Building IBM

Copyrighted Material
History of Computing
I. Bernard Cohen and William Aspray, editors

Editorial Board: Bernard Galler, University of Michigan, Ann Arbor,

Michigan; J. A. N. Lee, Virginia Polytechnic Institute, Blacksburg, Vir­
ginia; Arthur Norberg, Charles Babbage Institute, Minneapolis, Min­
nesota; Brian Randell, University of Newcastle, Newcastle upon Tyne;
Henry Tropp, Humboldt State College, Arcata, California; Michael
Williams, University of Calgary, Alberta; Heinz Zemanek, Vienna

Memories That Shaped an Industry, Emerson W. Pugh, 1984

The Computer Comes of Age: The People, the Hardware, and the Software,
R. Moreau, 1984
Memoirs of a Computer Pioneer, Maurice V. Wilkes, 1985
Ada: A Life and Legacy, Dorothy Stein, 1985
IBM's Early Com puters, Charles J. Bashe, L yle R. Johnson, John H.
Palmer, and Emerson W. Pugh, 1986
A Few Good Men from Univac, David E . Lundstrom, 1987
Innovating for Failure: Government Po licy and the Early British Computer In­
du stry John Hendry,
, 1990
Glory and Failure: The Difference Engines ofJoha n n Muller, CharlRs Babbag e
and Georg and Edvard Scheutz, Michael Lindgren, 1990
Johnvon Neumann and the Origins of Modern Computing, William
Aspray, 1990

IBM:5 360 and Ear�'V 370 Systems, Emerson W. Pugh, Lyle R. Jo hns on ,

andJohn H. Palmer, 1991

Building IBM: Shap i ng an Industry and Its Techn%gy, Emerson W.
Pugh, 1995

Copyrighted Material
Building IBM
Shaping an Industry and Its Technology

Emerson W. Pugh

The MIT Press

Cambridge, Massachusetts
London, England
Copyrighted Material
Third printing. 1996
© 1995 Massachusetts Institute of Technology

All rights reserved. No part of this book may be reproduced in any form by
any electronic or mechanical means (including photocopying. recording. or
information storage and retrieval) without permission in writing from the
publisher.

This book was set in Baskerville by DEKR Corporation and was printed and
bound in the United States of America.

Library of Congress Cataloging-in-Publication Data

Pugh. Emerson W.
Building IBM : shaping an industry and its technology / Emerson W.
Pugh.
p. cm. - (History o f computing)
Includes bibliographical references and index.
ISBN 0-262-16147-8
1. International Business Machines Corporation-History.
2. Computer industry-United States-History. I. Title.
II. Series.
HD9696.C64I4867 1995
338.7'61004'0973-dc20 94-21609
CIP

Copyrighted Material
Contents

Foreword Uy l. Bernard Cohen and William Aspray lX

Prefac e xi
Introduction xv

1
Hollerith: Inventar and Entrepreneur 1

Getting started. Addressing the census problem. The first practical

test. Obtaining patent protection. Entrepreneurial setbacks. Winning
the census business. Commercial applications. Entering foreign
markets.

2
Origins of IBM 19

Business stress. Battling the Census Bureau. The proposed

merger. Reaping the rewards.

3
Watson: A Man with a Mission 29
A gifted salesman. A lesson in antitrust law. The man.

4
Building an Engineering Organization 37
Competition emerges. Hiring creative engineers. Obtaining crucial pat­
ents. A drop in revenue. Managing engineers. Internal competi-
tion. The IBM card.

Copyrighted Material
vi Contents

5
Responding to the Great Depressicm 53
Coupling engineering to manufacturing. Emphasis on education.
Research for current needs. Qy,ality and production control. Methods
research. Hiring good people.

6
Support for Academic Research 67
Educational testing studies. An innovative science teacher. Automatic
computation. The Mark I computer.

7
Research for Patents and Devices 77
A prodigious inventor: Pioneering in electronics. Building electronic
calculators. Contacting outside researchers.

8
World War H Activities 89
Organizing for war: Mobile machine record units. Teleprocessing with
punched cards. The radiotype. Deciphering enemy codes. A trip to
remember. Lusting for electronics. The navy's task. A commercial
venture.

9
Future Demands 109
Lessons from NCR. Forming a committee. A department and its
demise. New leadership. The chief engineer.

10
Preparing for Peace 117
Planning for product improvements. Printer research. Emphasis on
electronics. The SSEC: a supercomputer. The Watson Laboratory.

Copyrighted Material
 Contents vii

11
Government-Funded Competition 131
Sou.nding the alarm. The ENIAC. Stored-program computers.
Promoting the concept. Commercial ventures. Provocative terminology.

12
IBM� Initial Response 145
Grooming two sons. Assignment in electronics. Getting started in
Poughkeepsie. A commercial electronic computer. The UNIVAC
threat. Applied Science representatives. Supercomputer strategies.

13
Watson, Jr., Takes Charge 163
Choosing team members. The Defense Calculator. Measures of suc­
cess. The 700 series. The most popular computer.

14
Programming Computers 183
Some early efforts. Sharing information and programs. Customer
support. Toward high level languages. The birth of FORTRAN.
COBOL arrives.

15
An Air Defense System 199
The Whirlwind computer. The new mission. A better memory.
Selecting IBM. Working together. The SAGE system. Programming.
An assessment.

16
Chasing New Technologies 221
A new laboratory. Inventing disk storage. The solid-state challenge.
Competition in supercomputers. Project Stretch. Reorganizing research.

Copyrighted Material
viii Contents

17
Legacy 243
Naming the industry. Some early business practices. Patent policies.
Dominating the industry. Settling the antitrust suit. Domestic
competition. The World Trade Corporation. Competition abroad.

18
Gambling on System/360 263
Organizing for growth Impact of the IBM 1401. Seeking
.

compatibility. The SPREAD Task Group report. Saved Uy emulation.

Announcing System/360

19
Commitment and Delivery 279
Entering the components business. Developing SLT. Memory and
storage. Achieving manufacturing goals. Stress at thp top. Software
support. Market impact.

20
Onrush of Technology 301
Going after monolithics. System/370. The FS failure. Strategic
concerns. The IBM Pc. RISC architecture.

21
Demands of the Future 317
New leaders. Antitrust issues. Coda.

Appendixes 323
Notes 331
Index 395

Copyrighted Material
Foreword

Beginning with the company's origins in the punched-card tech­

nology of the late nineteenth century, this well-researched volume
tells how IBM became so rapidly the dominant company in the
computer industry. In doing so, it provides refreshing new in­
sights on the origins and development of that industry.
The unique circumstances under which this book was written
have contributed to its remarkable qualities. It was begun as an
internally funded project to document research and development
at IBM and, as such, benefited from access to all people and
records within the company. However, with the downsizing in­
tended to make the company more competitive in the 1990s, the
project was divested and became the private property of the
author. Taking advantage of this opportunity, the author has gone
well beyond a technical history, adding insightful discussions of
internal business and cultural factors as well as external market
forces.
The author brings great knowledge and experience to this
book. Employed by IBM for thirty-five years, he has functioned
in a broad range of assignments including research scientist,
product development manager, and corporate executive. For
nearly a decade, beginning in 1983, he participated in IBM's
Technical History Project, heading it since 1985.
He is the author of three previous books published in this
series: Memories That Shaped an Industry (1984) and two volumes
written with others from the Technical History Project: IBM's
Early Computers (1986) and IBM's 360 and Early 370 Systems (1991).
In reviewing the first of these books, the Boston Globe said, "Pugh's
book is to good business magazine journalism as Jane Austen is
to New Yorker fiction." That accolade seems even better deserved
by the author's most recent effort.

Copyrighted Material
x Foreword

From its beginnings in punched-card technology through the

heyday of computer mainframes, this book dispels many myths
held within IBM and the larger computing community about the
origins, personalities, and operating structures of the company
and, as such, is a good antidote to both external journalistic
assessment and internal corporate lore. And the book offers
insight into how actions taken as early as the 1970s contributed
to the company's predicament today. For all these reasons, this
book will undoubtedl y stand as one of the major sources on IBM
well into the next century.

