atten- Department, were assigned to create an auto- dees of the September 1952 Association of matic coding system for the Ferut. They both Computing Machinery (ACM) conference in had considerable experience with the Toronto. Once operational, the Ferut was put machine, having authored numerous operat- to use on both scientific research and practical ing system modifications and subroutines. computational problems, from atomic energy They dubbed their project Transcode and fin- equations to St. Lawrence Seaway calculations ished writing the compiler within about a and insurance tables. year, taking advantage of every nuance in the The Ferut was a tricky computer to learn to hardware.41 program, and the staff devised improved input Transcode was an immediate success. Basic and output routines, debugging tools, and lessons could be taught in two hours, and the other techniques for their novice programmers. calculations could be returned to users in a The first year Ferut was on the campus, Gotlieb matter of days, not weeks, although Ferut was and Worsley held a small informal course on considered slow at the time. Articles by Worsley

October–December 2003 57 Beatrice Worsley: Female Computer Pioneer

and Hume appeared in both the Journal of the Detailed records and notes from her courses ACM and Physics in Canada announcing are available in the archives at Queen’s Transcode and describing the new and effective University. One of the earliest was a fourth-year tool for physicists and scientists.42 By 1958, industrial engineering course on programming when the aging Ferut was replaced by an IBM the IBM 650 and a senior physics course called 650, hundreds of faculty and students had been Programming Digital Computers. The lessons taught Transcode, and the Centre had put it to included theoretical and practical exposure to scientific use on behalf of dozens of research numerical analysis and related programming groups across Canada. challenges. Worsley had an excellent reputa- Because of the code’s simplicity, nearly bug- tion as a sharp and knowledgeable lecturer. As free programs could be submitted and the discipline of computer science began to sta- returned by mail. A direct teletype hookup bilize in the 1960s, the computer-related con- between Toronto and various Canadian uni- tent of Worsley’s courses began trickling versities was added later to provide a primitive downward from graduate levels to third- and remote operation.43 second-year studies. When the 650 was replaced by Canada’s first IBM 7090 in 1962 Articles and courses: A paper trail and made available to students, she modified Worsley’s focus and skill lay in adapting sci- her courses to accommodate newer languages entific problems to be solved using Transcode and techniques.49 All the while, Worsley main- and the Ferut. This subject area is the focus of tained an affiliation with the Physics most of the 17 technical papers she published Department, running first-year tutorial sections between 1952 and 1964. Her first such article for a number of years.50 described results obtained using Transcode to Worsley increased her professional activity solve nonlinear second-order differential equa- year by year as opportunities for women tions using a Runge-Kutta method, and custom improved. She had been an early member of floating-point routines that she developed.44 the ACM and in the 1960s joined special-inter- Most of the other articles flowed directly out of est groups on university computing and infor- her Cambridge work on self-consistent field mation retrieval. Worsley also served as the calculations. Toronto region correspondent for the Quarterly In September 1955, the Pure Physics Bulletin of the Computing and Data Processing Division of the NRC and the Computer Centre Society of Canada from 1962 to 1965, as the launched a joint project to develop the Hartree- director of the national executive in 1968 (now Fock formulations for digital computers and cal- renamed the Canadian Information Processing culate specific atomic wave functions. Working Society), and technical editor for the Bulletin in conjunction with J.F. Hart of the NRC, who from 1970 to 1971. coauthored several of the papers, Worsley coded One of Worsley’s long-term interests was the routines for the Ferut.45 Hartree consulted subroutine libraries, which she had first written on the project until his death in 1958, and the about in conjunction with universal instruc- general scheme Worsley used can be found in tion sets in her dissertation. Throughout the his 1957 text The Calculation of Atomic 1950s, part of her job as staff mathematician at Structures.46 The work led to a series of short arti- the Computation Centre was developing and cles on the wave functions of several elements.47 maintaining a variety of low- and high-level In the early 1960s, she coauthored a number of subroutines for the Ferut and Transcode. The biophysical articles that explored numerical lessons from these years were sketched out the- computational techniques available for such oretically in a 1959 article, “Blueprint for a research.48 Although most of the staff at the Library.” As she saw it, the university comput- Centre had advanced scientific degrees, if meas- ing center library requirements were twofold: ured by the number of publications credited, “a sophisticated coding scheme for staff Worsley was by far the most productive mem- engaged in research projects and a simplified ber of the Computation Centre in the 1960s. coding scheme for teaching purposes and stu- By 1960, Worsley’s career was heading away dent applications.”51 Worsley wrote that the from research toward teaching. She began coding scheme should be as flexible as possible teaching noncredit university extension cours- yet provide complex mathematical functions es in the mid-1950s. They became graduate and and operations. It should also provide direct undergraduate courses and the load increased machine addressing if necessary. Worsley when the IBM 650 was installed in 1958 and described the second scheme as a subset of the the Institute of Computer Science was founded first and available to all programmers. It should at the University of Toronto in 1962. provide double- or multiple-length arithmetic,

58 IEEE Annals of the History of Computing floating-point routines, complex and matrix As a result, in 1965 Worsley left Toronto to algebra, and statistical functions. Most impor- take a new position at the Computing Centre at tantly, the schemes should be universal and Queen’s University in Kingston, Ontario. The machine independent, within the bounds of departure caught her Toronto colleagues off the topology of any given computing center. guard, and it would not have been an easy deci- In some ways, the “Blueprint” article was an sion for her. At Queen’s she would no longer be a updated realization of axiomatic programming. professor, but computing advisor to the Com- But now Worsley’s intention was to abstract puting Centre with teaching duties. Compared above the machine’s instruction set, creating in to the well-funded, well-equipped, and commit- essence a virtual machine. It made possible ted research and teaching program in Toronto, optimized libraries that every programmer Queen’s was a considerable step down. There was could use and benefit from, regardless of no computer science program, and the only whether one needed procedures for data pro- computer in evidence was an outdated IBM 1620 cessing or for numerical analysis functions. in the Civil Engineering Department. A secondary interest of Worsley’s was library On the other hand, Worsley was lured to automation, which developed through exter- Queen’s specifically to help launch and man- nal circumstances. In 1963, the Ontario New age a new Computer Centre with the 1620. University Library Project staff began the task Besides herself, just two people were involved: of acquiring and cataloging more than 35,000 John Lindsay, with a master’s degree from the new books within three to four years. One of University of Toronto and several years’ indus- the participants, the University of Toronto try experience, and Mers Kutt, who had been Library (UTL), turned to the Institute of recruited earlier by Queen’s to get computing Computer Science for assistance.52 Worsley did off the ground.55 For Worsley, it was a chance not help immediately, but did work at the sub- to start afresh, and perhaps earn respect the committee level on the Serials Systems Design way she had in Halifax: by her merits, regard- project, tying automated serials processing with less of gender. Unfortunately, her career, which acquisition, circulation, and catalogue genera- had already veered from scientific research, tion.53 In 1965, the NRC awarded her Grant quickly narrowed at Queen’s to that of teacher Number A-2654 for “An Information Retrieval and administrator. System for Computer Science Literature.” The one-year project was to implement a computer Queen’s University: Many hats system for retrieving references and abstracts A 1967 letter to Worsley from Kutt, the direc- from a computer science literature database. tor of computing, outlined her teaching respon- Worsley’s proposal was based on an earlier sibilities at Queen’s, which besides lecturing paper by Guy J. Groen, a colleague at the included serving as curriculum advisor and Arts Institute of Computer Science at Toronto.54 The and Science Faculty representative. Worsley was grant was to cover the costs of developing a also tasked with providing material for the uni- small prototype using a sample set of data. versity calendar, self-education, and drafting However, Worsley’s participation on both reviews for Communications of the ACM and library projects was limited, because she left Applied Mechanics Reviews. In her role as com- Toronto permanently before the summer was puter science advisor to the computing center, out to join Queen’s University. she was to select book and journals for the small The reason for Worsley’s departure might be library, coordinate the programming staff, act as guessed. Although she had a PhD from Queen’s computing advisor to local area high- Cambridge, was a core member of the school teachers, arrange seminars, and author- Computation Centre teaching staff with more ize computer time. Beyond that, she would be publications than anyone else, and had been the local SHARE user group representative.56 performing valuable research for the universi- Immediately after arriving at Queen’s in ty for over a decade, it was not until 1960 that 1965, Worsley began teaching undergraduate Worsley was promoted from Computation computing courses, which continued until her Centre mathematician to assistant professor of first sabbatical in 1971. These courses were physics. Eventually, when the University of heavily grounded in numerical analysis, cov- Toronto created a graduate department of com- ered programming techniques using Fortran, puter science in 1964, she was promoted to Watfor, and assembly, and included material associate professor of physics and computer sci- on the mathematical principles of computing. ence. In comparison to other staff members, Many of her records, notes, and lecture outlines the lack of official recognition is conspicuous from this period survive in the Queen’s and is almost certainly because of her gender. University archives.