I. Bernard Cohen
William Aspray

Copyrighted Material
Preface

When I joined IBM in 1957, its major products were punched­

card accounting machines, time recording equipment, and elec­
tric typewriters . The rapidly growing commercial electronic
stored-program computer segment accounted for only about 10
percent of the company's gross revenue, whereas military con­
tracts still accounted for more than 15 percent. The company's
popular electric typewriters were making IBM better known to
the general public, but several years would pass before IBM would
be as familiar a name as DuPont, General Electric, General Mo­
tors, RCA, U.S. Steel, and many others. The company was well
known to the investment community, however, because of its
sustained rapid growth.
My first year's salary of $11,000 seemed enormous compared
to the small stipend I had received while working toward a doc­
torate in solid-state physics. Within a year I was promoted to
manager of a small research group. I wore a suit, white shirt, and
necktie to work as did all managers and most other employees.
We joked about the singing of IBM songs by salesmen at meetings
they occasionally held in a small conference room in the building.
But we did not joke about the policy on alcohol. It affected all of
us. Managers even worried about serving alcoholic drinks at
parties in their homes. It was good to have non-IBM employees
in attendance to be sure a private party could not be classed as a
company function.
Like most young people, I was cynical about such policies and
quick to criticize the manner in which the company was managed.
But as I gained experience managing projects in several of the
company's laboratories and advanced to executive positions with
companywide responsibilities, I came to appreciate how difficult
it is to manage any large organization and how well managed IBM
was. The enormous manae:eroeot oroblems I observed made me
c.;0PYNgntea Malenal
xii Preface

increasingly interested in the company's history, and especially in

its product development efforts. I published my first book on the
subject, Memories That Shaped an Industry, in 1984.
Since then I have had the privilege of working with others in
producing two volumes that cover major product developments
between 1945 and 1975: IBM's Early Computers and IBM's 360 and
Early 370 Systems. Published in 1986 and 1991, respectively, these
books were sponsored by IBM for the purpose of documenting
the company's technological developments as accurately as possi­
ble. They were written to appeal to specialists in the field and to
those desiring a reliable source of information about IBM and its
early computer products.
Now I have chosen to write a book that is relatively short, covers
a longer time frame and a wider range of topics, and is intended
for anyone with a desire to learn about the origins of IBM and
the computer industry. Although no longer employed by IBM, I
have been granted the use of certain facilities and access to
corporate archival materials. Without this access I could not have
written the book. I am indebted to the corporation for plac­
ing no constraints on the manner in which I tell the story and
for renouncing any right to review the manuscript prior to
publication.
The quality of the book has been aided by three people with
whom I have written books in the past: Charles J. Bashe, Lyle R.
Johnson, and John H . Palmer. They have read and commented
on the chapters as they were written and have offered numerous
suggestions and corrections. John A. Armstrong, William Aspray,
James W. Cortada, Martin Campbell-Kelly, Robert A. Myers, Ar­
thur L. Norberg, and Charles E. Pankenier read early versions of
the manuscript and offered many helpful comments. So many
others have provided personal recollections and documents,
given me tutorials on specific topics, or read and commented on
individual sections that it is impractical to list them all here.
Although I take full responsibility for the content of the book, I
am indebted to all of these individuals for helping me make it
better.
The book has also benefited from the availability of many
recently published books. In particular I want to acknowledge the
use I have made of Father, Son & Co.: My Life at IBM and Beyond
by T. J. Watson, Jr., and P. Petre; Herman Hollerith: Forgotten Giant
of Information Processing by G. D. Austrian; and the books I myself

Copyrighted Material
 Preface xiii

wrote with C. J. Bashe, L. R. Johnson, and J. H. Palmer. Finally I

would like to express my gratitude to Robert H. Godfrey andJohn
F. Maloney of the IBM Corporate Archives for helping me locate
historical documents and photographs, to Phillip D. Summers for
helpful discussions and a preliminary draft of one of the sections,
and to Caroline C. Coppola for her consultation and assistance
with text processing.

Emerson W. Pugh

Copyrighted Material
Introduction

No company of the twentieth century achieved greater success

and engendered more admiration, respect, envy, fear, and hatred
than IBM. This book tells the story of that company-how it was
formed, t
how i grew, and how it shaped and dominated the
s
information processing i n du try .

Challenging the widely held perception that the origins of the

computer industry lie in specialized electronic devices created
during World War II, the book reveals how punched-card equip­
ment created the market into which early electronic computers
were placed and defined many of the processes by which comput­
ers did their work. It tells how the technologies of IBM and its
industry evolved from electromechanical devices and punched
cards to integrated semiconductor circuits and magnetic disk
s
storage devices of previou ly unimagined capabilities. Finally it
reveals conditions and events of the 1970s that led to the severe
difficulties IBM experienced more than a decade later.
The emphasis is on the development of new technologies and
their application in products. Maintaining leadership in these
activities was crucial to IBM's success. Other topics such as manu­
facturing, marketing and sales, customer service, education, per­
sonnel practices, and organizational structure are sufficiently
discussed to provide a general overview of the evolution of the
p
company and its ractices Although the primary focus is IBM,
.

the background coverage of other organizations provides a broad

view of the origins and development of the information process­
ing industry.
Rather than accepting the popular myth that T homas J. Wat­
son, Sr. (1874-1956), founded IBM in 1914, this book shows how
a business established a quarter of a century earlier by Herman
Hollerith (1860-1929) providedthe primary basis for IBM. The
s
story b egin shortly cl3PWrQhtMP�f�Wa/first learned about the
xvi Introd 11 ction

difficulties of processing census data. It progresses through his

inventive and entrepreneurial efforts that created punched-card
equipment for processing data from the census of 1890 and that
led to improved systems and wider applications.
Chapter 2 finds Hollerith reaching the limits of his managerial
capabilities and welcoming an opportunity to sell his company.
In 1911 his business was combined with two others to create the
company Watson was hired to manage three years later. Chapters
3 through 9 provide the background of the new manager and
describe how he transformed Hollerith's small entrepreneurial
organization into the vertically integrated IBM Corporation.
These chapters emphasize Watson's efforts to strengthen research
and development and to couple them closely with manufacturing.
They tell how IBM secured a dominant position in its industry
during the Great Depression and World War II.
The problems facing the company in the postwar and cold war
years are topics of chapters 10 through 17. Of particular note is
IBM's response to the rapid transition from electromechanical to
electronic technologies. Key elements of this stor y are the reluc­
tance of Watson to participate in government-funded develop­
ment projects and the manner in whic h he and his older son
interacted . Making use of a variety of evidence, including minutes
of engineering meetings chaired by him long before his son
returned from service, the book challenges the widely accepted
view that he was slow to move into electronics.
Bringing out the company's first large-scale electronic comput­
ers was a major test ofThomasJ. Watson,Jr. (1914-1993). Trained
at the knee of IBM's legendary leader and driven by a strong
competitive spirit, the young Watson rose to the task and guided
the company through an era of unprecedented technological
change. His crowning achievement, discussed in chapters 18 and
19, was development and introduction of the IBM System/360
line of computers during the 1960s. Adopted as an industr y
standard, System/360 provided the basis for sustained growth of
IBM and the industry for nearly two decades.
In order to place the first nineteen chapters in historical per­
spective for the modern reader, chapters 20 and 21 provide a brief
overview of major activities from the mid-1960s to the end of the
1980s. Topics include IBM's response to the onrush of technology
during the 1970s and 1980s and the underlying causes of the
severe problems faced by the company during the 1990s.

Copyrighted Material
Building IBM

Copyrighted Material
1
Hollerith: Inventor and Entrepreneur

"To all whom it may concern: Be it known that I, Herman Hollerith,

a citizen of the United States, residing at New York city, in the
c ounty and State of New York, have invented certain new and
useful Improvements in Apparatus for Compiling Statistics. "
These opening phrases preceded several pages of text and
figures describing the invention, followed by the all-important
assertion: "What I claim as new, and desire to secure by Letters
Patent, is .. .. " Then came a series of specific invention claims.
This first of Hollerith's many patent applications was filed in
September 1884. He later divided it into two applications and
renewed it twice before it was finally issued as two patents (simul­
taneously with a third patent) in January 1889.1
Not yet twenty-nine, Herman Hollerith had been granted three
patents covering inventions essential to punched-card equipment
that was to tabulate data from the U.S. Census of 1890 and would
later provide the basis for the information processing industry.
The path was not easy, however. It required all of Hollerith's
entrepreneurial skills, severely strained his family life, and ulti­
mately required new management to ensure the survival of the
enterprise he had created.

Getting Started

Hollerith's family background, education, and early work experi­

ence flowed so naturally toward his lifetime career as an inventor
and entrepreneur that only the level of success he achieved can
be considered surprising. He was born in Buffalo, New York, in
February 1860. His father, a teacher of Latin and Greek, died in
an accident when the boy was only seven. His mother came from
a family whose members had been skilled locksmiths in Europe
prior to establishing themselves jv. tht e .business of designing and
C"opynghrea Ma enal
2 Building IBM

manufacturing horse-drawn carriages in America. Left to raise

her five children alone, she was guided by values that emphasized
industriousness, integrity, thrift, education, and self-reliance. Re­
fusing to ask her family for financial help, she turned her hobby
of making ladies' hats into a successful millinery business in
Buffalo.
Except for a difficulty in spelling that plagued him throughout
life, Hollerith was a good student. His mental processes tended
to be visual rather than abstract, and he had a remarkable ability
to assimilate and logically organize large quantities of informa­
tion. He attended the College of the City of New York and the
Columbia School of Mines, from which he received perfect grades
in drawing, geometry, graphics, and surveying, and was graduated
at age nineteen.2
In the fall of 1879, soon after graduating from the Columbia
School of Mines, Hollerith moved to Washington, D.C., to begin
work as a special agent for the U.s. Census of 1880. The job was
offered by one of his professors who had been appointed a chief
special agent. Hollerith's assignment was to collect and analyze
statistical information on the use of steam and water power by
the iron and steel industries.3 His academic record alone would
have qualified him for the job, but the ease with which he en­
gaged in discussions with his professors and the enthusiasm he
showed for each new task were important factors. Indeed the
pleasure he derived from the companionship of older men was
an important asset throughout the early part of his career.4
As originally conceived, the census was relatively simple. Article
I, Section 2, of the U.S. Constitution requires the Congress to
conduct an enumeration of population every ten years, "in such
Manner as they shall by Law direct." Its purpose was to ensure
fair representation and taxation among the states. The first U.S.
Census in 1790 revealed a population of just under 4 million.By
1870 the population had grown tenfold and the questions to be
asked of each person had increased manyfold to include age, sex,
race, place of birth, occupation, and more.
Responding to congressional interest in the country's growing
industries and the movement of people from farms to cities, the
census increased yet further the range of information it collected
and analyzed.The census had become a tool for policy-making
well beyond its original purposes. Because of their political sig­
nificance, census results had often been challenged. But as the
amount of information grew and the analyses became more com-
Copyrighted Material
 Hollmth: Inventor and Entrepreneur 3

plex, it was increasingly difficult to achieve results that were

meaningful and accurate. Clearly some improvement was needed
in the error-prone manual methods used for collecting, record­
ing, and analyzing the information.
Francis A. Walker, superintendent of the census of 1870 and
1880, was committed to improving the process. He increased the
number of staff members in the Washington office for the 1870
census and tripled that number to almost 1,500 for the 1880
census.s His quest for expert assistance in the classification of
diseases for the census of 1870 led to the assignment of two
assistant surgeons by the surgeon general. One of these was John
S.Billings, a talented young physician who continued his involve­
ment with the census and headed the Division of Vital Statistics
for the 1880 census. Walker's policy, that agents hired to analyze
data on the country's industrial activities must have scientific and
technical knowledge, led to the hiring of many qualified men for
the 1880 census. Among them was Herman Hollerith. Billings was
in his early forties when the nineteen-year-old Hollerith arrived.
Nevertheless the two became acquainted and quickly learned to
respect each other's capabilities and knowledge.6
To enhance the friendship and as a diversion from his other
work, Hollerith says he "computed a lot of life tables for Dr.
Billings." It was the beginning of a long relationship with pro­
found significance. In particular Hollerith credits Billings with
starting him thinking about solutions for the census problem:
"One Sunday evening at Dr. Billings' tea table, he said to me there
ought to be a machine for doing the purely mechanical work of
tabulating population and similar statistics. We talked the mat­
ter over.... He thought of using cards with the description of
the individual shown by notches punched in the edge of the
card."7
Encouraged by Billings and Superintendent Walker, Hollerith
began spending part of his time seeking ways to improve the
processing of census data. To better understand the require­
ments, he even took a temporary assignment in the population
division, working as "a clerk reporting for duty." Mter studying
the problem, Hollerith recalls, "I went back to Dr. Billings and
said I thought 1 could work out a solution for the problem and
asked him would he go in with me." But the older man was well
established in his own profession and responded that he was "not
interested any further than to see some solution of the problem
worked out."8
Copyrighted Material
4 Building IBM

In the fall of 1882, Hollerith moved to Boston to become an

instructor of mechanical engineering at the Massachusetts Insti­
tute of Technology (MIT). The position was urged upon him by
Walker, who had left the census a year earlier to become president
J
of the institute. Hollerith thrust himself enthusias cally into his
new role and soon gained the respect of his students.But before
the year was out, he knew teaching would not interest him in the
long run. He had no desire to repeat the same material for the
next group of students.
Even more important, he wanted to solve the census problem.
During his spare time, he had made use of MIT laborator y
facilities to conduct his first real experiments.The results were
encouraging, but even if he failed with the census problem, he
now knew he wanted to be an independent inventor. He had all
the tools he needed except one: knowledge of patent law and
practice.
With help from colleagues at the census, Hollerith obtained an
appointment as an assistant examiner in the U.S. Patent Office
in Washington, D.C. He reported for work as soon as his obliga­
tions at MIT had been fulfilled.His new duties offered him the
opportunity to study the structure and wording of patents and to
learn how to use the patent processes to enhance the quality and
duration of patent protection.He also reviewed the patent art in
areas pertaining to his own ideas. In less than a year he had
accomplished his objectives.
He resigned from the Patent Office in March 1884 and went
into business for himself as an "Expert and Solicitor of Patents."
Helping other inventors with their patents would provide
sufficient income while he worked on his own inventions.9

Addressing the Census Problem

It is impossible to reconstruct with certainty the inventive steps

by which Hollerith conceived the equipment used to tabulate the
census of 1 890. His recollections and those of others leave little
doubt, however, as to what got him started. It was Billings who
first urged him to work on the problem, and it was Billings who
suggested using cards to store data. There were many problems
to be solved, however, and several false starts before a practical
solution was achieved.
The difficult problem of handling individual cards led Hol­
lerith to consider using a long roll of paper tape that could be
Copyrighted Material
 Hol/erith: Inventor and Entrepreneur 5

Figure 1.1. Herman HoUerith

Photographed in Washington. D.C., in 1888, Herman Hollerith had already
invented equipment that would be used to tabulate the census of 1890. He
was born in 1860 in Buffalo, New York. Hollerith's punched-card tabulating
equipment altered forever the way government and bu si ness records were
kept and was the basis of the early data processing industry.

moved for ward and back by simple rollers. Ten years earlier the
chief clerk of the census had proposed using a continuous roll of
paper, threaded over many rollers to facilitate the manual writing
of information so as "to for m condensed tables of figures or
characters." The invention, according to his patent, "consists in
the method of bringing the widely separated columns or parts of
the paper into close proximity." Data collected in convenient
tables in this manner were subsequently analyzed by hand.1O
In contrast Hollerith proposed storing information by means
of round holes punched in a roll of wide paper tape.Each line
of holes across the width of the tape would represent information
about one individual. To tabulate the information, Hollerith pro­
posed that the tape be passed between a metal drum and a set of
wire brushes. Electromechanical counters, associated with spe­
cific hole locations, would be advanced one unit each time the
correspondingly positioned metal brush made electrical contact
with the metal drum through a hole in the tape. II