October–December 2003 59 Beatrice Worsley: Female Computer Pioneer

Much of Worsley’s day-to-day work also Conclusion dealt with the operation and long-range plan- The parallel career of Grace Murray Hopper ning for the Computing Centre. Like many is sure to cross the mind of many readers. other Canadian universities, Queen’s had an Hopper and Worsley were apparently acquaint- IBM 1620 installed in the early 1960s. Unlike ed, although casually. Both studied mathemat- most, it was not upgraded to an IBM 7040 or ical physics in graduate school, both enlisted in 7090 mainframe. By the time Kutt, Worsley, the navy during World War II, and both worked and Lindsay arrived at the Computing Centre, on first-generation computers, although neither a 7040 no longer made sense economically, was a hardware specialist. Both were profes- and so they held off upgrading until the IBM sionally active early on, advocating compilers 360/40 became available in 1966. Queen’s was and the value of higher-level languages. one of the first universities in Canada to Worsley’s interest, however, was academic and announce such a prized acquisition, but geo- scientific computing, while Hopper’s was busi- graphical isolation and static university bureau- ness computing and data processing. cratic structures created a computer-related Consequently, Worsley helped write apathy at Queen’s. The field lacked a high pro- Transcode, earning the gratitude of physicists file on campus, and in the early 1960s the across Canada, while Hopper helped design mathematics department signaled they were Cobol, earning accolades from business pro- not interested in computing or applied mathe- grammers across the world. Finally, whereas matics, although this eventually changed with Hopper had an outgoing personality and was a new faculty recruits. During Worsley’s time at colorful spokesperson for the US Navy and for the Computing Centre, most of the impetus for computer companies, Worsley was virtually the computing came from the engineering depart- opposite, working quietly in an academic envi- ment, which sought undergraduate training ronment. In the words of her sister-in-law, “She and programmers experienced with numerical was so very quiet and did not talk about her techniques.57 work too much. I guess she realized that it was Worsley worked hard to improve the situa- beyond our comprehension.”59 tion. Without the larger budgets of other uni- “Trixie” Worsley left a fascinating legacy as versities, computing power was slowly a computing pioneer. Her natural appreciation broadened with microcomputers, with three for what computers were capable of doing was DEC PDP-11s and two PDP-15s and an ancient reflected in a lifelong interest in the develop- IBM 360/20 in place by 1971.58 Lindsay was the ment of computer libraries and scientific com- programming guru, and Worsley was the librar- putation. A skilled mathematician and ian. In addition to maintaining routines and unquestionably Canada’s first female comput- programs, she also wrote most of the computer scientist, she found a successful calling in a er manuals and related documentation. profession dominated by men. Throughout her Queen’s administration implicitly expected life, she struggled against this barrier. The that by gradually increasing the role of comput- strength of her intellect and accomplishments ers, computer science studies would evolve on is undeniable, but official recognition has taken campus. A master’s program at Queen’s was cre- too many years. ated in 1968, together with a new Department of Computing and Information Science. References and notes Worsley was a driving force behind the creation 1. The year before her death, Worsley donated a of an academic home for computing, and she number of her early papers and records, from was again promoted to associate professor. 1946 to 1959, to the Smithsonian Institution. In September 1971, Beatrice Worsley took a These items are available through the National long-deserved sabbatical year at the Depart- Museum of American History. The bulk of the ment of Applied Analysis and Computer Sci- material for this article came from two alternate ence at the University of Waterloo. While at sources. The first is Worsley’s own notes, held at Waterloo, she studied assembler coding as it the Queen’s University Archives. These boxes con- related to computer architecture, which tain copies of her publications including theses, appears to have been an attempt to return to notes from high school to graduate school, pro- her earlier research interests. However, on Mon- fessional correspondence, her teaching notes day, 8 May 1972, while in Waterloo, she suf- from both Toronto and Waterloo, and a great deal fered a fatal heart attack. Survived by her of administrative paperwork from her final years brother Charles, the funeral was held that at Queen’s. The second source is sister-in-law Alva Thursday, 11 May, in Toronto, where she was Worsley, who provided a wealth of personal infor- interred in the Mount Pleasant Cemetery. mation and contributed the photographs.