Copyrighted Material
6 Building IBM

Hollerith's decision to use electrical signals rather than me­

chanical methods to detect the holes is somewhat surprising. He
was well versed in mechanical devices, but he had taken no
courses in electricity.Indeed Thomas A. Edison had developed
his incandescent lamp only a few years earlier, and the supporting
electric power system was still being developed. There was a solid
precedent or basis for Hollerith's approach, nevertheless. The
automatic telegraph system, developed about two decades earlier,
made use of perforations in paper tape to represent information
to be transmitted electrically. Like Hollerith's proposal, the tape
was drawn over a metal cylinder, and electrical signals were cre­
ated as the perforations passed under electrical contacts.21
As his experiments and studies progressed, Hollerith became
increasingly aware of the disadvantages of paper tape.Once in­
formation was punched into the tape in some predetermined
order (for example, people listed alphabetically by name) all
further use of the information had to proceed in the same order.
If one wanted to study characteristics of one age group or ethnic
group, the entire tape would have to be scanned again and again.
Returning to Billings's original speculation, Hollerith decided
to use one card to hold all information pertaining to each indi­
vidual. A group of cards could be ordered in any desired se­
quence, and groups of cards could be combined into larger
groups or subdivided into smaller groups, for statistical analyses.
The idea of representing the characteristics of an individual by
holes punched in a card was suggested by a "punch photograph"
he observed being used in the West for identifYing railroad ticket
holders. On each ticket were locations for the conductor to
punch out holes to identifY the individual as having "light hair,
dark eyes, large nose, etc. So you see," Hollerith modestly ob­
served, "I only made a punch photograph of each person. "13
It seems likely that he was also influenced by an automated
loom invented by the Frenchman, Joseph Marie Jacquard, about
eighty years earlier. Jacquard looms in Hollerith's time routinely
produced ver y intricate patterns guided by a sequence of thou­
sands of punched cards, each with holes punched so as to specify
one step of the process. If not from other sources, he would surely
have learned about the Jacquard loom from a brother-in-law who
was involved in the silk-weaving business.
The planned function of what came to be known as the Hol­
lerith card was of course different from either of these uses.Its

Copyrighted Material
 Hollerith: Inventor and Entrepreneur 7

function was to facilitate the statistical analysis of large quantities

of information. A closely related use of punched cards had been
proposed half a century earlier by the British mathematician,
Charles Babbage. Data and instructions stored on punched cards
were to operate "analytical engines" devised by him for carrying
out mathematical computations. Hollerith was apparently not
aware of this work-probably because Babbage never completed
a working model.14

The First Practical Test

With the help of Billings, Hollerith arranged for the first practical
test of his equipment in 1886. The task was to record and tabulate
vital statistics for the Department of Health in Baltimore. Hol­
lerith recorded each death on a 31/4-by-85/s-inch (8.26-by-21.9-
centimeter) card by selectively punching out hole locations in
three rows along each of the two long edges of the cards.Each
hole represented information about the deceased, such as age,
place of birth, occupation, and cause of death.15
To tabulate data from the cards, Hollerith provided a flat work
surface behind which were several rows of dials, each dial being
associated with a specific hole location in a card. Dials could
,
record numbers from 0 to 9999 by means of two hands that
rotated in a manner analogous to the minute and second hands
of a clock, each hand pointing to one of the one hundred marked
positions on the circumference of the dial. The operator began
the tabulation process by placing a punched card on a hard
rubber surface that contained an array of tiny cups of mercury,
so spaced that there was a cup directly under each of the possible
hole locations. Then a..<; the operator pulled down the handle of
a hinged "pin box" that held an identically spaced array of spring­
loaded pins, each pin either passed through one of the punched
holes into a cup of mercury or was restrained by the card. Pins
that contacted mercury closed an electrical circuit that advanced
the corresponding counting dial by one unit. Mter a number of
cards were processed, the operator wrote down the dial readings
and reset the dials to zero.
Hollerith also devised an entirely new apparatus for sorting
cards into categories. It consisted of a wooden box with many
compartments. Each time a card was tabulated, the electrical
signals that advanced counter dials (based on which holes had

Copyrighted Material
8 Building IBM

been punched) were also used to open the lid of one of the
compartments. The operator manually removed the card from
the rubber surface, placed it in the open compartment of the
sorter, and closed the lid. He then placed the next card on the
rubber surface and repeated the process. Using this equipment
it was possible to perform a variety of statistical analyses. For
example, it was possible to sort cards according to types of occu­
pation and then determine the most common causes of death for
people in different occupations.16
The results of the test were very encouraging not only for the
young inventor but also for Billings who took pride in having
initiated the activity. In describing Hollerith's equipment and its
capabilities in a paper published by the American Public Health
Association in 1887, Billings enthusiastically asserted: "I have
watched with great interest the progress in developing and per­
fecting this machine, because seven years ago I became satisfied
that some such system was possible and desirable, and advised Mr.
Hollerith who was then engaged on census work, to take the
matter up and devise such a machine. . . . I think that he has
succeeded, and that compilers of demographical data will be glad
to know of this system. "17

Obtaining Patent Protectitm

Storing information by punching holes in cards was not new, nor

was Hollerith's proposed electrical method for detecting the
holes. Still he could obtain patent protection if he could show
that his combination of well-known devices created a new device
with useful capabilities not obvious to those "skilled in the art."
He also needed a workable embodiment of his invention.
Since no one before him had used punched cards in conjunc­
tion with electromechanical means for detecting holes and tabu­
lating data, and because the result was useful, Hollerith could
surely obtain a patent.The challenge for him was to obtain broad
patent protection that would hold up against any future chal­
lenge. The embodiment he chose for his first patent application
made use of paper tape rather than cards, but the wording of his
claims was broad enough to cover the use of cards as well. Filed
in September 1884, only six months after he resigned from the
Patent Office, this patent application was divided by him into two
applications in October 1885. There were nine invention claims

Copyrighted Material
 Hollerith: Inventar and Entrepreneur 9

0 0000
000 0000000000
0000000
00 0 0000
ooo
ooo oooo
ooooooo�
ooooooooo
ooooo
00000
000000000
0000 0000 000
00

::::: 000000
00000
:::: ::::: :::::::
ooo
ooooo ,
oooo
000 0000
QOov p
00 0
00000
ooov00
000 oo
000000 ooooo
o
ooo
oooooo
oo oo o�
000 00 00 00
0000 00000
000 000
000 00 0 0
00
000000
0000 • \
o oooo 00
ooooo
ooo oo
ooou

Figure 1.2. Early Hollerith Inventions

Top: Pantograph punch as depicted in U.S. Patent 487,737, filed by Hol­
lerith in March 1891. Bo ttom : The tabulating apparatus as illustrated in U.s.
Patent 395,781, filed by Hollerith in June 1887. The 1890 census was m ech a­
nized by apparatus similar to that illustrated here, except that there were 26
rather than 22 compartments in the sorter and 10 ra ther than 8 counter di­
al s in each of the four rows. Also the boards of electrical posts (p3 and p7)
were out of sight, and the array of counter dials was directly behind the
table.

Copyrighted Material
10 Building IBM

in the two patents. Each claim was broadly written to cover as

many different implementations as possible. IS
Hollerith's third patent was not filed until June 1887 after he
had built fully operable equipment and tested it in the Baltimore
study. The description of the equipment is therefore quite de­
tailed. There are twenty-one invention claims in the patent, rang­
ing from very broad ones to some that are quite specific to the
embodiment shown. This patent was issued in Januar y 1889,
simultaneously with the two earlier patents, applications for which
Hollerith had renewed twice in the interim. By delaying the
patent process on his first two applications, Hollerith was able to
give himself the longest possible protection (17 years from date
of issue) on all three patents. The more detailed claims of his
third patent would prevent others from simply copying his equip­
ment, even if some of his broader claims were rejected or success­
fully challenged at a later date.19
A critical flaw in the equipment was painfully revealed to Hol­
lerith during the tests in Baltimore. To punch holes in the cards,
he had used a small hand-held punch of the type employed by
train conductors. Although the conductor's punch was satisfac­
tor y for intermittent use, it was not suitable for punching thou­
sands of holes in cards. Hollerith's hand and arm became nearly
paralyzed after spending a day continually punching holes in
cards.20 To solve the problem, he devised a new punch, commonly
called a pantograph punch because large motions of a manually
moved stylus were mimicked by smaller motions of a punch that
cut round holes in the card.
Near the operator was a large flat plate with holes correspond­
ing in their locations to possible hole locations in the much
smaller card. By moving the stylus by hand and pushing it into
the selected hole in the plate, the operator caused a punch on
the levered arm to create a similarly located hole in the card.
Ready for the next test of Hollerith's system, this device permitted
an operator to punch an estimated 500 cards per day with good
accuracy and little physical strain.21 Although the improved
punch was crucial to the efficient use of Hollerith's system, it
would be impossible to get broad patent coverage because there
were already many types of punches in use. He therefore delayed
filing for a patent on the punch until March 189 1 after he had
tested it in practical applications, modified it, and used it in
processing data for the census of 1890. The patent was issued in
December 1892.22
Copyrighted Material
 Hollerith: Inventor and Entrepreneur 11

Entrepreneurial Setbacks

Shortly before filing his first patent application in 1884, Hollerith

requested a loan from his sister's husband to finance the devel­
opment of experimental equipment. It was not the first time the
two men had considered joint business opportunities. Their pre­
vious discussions had involved possible improvements in silk-weav­
ing equipment used in his brother-in-law's business.
Concerning the tabulating equipment venture, Hollerith as­
sured his brother-in-law, "I can secure what is technically known
as a 'foundation patent' covering the ground broadly and, there­
fore, all subsequent improvements would be subject to my broad
patent. "23 He also advised him that the new superintendent of the
census was willing to assign clerks to test an experimental model
of his tabulating equipment on unprocessed data from the pre­
vious census. Hollerith estimated the cost of an experimental
machine at $1,500. Another $1,000 should cover all other costs,
including the securing of domestic and foreign patents. He of­
fered to pay his brother-in-law 6 percent on the loan plus half of
the profits up to $12,000.
It was an attractive offer, and Hollerith soon received enough
money to get started. Less than a year later, however, the deal had
soured. No longer convinced it would be a profitable venture, the
brother-in-law refused to supply additional money. Hollerith pro­
tested, but to no avail. When his efforts to secure funding from
other family members also failed, he became so embittered that
he broke off all further relations with them.
24
The loss of financial support for developing tabulating equip­
ment only increased the urgency with which he pursued his
newest venture. On the last day of April 1885, he had filed three
separate patent applications with the same title: "Electro-Magneti­
cally-Operated Air-Brake for Railway-Cars." All three patents were
issued on the same day in January 1886. Although no models
were provided, the patent examiner was evidently convinced of
their operability and uniqueness.25
The patents list Hollerith's address as St. Louis, Missouri. He
had moved there to work on railway brake improvements as an
employee of an older cousin, an accomplished engineer, who had
already acquired seven patents of his own on electromagnetic
devices for railway brakes.26 Electrical devices offered the possi­
bility of activating brakes on all cars nearly simultaneously, rather
than waiting several seconds for air pressure from a single tank
Copyrighted Material
12 Building IBM

in the locomotive to make its way through long air hoses and
tubes that ran from car to car in the system George Westinghouse
had introduced more than a decade earlier. The newness of the
technologies and the complexities of the systems left considerable
uncertainty, however, as to the best solution.
During railway tests conducted at Burlington, Iowa, by the
Master Car Builders' Association in 1886 and 1887, electrically
activated air brakes proved to be superior to purely pneumatic
ones. Moreover it appears that a system using Hollerith's inven­
tions was superior to all others.But that is where his superiority
ended. The benefits of rapid standardization across the industry,
combined with the dominance of the Westinghouse Air Brake
Company, pushed Hollerith and other independent inventors
and entrepreneurs out of the field. Westinghouse offered to pur­
chase Hollerith's patents, but he refused to sell. He believed
Westinghouse would have to pay him more later, but that never
happened.27
Deeply discouraged, even embittered, Hollerith returned to
Washington. He now had no option other than to pursue his
inventions for compiling statistics on a full-time basis. It proved
to be a fortunate circumstance.

Winning the Census Business

Successful use of his equipment by the Department of Health in

Baltimore and the enthusiastic support of Billings helped Hol­
lerith gain other opportunities to demonstrate his equipment.
New Jersey adopted the system soon after Baltimore, and the
Surgeon General's Office of the War Department contracted for
a system late in 1888. During the spring of 1889 Hollerith exhib­
ited his equipment in Berlin and in Paris, and in July he installed
a system in the New York City Health Department.
Adding to the good news that year was the appointment of
Robert P. Porter as superintendent of the upcoming census. Por­
ter, like Hollerith, had worked on the U.S. Census of 1880, and
the two were quite friendly. During the intervening years, Porter
had become editor of the New York Post as well as an enthusiastic
advocate of Hollerith's equipment. He had even written in the
Post that the next census might well "be tabulated by electricity. "28
One of Porter's first actions as superintendent was to appoint
a commission of three experienced statisticians to evaluate equip­
ment for improving the census process. Headed by Hollerith's
Copyrighted Material
 Hollerith: Inventor and Entrepreneur 13

good friend, the widely respected Dr. Billings, the commission

limi ted its evaluation to H ollerith s system and two others.Using
'

1880 c ensus enumeration data from four districts of the city of

St. Louis, it was determined that the transcription of data with
Hollerith's system was at least 50 percent faster and tabulation
was more than e i ght times faster than with the second place
contender. More than two times faster overall in the test, Hol­
lerith's system was expected to be far superi or in meeting the
growing requirements for analyzing the same data in many dif­
ferent ways. On t he basis of the commission's report, Hollerith's
equipment was selected and used for the census of 1890.29
Even before the commission's final report had been submitted
in N ovember 1889, Porter authorized the installation of six Hol­
lerith machines for the Division of Vital Statistics. Another one
hundred machines were expected to be needed to process all
census data. At $ 1,000 rental per year per machine, the cost
seemed enormous. To reduce the costs, Porter insisted on the
right to use the equipment day or night. In th is way only fifty
machines would be needed. Furthermore Hollerith was to forfeit
$10 per day for any machine that was not repaired (or replaced
with an operable one) within twenty four hours after notification
-

of its malfunction. When working at night proved to be undesir­

able for all par ties Hollerith agreed to install another forty ma­
,

chines for an annual rental of only $500 each. To compensate for

the lower fee, Porter agreed that no machine would be used
outside of normal office hours.
Hollerith was fortunate in his choice of manufacturers. The
Pratt and Whitney Company manufactured the pantograph
punches and the rest of the equipment was made by the Western
El ectric Company Partly because of the quality of their construc­
.

tion and partly because of Hollerith's personal commitment and

penchant for detail, he never had to forfeit the stipulated $10 per
day for m alfunc tioning equipment.
3o
There was good reason to question the impartiality of those
who selected H ollerith s equipment for the
' 1890 census, and the
press of the time enjoyed printing abusive stories. In retrospect,
however, its superio rity over previous methods is beyond dispute.
It saved both time and money. But the system's most significant
achievement was that it made possible the analysis of data in many
different ways and with far fewer errors than had been previously
possible .