60 IEEE Annals of the History of Computing 2. A. Worsley, telephone conversation with author, Toronto’s The Globe and Mail, 15 Dec. 1951. 4 May 2003. 21. Readers are again encouraged to consult the 3. K.W. Clarke, “The Story of the Bishop Strachan Vardalas and Williams works for technical and School,” (not published, c. 1950?). political details surrounding the construction of 4. Interview form, 29 Feb. 1944, Lieutenant O- the UTEC and its replacement, the Ferut. 79817 Beatrice Helen Worsley military personnel 22. The original teletype paper output can be found file, Nat’l Archives of Canada. at the Smithsonian Am. History Archives, in the 5. BSS Magazine, Bishop Strachan School yearbook, Beatrice Worsley Collection, signed and dated by Toronto, for the years 1934–1939. Worsley. 6. Nat’l Research Council of Canada Application for 23. B.H. Worsley, The EDSAC Demonstration: Report on Employment, Worsley personnel file. a Conference on High Speed Automatic Calculating 7. I.R. Pounder to M.N. Bedard, Nat’l Research Machines, Cambridge University Mathematical Lab- Council, Worsley personnel file. oratory, January 1950, tech. report, available in B. 8. Torontonensis, Univ. of Toronto yearbook, 1944. Randall, ed., The Origins of Digital Computers, 2nd 9. By 1944 research or lab work would not have ed., Springer-Verlag, 1975, pp. 395-401. been an entirely unusual request; when the 24. M.V. Wilkes, Memoirs of a Computer Pioneer, MIT Canadian Wrens were formed in late 1942, the Press, 1985, p. 147. women were limited to clerical duties, but their 25. J.P. Stanley and M.V. Wilkes, Table of the Reciprocal role in the navy was quickly expanded to include of the Gamma Function for Complex Argument, almost all of the men’s duties except, normally, Computation Centre, University of Toronto, 1950. serving at sea. See “History,” The Wren Associa- 26. M.R. Williams, “UTEC and Ferut…”, pp. 7-8. tion of Toronto, www.naval-museum.mb.ca/ 27. Worsley Archives, box 2, folders 6-10. history/exhib10.htm. 28. C. Froese Fischer, “Reminiscenses at the End of 10. J.R. Longard, Knots, Volts and Decibels: An Informal the Century,” Molecular Physics, vol. 98, no. 16, History of the Naval Research Establishment, 2000, p. 1050. 1940–1967, Defense Research Establishment 29. B.A. Griffith, “My Early Days in Toronto,” IEEE Atlantic, 1993. Annals of the History of Computing, vol. 16, no. 2, 11. J.M. Coade, “Wren Gets Plenty of Time at Sea,” Summer 1994, p. 56. not published, 1946. 30. Ibid., p. 62. 12. B. Worsley fonds, Queen’s Univ. Archives, 31. B.H. Worsley, Serial Programming for Real and Ide- Kingston, Ontario, Canada, (hereafter referred to alized Digital Calculating Machines, doctoral disser- as Worsley Archives) box 1, folders 18-21; box 2, tation, 1951, Worsley Archives, box 2, folder 18. folders 1-4. 32. B.H. Worsley, “On the Second Order Correction 13. B.H. Worsley, A Mathematical Survey of Computing Terms to Values of Gravity Measured at Sea,” Devices with an Appendix on Error Analysis of Differ- Proc. Cambridge Philosophical Soc., vol. 48, no. 4, ential Analyzers, master’s thesis, MIT, Cambridge, 1952, pp. 718-732. Mass., 1947. 33. B.H. Worsley, Serial Programming for Real and Ide- 14. J.N. Vardalas, The Computer Revolution in Canada: alized …, doctoral dissertation, Worsley Archives, Building National Technological Competence, MIT box 2, folder 18, p. 133. Press, 2001, pp. 16-43. 34. A. Hodges, Alan Turing: The Enigma, Walker & 15. M.R. Williams, “UTEC and Ferut: The University of Co., 2000, p. 317. Toronto’s Computation Centre,” IEEE Annals of 35. S. Millman, ed., A History of Engineering and Science the History of Computing, vol. 16, no. 2, Summer in the Bell System: Communication Sciences (1925– 1994, p. 5. 1980), AT&T Bell Laboratories, 1984, pp. 56-58. 16. Ibid., p. 6. 36. Email from David and Joyce Wheeler to author, 1 17. D. Hartree and A. Porter, “The Construction and June 2003. David Wheeler was a graduate Operation of a Model Differential Analyzer,” student in the Mathematical Laboratory at the Memoirs and Proc. Manchester Literary and Philo- time and one of the first EDSAC programmers. sophical Soc., vol. 79, 1935, pp. 51-74. 37. S.E. Lee, telephone conversation with author, 26 18. B.H. Worsley, Construction of a Model Differential May 2003. Lee, a Canadian and professor in the Analyzer, tech. report, Worsley Archives, box 3, Computer Laboratory at Cambridge, UK, in the folder 10, “Toronto Computation Centre,” 10 1990s, confirms that at the EDSAC 50th anniver- Sept. 1948. sary celebration in Apr. 1999 few could remem- 19. A. Porter, “Building the Manchester Differential ber Worsley. Analyzers: A Personal Reflection,” IEEE Annals of 38. Cambridge University Reporter, 12 Nov. 1975. the History of Computing, vol. 25, no. 2, Apr.–June 39. Univ. of Cambridge, Statutes and Ordinances of 2003, pp. 86-92. the University of Cambridge, Cambridge Univ. 20. A photograph of this analyzer can be found in Press, 1988, p. 941.

October–December 2003 61 Beatrice Worsley: Female Computer Pioneer

40. B.H. Worsley, “Computer Training at Toronto,” 51. B.H. Worsley, “Blueprint for a Library,” Computers Proc. First Conf. Training Personnel for the Comput- and Automation, vol. 8, no. 4, Apr. 1959, p. 17. ing Machine Field, A.W. Jacobson, ed., Wayne 52. R. Bregzis, C.C. Gotlieb, and C. Moore, “The Univ. Press, 1955, pp. 71-72. Beginning of Automation in the University of 41. For a more detailed look at the history of the Toronto Library, 1963–1972,” IEEE Annals of the Ferut and Transcode, see J.N. Patterson Hume, History of Computing, vol. 24, no. 2, Apr.–June “Development of Systems Software for the Ferut 2002, pp. 50-70, and Worsley Archives, box 3, Computer at the University of Toronto, 1952 to folders 25-27. 1955,” IEEE Annals of the History of Computing, 53. Worsley Archives, box 3, folders 25-27. vol. 16, no. 2, Summer 1994, pp. 13-19. 54. G.J. Groen, “An Information Retrieval System for 42. J.N.P. Hume and B.H. Worsley, “Transcode: A Sys- References and Abstracts in the Computer tem of Automatic Coding for Ferut,” J. ACM, vol. Sciences,” Third Canadian Conf. for Computing 2, no. 4, 1955, pp. 243-252; B.H. Worsley and and Data Processing, Univ. of Toronto Press, 1962, J.N.P. Hume, “A New Tool for Physicists,” Physics in pp. 136-143. Canada, vol. 10, no. 4, Summer 1955, pp. 11-20. 