Copyrighted Material
14 Building IBM

Two features are particularly worthy of note.First, the counters

and sorting box lids could be wired to be activated either by a
single hole punched in a predesignated location or by combina­
tions of several preselected holes. Thus it was just as easy to
tabulate information based on several parameters as it was on a
single parameter. Second, the machine could reject cards with
many types of punching errors. For example, a card failing to
have either the "male"or the "female" locations punched would
result in the monitor bell just to the left of the press, refusing to
give "its cheery signal of correctness." As reported in 1891 by a
publication devoted to theoretical and applied electricity, "the
apparatus works as unerringly as the mills of the Gods, but beats
them hollow as to speed. "13

Commercial Applications

Even while Hollerith was preparing for the U.S. Census of 1890,
he had received inquiries from several foreign countries about
using his machines in their census work. Austria, Italy, Norway,
and Canada were among the first to adopt the equipment.W ith
few exceptions, the equipment was hailed for its performance and
accuracy. The director general of Italy's Imperial Bureau of Sta­
tistics evinced remarkable vision by predicting, "The time will
come when the railroads, the great factories, the mercantile
houses, and all the branches of commercial and industrial life will
be found using the Hollerith machines as a matter of not only
economy but necessity."32
Accomplishing this prediction would not be easy, however, be­
cause new techniques would have to be developed for commercial
applications. On top of the technical problems, a decline in
Hollerith's business following completion of the U.S. Census in
the early 1890s forced him to lay off all but four employees and
move his family into his mother-in-Iaw's home.33
Hollerith's initial commercial thrust was into railroads, which
during the mid-1800s had led the way in using sophisticated
accounting techniques.34 His primary target was the New York
Central, which had become the second largest railroad system in
the United States. Nearly 4 million freight waybills per year were
processed by hand. Hollerith's goal was to automate this process
so that freight movements and revenues could be handled more
economically and monitored on a weekly rather than a monthly
basis. He redesigned his cards and his equipment for railroad use.
Copyrighted Material
 Holkrilh: Inventor and Entrepreneur 15

Unlike cards used for census work in which punched-hole

locations represented personal characteristics such as age or mari­
tal status, the new cards were entirely numeric; a single digit could
be recorded in each column by a hole in one of ten positions. To
help operators read the cards manually, each column was printed
with the numerals 0 through 9 from top to bottom, and groups
of vertical columns reserved for particular classes of information
were set off from other columns by printed lines. A region be­
tween two lines was called a field, a term that endured for dec­
ades. A field of several columns was allocated to each class of
information.For example, Hollerith allotted five columns for the
weight of shipments to permit values up to 99,999 pounds.When
an amount of money was recorded, the right two columns indi­
cated cents and the adjacent columns indicated dollars up to the
largest number anticipated. A three-column field, for example,
could record values up to $9.99.
No longer did simple counting suffice. Hollerith made substan­
tial modifications to his tabulating equipment to perform addi­
tion-that is, to add a number recorded in a field on one card
to those in the same field on preceding cards. The dials were
replaced with accumulators having sufficient digit positions to
handle sums from the assigned fields. Each digit position was
represented by a ten-position wheel with visible numerals. The
sizes of the fields and associated accumulators were selected for
the needs of the New York Central,35
A significant innovation that helped reduce the cost of this new
function was introduced soon after Hollerith replaced the pin­
box card reader with an automatic card feeder.The new method
for performing addition made use of the distance between
punched holes to represent number size. Represented by holes
punched in a row near the top of the card were Os, and holes
representing 6s, for example, were six times further from the Os
than were holes representing Is. This arrangement made it pos­
sible to advance each digit in the accumulator by the desired
amount using an electrically actuated clutch that mechanically
coupled the rotation of the ten-digit wheels of the accumulator
to the motion of the cards.36
Hollerith had moved his business from downtown Washington,
D.C., to a former cooper's shop in Georgetown in 1892. The
two-story brick building, to which a third stor y and other expan­
sions were subsequently added, served as the card manufacturing
plant, final assembly plant, repair shop, and development labora-
Copyrighted Material
16 Building IBM

tor y. Subassemblies were obtained from numerous suppliers. The

shop also served as a training facility for salesmen, whom Hol­
lerith trained by requiring them to take apart and reassemble
equipment. In September 1896 Hollerith signed a contract agree­
ing to supply the New York Central with sufficient punched-card
equipment to process up to 4 million waybills during the next
year. That December he incorporated his business as the Tabulat­
ing Machine Company.37
Success in the commercial field came slowly, partly because
desk-top calculating machines were more easily absorbed into
standard business routines than were tabulating machines. Hol­
lerith was therefore grateful for a large order for census equip­
ment from the Russian government that helped carry the
business until the next U.S. Census. Tabulators used in the 1890
census again served in the 1900 census of population, but tabu­
lators of the type developed for the New York Central Railroad
were needed to process farm crop data. For these newer tabula­
tors, Hollerith also developed an automatic sorter. His manually
operated sorter was becoming impractical owing to the large
volume of cards being processed.38

Entering Foreign Markets

Despite the rapidity with which Hollerith's equipment was

adopted for census purposes by several foreign governments, the
broader development of foreign markets progressed slowly. Then
in 1901 Robert Porter fortuitously came back into Hollerith's life
to create his first permanent business relationship outside the
United States. Porter, it win be recalled, had first met Hollerith
when they worked together on the V. S. Census of 1880. Later as
editor of the New York Post, he had promoted Hollerith's system.
Appointed superintendent of the census of 1890, Porter not only
supervised the first significant use of the equipment, but he had
heen instrumental in its selection.
Leaving the census in 1893, Porter was appointed special fiscal
adviser and tariff commissioner for Cuba and Puerto Rico. He
also arranged to join Hollerith in giving presentations on tabu­
lating the U.S . Census to the Royal Statistical Society in London.
Porter's professional credentials were no doubt sufficient to war­
rant this opportunity, but the fact that he was a naturalized U. S.
citizen of British birth added interest to the presentations and
contributed to the warm reception the two men received.39

Copyrighted Material
 Hollerith: Inventor and Entrepreneur 17

Hollerith preceded this event with an arduous tour of Euro­

pean capitals to promote his system. He arrived in Berlin in
November 1893 and then went on to Rome, Paris, and London,
where he joined Porter. When his tabulating equipment failed to
arrive in time for his presentation, Hollerith spoke extemporane­
ously using lantern slides prepared for such an emergency.40
Porter's wife, who attended the event, graciously wrote to Hol­
lerith's wife, telling her how brilliantly her husband had over­
come the unexpected problem. ''You would have been proud of
Mr. H's appearance before that scientific and august bald-headed
body," she wrote. "He rose to the occasion and in the easiest
possible manner and in the most lucid language gave his distin­
guished audience a clear idea of the machine."41 Hollerith was
subsequently elected a member of the society-one of many
honors he sought and received throughout his life.
Despite this favorable introduction to the British scientific com­
munity, Hollerith failed to find suitable backers for a permanent
outlet in Britain for several years. Then in 190 1 President W illiam
McKinley was assassinated, bringing to an end Porter's political
career. Porter accepted an offer to return to England to become
the first editor of the Engineering Supplement of the Times.
Ver y likely Porter had already given thought to establishing a
British outlet for Hollerith machines. In any event he acquired
an exclusive option to establish one from Hollerith in February
1902. The British company was finally incorporated as the Tabu­
lator Limited in June 1904. Porter was chairman and Raleigh
Phillpotts, a thirty-year-old specialist in commercial law, was gen­
eral manager. Phillpotts's most relevant prior experience was
three years as secretar y of the British Westinghouse Company.
Another important asset, especially for a businessman in England,
was that he came from a prominent family.
Progress in getting equipment accepted by customers was slow,
in part because of the inexperience of the company's manage­
ment. For example, it was about a year before they realized that
the equipment's decimal counters could not handle the nondeci­
mal British currency then in use. Since they had raised only 10
percent of the 20,000 pounds specified by the original agreement,
operating funds were in short supply. To solve this problem,
Phillpotts negotiated a new agreement that replaced a one-time
payment of 10,000 pounds for use of Hollerith's patents with a
25 percent royalty on all future revenues. These high royalty

Copyrighted Material
18 Building IBM

payments were later viewed as a primar y reason for the relatively

poor success of the British firm.42
In 1905 the company established its first machine installation
with the Vickers company, and Porter used his influence with the
general manager of the Lancashire and North Yorkshire Railway
to arrange for a trial installation. A large installation resulted,
involving five tabulators and many key punches. Now the com­
pany needed more money to purchase equipment to place on
rental. With its better business prospects, it was able to sell
enough shares of stock to solve the immediate problems, al­
though friends and relatives again had to be solicited. To make
the company more attractive to investors and customers, its name
was changed in October 1907 to the British Tabulating Machine
Company (BTM).
The selling of each new installation continued to require con­
siderable effort, and business expanded slowly.In 1910 BTM still
had only five employees when a major effort was undertaken to
win the contract for the British census. Trials that summer indi­
cated modified equipment with thirty-six counters would be re­
quired. Despite the short time available to make these
modifications, the contract was signed in December. Calling for
eight counting machines, fifteen sorters, and sixty punches, the
contract represented more revenue than all previous installations
combined. Inevitably problems were encountered, but the census
was successfully tabulated during a two-year period beginning in
mid-1911.43
In 1910 a German businessman obtained from Hollerith the
right to establish a firm to serve the German market. Headquar­
tered in Berlin and named the Deutsche Hollerith Maschinen
Gesellschaft, the new company was more commonly referred to
by the acronym Dehomag. Like BTM, Dehomag paid the Ameri­
can company 10 percent over the cost of the machines plus a
royalty of 25 percent of gross rentals. Perhaps learning indirectly
from the experiences of the British company, the German com­
pany was quickly funded and profitable. Soon it was paying Hol­
lerith's Tabulating Machine Company an additional annual fee
to operate in Denmark, Nor way, Sweden, and Switzerland as
well.44

Copyrighted Material
2
Origins of IBM

After tabulating the U.S. Census of 1900, Hollerith enjoyed con­

siderable success in the commercial market. He had also estab­
lished an outlet for his equipment in Great Britain and Germany
by the end of the decade. His pleasure in these achievements,
however, was undercut by growing problems with his primary
customer, the Bureau of the Census.
More an inventor and entrepreneur than a businessman or an
executive, Hollerith had never been able to remove himself from
the details of his entire operation. His devices were his personal
creation. Their successes and failures were his own. When ma­
chine part.. did not arrive on time, he became an expeditor. When
money was needed, he sought a loan. When there were legal
problems, he handled them with his attorneys. His health suf­
fered from the strain, and he longed to have more time with his
family. Nearly over whelmed by his many problems, Hollerith
accepted a merger proposal in 1911 that made him a wealthy
man. The merger also triggered a series of events that were
destined to propel his small business into one of the larger and
more successful business enterprises of the century.

Business Stress

In spite of the intensity with which Hollerith pursued his intellec­

tual and business interests, he had always devoted some time to
social activities. Frequently this took the form of discussions with
older men or women, but he also participated in leisure activities
with others his age. His first fiancee died of typhoid fever in 1886,
one year after their engagement was announced. Two years later
he became engaged again, this time to Lucia Talcott, a young
woman whose widowed mother had been married to an engineer.
Estranged from his own family, Hollerith derived considerable
Copyrighted Material
20 Building IBM

comfort not only from Lu but also from her mother who had
much in common with his own.
For nearly two years Hollerith and Lu were formally engaged,
but he continually postponed the marriage, hoping he could first
get out from under his business uncertainties and pressures.
Winning the contract for the census of 1890 reduced the uncer­
tainties, but it dramatically increased the pressures. Ordering,
installing, and servicing so many machines allowed no time for a
wedding. So great was Hollerith's stress that he suffered from
severe headaches, inability to sleep at night, and an uncon­
trollable temper by day. When his doctor ordered him to leave
town for a complete rest, he concluded there could be no rest
for him without Lu. He contacted her immediately and a wedding
was hastily arranged. Three days later he and his new wife headed
off for their honeymoon and his complete rest. It appears to have
been a good marriage-strong enough to withstand the vagaries
of Hollerith's business as well as the peaks and valleys of his
emotions.'
Except for its sharp ten-year cycle, the Census Bureau work
provided an ideal beginning for his business. All of the equip­
ment was installed in the same geographic location and all of the
systems were essentially identical. Hollerith, the one-man entre­
preneur, could supervise their installation and maintenance,
while leaving Hollerith, the inventor, time to work on major
improvements. With business customers by contrast, he had to
spend more of his time learning about diverse applications and
making minor modifications to his equipment rather than mak­
ing more basic improvements.
To reduce the burden of managing his business, Hollerith
needed more employees on whom he could depend, but he
lacked the personal skills to attract and retain them. His relation­
ship with his first employee, Edmund Talcott, who was also his
wife's younger brother, is particularly revealing. Soon after the
contract for the 1890 census was signed, Hollerith hired him to
install and service machines and to rewire them for different
applications. Rewiring consisted of unsoldering selected copper
wires from one set of terminals and resoldering them to others.
Servicing the equipment included a daily recharging of two-foot­
high storage batteries, using Edison power lines that came into
the building. Batteries were used to power the equipment because
of the uncertain voltage levels and continuity of service of the
power lines.
Copyrighted Material
 Origins of IBM 21

Ned Talcott learned to perform all chores with diligence and

was essential in achieving the rapid buildup of equipment re­
quired by the Census Bureau. He was also exceptionally loyal.
Termination of the census work in 1894 coincided with a nation­
wide financial depression, forcing Hollerith to lay off all but Ned
Talcott and three other employees. When Ned obtained another
job so his brother-in-law would have fewer people on the payroll
but did not explain the move to Hollerith, the latter became so
furious that the younger man felt obliged to abandon his new
job.
Two years later an angry dispute erupted over Ned Talcott's
lack of initiative-and his wages, which Hollerith had been un­
able to pay. In a fit of anger, Hollerith fired him. It was a decision
the two regretted almost immediately and for years to come, but
neither one was able to bring about the mutually desired recon­
ciliation.2

Battling the Census Bureau

Probably no events were more frustrating and emotionally drain­

ing for Hollerith than those that followed creation of a perma­
nent U.S. Bureau of the Census in 1902. Simon Newton Dexter
North, its newly appointed director in 1903, became concerned
that the government might be paying Hollerith excessively high
prices.
As the record revealed, the cost per capita of the 1870 census
had been 8. 8 cents; of the 1880 census, ll.5 cents; and of the
1890 census with Hollerith's equipment, 18.4 cents. The in­
creased costs do not seem unreasonable in view of the larger
amount of data collected and analyzed per capita.3 From another
perspective, however, the rates were high. The rental Hollerith
charged was based on the number of cards processed: 65 cents
per 1,000 cards tabulated and 18 cents per 1,000 cards sorted.
Using the old machine, an operator could typically tabulate 8,000
cards per day, resulting in an annual rental of about $1,500 per
machine. The new automatic machines could process four to five
times more cards, yielding an annual rental greater than $6,000.
This was more than the cost of building a machine. Hollerith was
also the sole supplier of the cards. The faster his machines
worked, the more cards he sold.
The potential political embarrassment for North was height­
ened in 1904 when his predecessor accepted the position of
Copyrighted Material
22 Building IBM

president of the Tabulating Machine Company. Any future con­

tracts between the company and the Bureau of the Census might
well come under considerable scrutiny.4
To break the hold Hollerith had over the Census Bureau, North
obtained a $40,000 appropriation from Congress in July 1905 to
cover experimental work to develop new tabulating machinery.
Almost simultaneously all of Hollerith's machines were removed
from the Census Office. Explaining this event, North cited the
failure to reach a mutually acceptable rental agreement, whereas
Hollerith cited his company's reluctance "to leave its patented
devices where they could serve as models for experimental
work.u5
As events unfolded, North established a machine shop in the
Bureau of Standards (later moved to larger quarters in the Census
Bureau) and staffed it with people, several of whom had pre­
viously worked for Hollerith. Learning that Hollerith's primary
patents would soon expire, North set the objective of copying the
basic designs and then improving on these in ways that would
circumvent more recent patents.
Meanwhile Hollerith was becoming increasingly incensed by
the government's effort to prevent him from profiting from his
inventions. Even if he could show that the Bureau of the Census
was infringing his patents, the laws of the time made it doubtful
that he could stop the bureau or collect damages as he might
from a private company. Almost more galling was a new policy
that permitted Census Bureau employees to acquire patents in
their own name for their own use, so long as the government also
had free use. In addition to driving Hollerith out of the census
business, the government was in effect paying to develop patents
that could be used by individuals or private companies in direct
competition with him in the commercial market. Hollerith be­
lieved the practice was grossly unfair. He lobbied influential
friends and congressmen. and he took his case directly to Presi­
dent Theodore Roosevelt and later to President William Howard
Taft.
His activities contributed to the ouster of North as director of
the Bureau of the Census. He even obtained a temporary court
injunction in Januar y 1910 to prevent the bureau from making
equipment modifications essential to carrying out the constitu­
tionally mandated census of that year. Ultimately his efforts failed.
The census of 1910 was taken using equipment built by the
Bureau of the Census.Hollerith received no compensation for
Copyrighted Material
 Origins of IBM 23

patent infringement-if indeed his patents had been infringed.