55. Z. Stachniak, “The Making of the MCM/70 43. M.R. Williams, “UTEC and Ferut…”, p. 11. Microcomputer,” IEEE Annals of the History of 44. B.H. Worsley, “Solutions of a Nonlinear Differen- Computing, vol. 25, no. 2, April–June 2003, p. 64. tial Equation Arising in the Theory of Diffusion 56. Letter from Dr. Kutt to Dr. Worsley, 25 Jan. 1967, Flames,” Mathematical Tables and Other Aids to Worsley Archives, box 4, folder 27. Computation, vol. 9, no. 51, 1955, pp. 112-116, 57. J. Lindsay, telephone conversation with author, and B.H. Worsley, “Numerical Representation in 26 May 2003. Fixed-Point Computers,” Computers and Automa- 58. “Canadian Computer Census,” Canadian tion, vol. 4, no. 5, 1955, pp. 10-13. Information Processing Soc. Bull., 1974, p. 76. 45. J.F. Hart and B.H. Worsley, “Self-Consistent Field 59. A. Worsley, email to author, 13 May 2003. Calculations at the University of Toronto,” Proc. First Canadian Conf. for Computing and Data Pro- Scott M. Campbell is a PhD can- cessing, Univ. of Toronto Press, 1958, pp. 298-306. didate at the Institute for the 46. D.R. Hartree, The Calculation of Atomic Structures, History and Philosophy of John Wiley & Sons, 1957. Science and Technology at the 47. B.H. Worsley, “The Self-Consistent Field with University of Toronto. He is cur- Exchange for Neon by Ferut Program,” Canadian rently studying the origins of J. Physics, vol. 36, 1958, pp. 289-299; B.H. Wors- computer science as a discipline ley, “Radial Wave Functions with Exchange for in Canada for his dissertation. V2+, Kr, and Ag+,” Proc. Royal Soc., series A, vol. Research interests include the history of computing, 247, 1958, pp. 390-399; J.F. Hart and B.H. Wors- computer science, and software and the history of ley, “Approximate Wave Functions of Pb+++ by the Canadian science and technology. His master’s thesis Method of Self-Consistent Field Without was on the history of Watfor, the popular student- Exchange,” Canadian J. Physics, vol. 37, 1959, oriented load-and-go Fortran compiler written at the pp. 983-988; and B.H. Worsley, “A Note on the University of Waterloo in the mid-1960s. He also has a Wave Functions of Krypton,” Proc. Royal Soc., bachelor of mathematics and computer science from the series A, vol. 276, no. 1328, 1962, p. 146. University of Waterloo. 48. D.B.W. Reid, L.C. Lax, and B.H. Worsley, “Error Estimation in the Transfer Rates of Plasma Readers may contact Scott Campbell at the Inst. for Constituents,” Proc. 2nd Canadian Conf. for Com- the History and Philosophy of Science and puting and Data Processing, Univ. of Toronto Technology, Univ. of Toronto, Rm. 316, Victoria Press, 1960, pp. 158-174, and L.C. Lax and B.H. College, 91 Charles Street West, Toronto Ontario M5S Worsley, “Selection of a Numerical Technique for 1K7, Canada; [email protected]. Analyzing Experimental Data of the Decay Type, etc.,” Biochimica et Biophysica Acta, vol. 59, no. 1, 1962, pp. 1-24. For further information on this or any other com- 49. Worsley Archives, box 3, folders 31-33. puting topic, please visit our Digital Library at 50. Worsley Archives, box 4, folder 2. http://computer.org/publications/dlib.

62 IEEE Annals of the History of Computing