Even before his suit against the government ended, Hollerith
decided not to seek damages in the Court of Claims. He preferred
to concentrate on new business activities rather than spending
his limited resources seeking an uncertain outcome.6
"Mter my row with North," Hollerith stoically recalled some
years later, "I devoted my attention entirely to commercial work.
However, I always have regretted that I could not stay in census
work long enough to carry out my ideas regarding verification
machines."7 It was the lament of a man whose driving motivations
were to solve engineering problems and create useful inventions.
He enjoyed his business primarily because it was a means for
proving the worth of his inventions. Increasingly, however, he saw
his business obligations as preventing him from doing interesting
work.

The Proposed Merger

In the past Hollerith had turned down offers to sell his business.
But in 1911 any reluctance he had to sell the business was over­
come by his frustration with the actions of the Bureau of the
Census and by a very attractive proposal presented by the finan­
cier Charles Ranlett Flint.
Born in Maine in 1850 to a family in the shipping and ship­
building industries, the young Flint graduated from Brooklyn
Polytechnic at age eighteen. Declining his father's offer to send
him to college, he began his career as a dock clerk in New York
City. Four years later he was a partner in the W. R. Grace and
Company shipping firm. By the time he was thirty, he had become
an independent businessman, procuring and shipping munitions
to Peru and Brazil and serving as consul for Chile in New York,
among other activities. His diverse involvements included an un­
successful effort, as president of the United States Electric Com­
pany in 1880, to become part of the national surge in business
consolidations by combining major elements of the electric light
and power industries. Attributing his failure to his position as
president of one of the involved companies, Flint decided to
pursue future consolidations in the role of adviser and facilitator.s
Many organizational structures were available, ranging from
loose federations to a full merger of the operations of two or more
companies. Among the structures in use in the 1880s was the
"trust," a device in which several companies turned their stock
Copyrighted Material
24 Building IBM

Figure 2.1. Charles R. Flint

Known to the press as "the father of trusts," Charles Ranlett Flint was enter­
ing his sixties when he merged Hollerith's company with two others in 1911
to form the Computing-Tabulating-Recording Company. The name of the
company was changed to the International Business Machines Corporation
in 1924.

over to a board of trustees in return for "trust certificates" of

equal value. The board of trustees was empowered to make op­
erating and investment decisions on behalf of the companies
within the trust. Large financial resources, and control of major
industry segments, provided trusts with opportunities to squeeze
out smaller companies and ultimately gain monopoly power.
Exploitation of monopoly powers by the railroads and other
industries had caused the federal government to become involved
in protecting local regions and small companies from ruthless
exploitation. The Interstate Commerce Act of 1887 and the Sher­
man Antitrust Act of 1890 were intended to stop undesirable
monopoly practices, but they had only limited effect until Presi­
dent Theodore Roosevelt initiated his aggressive "trust-busting"
activities in the early 1900s. Thereafter business consolidations
had to be carefully structured to reduce the likelihood of antitrust
action by the government.9
By the time he approached Hollerith in 1911, Flint was among
the better-known speculators and promoters of industrial merg­
ers. Well on his way toward a lifetime total of two dozen industrial
consolidations, Flint had socialized and negotiated with many of

Copyrighted Material
 Origins of IBM 25

the countr y's leading inventors, entrepreneurs, financiers, and

industrial leaders. He had also found time to become a skilled
aviator, to own and operate one of the fastest racing yachts in the
country, and to join with six others in founding the Automobile
Club of America.
Flint favored large corporations because they had economies
of scale that created profits needed to improve products and
production methods and to expand business. He derided compe­
tition as a destructive influence that forced manufacturers to
lower the quality of products in order to meet competitive prices.
Among companies created by his mergers were the United States
Rubber, American Woolen, and American Chicle companies.
Flint obtained substantial profits from mergers by selling stock at
prices well above the tangible assets of the corporations. For
example, in the case of American Chicle, "the chewing gum
trust," the tangible assets of the six merged companies were $0.5
million, whereas Flint capitalized the corporation at $9 million
based on his estimate of the value of their trademarks.
The plan Flint offered Hollerith and his fellow owners of the
Tabulating Machine Company was a merger with the Interna­
tional Time Recording Co mpan y and the Computing Scale Com­
pany of America. It would be Flint's first consolidation of
manufacturers of similar but not identical products.10 A ten-year
series of mergers and acquisitions, facilitated by Flint, had helped
the International Time Recording Company dominate its indus­
try with annual sales throughout the United States and Europe
exceeding $1 million. Flint had also been involved in creating the
Computing Scale Company of America, but its business was less
successful. By merging the two firms, Flint hoped the marketing
and financial resources of the former would strengthen the latter.
The subsequent decision to include Hollerith's Tabulating Ma­
chine Company could be rationalized by the company's chronic
shortage of funds to support its growth-a shortage exacerbated
by Hollerith's policy of renting machines rather than selling them
outright. Flint also believed there would be marketing benefits
since the three companies produced products sold directly to
other businesses. Finally, there was some hope of technological
synergism be�ause all three companies were dependent on new
inventions. Each could trace its origins to patents issued and
entrepreneurial activities begun in the latter half of the 1880s.11
The first of these patents was issued in 1885 to Julius E. Pitrat
of Gallipolis, Ohio, for a scale equipped with a chart to enable
Copyrighted Material
26 Building IBM

clerks to determine an item's price at the same time it was being

weighed.12 Through the efforts of those who bought his patent,
the world's first computing scales were manufactured in Dayton,
Ohio, beginning in 1889, and the Computing Scale Company was
formed. In 1899 the Moneyweight Scale Company was organized
to serve as the selling agent in the United States and Canada.
Under Flint's guidance in 1901, the Computing Scale Company
of America was incorporated to take over the properties and
businesses of the Computing Scale Company, the Moneyweight
Scale Company, and the Detroit Automatic Scale Company.
The second corporate entity to be merged with Hollerith's
company began operations in September 1889 when the Bundy
Manufacturing Company was incorporated through the efforts of
Harlow E. Bundy. Its product was a time clock invented by his
brother, Willard L. Bundy, a jeweler of Auburn, New York. Willard
Bundy's patent on time clocks had been issued in 1888.13 To use
the Bundy device, each workman was given his own key bearing
his number. To check in or out, he inserted his key in the
mechanism and gave it a quarter-turn. This action printed his
number and the time on a roll of paper that became the plant
time record.
In 1896 Harlow Bundy persuaded George W. Fairchild, a self­
made businessman and publisher of a small-town newspaper in
upstate New York, to become a substantial stockholder and active
participant in the affairs of the Bundy Manufacturing Company.
Three years later, the Bundy company acquired the Standard
Time Stamp Company. Then followed a series of structural
changes that by 1907 had created the International Time Record­
ing Company as it was when the three-way merger was proposed.
The restructuring had included acquisition of the Willard and
Frick Manufacturing Company in 1900, the Chicago Time Regis­
ter Company in 1901, and the Dey Time Register Company in
1907. The Willard and Frick company manufactured the world's
first card time-recorder, and Dey Time Register Company manu­
factured a popular dial time recorder that did not require each
worker to be issued a key. These products were based on patents
issued in 1894 and 1888, respectively.14
Hollerith's patents and business formed the third component
of Flint's proposed three-way merger. Eventually all parties agreed
to the plan, and inJune 1911 the Computing-Tabulating-Record­
ing Company (CTR) was incorporated. Its name was derived from
�he three companies whose businesses were merged: the Comput-
Copyrighted Material
 Origins of IBM 27

ing Scale Company, the Tabulating Machine Company, and the

International Time Recording Company. The value Flint ascribed
to CTR was $17.5 million, only $1 million of which was based on
tangible assets.I5

Reaping the Rewards

The price set for the stock of the Tabulating Machine Company
was $2.3 million. Hollerith received just over half. In those days
majority stockholders often negotiated more for their own shares
than was subsequently received by others, hut Hollerith insisted
that all stockholders receive the same amount per share. Among
those henefiting from Hollerith's fairness was Eugene A. Ford, an
engineer and inventor who had helped Hollerith implement
some of his designs and who later played a key role in the merged
organization. To Hollerith he wrote: "Please accept my thanks for
procuring me the same price for my stock as for your own. I have
been wondering how much more you could have gotten for yours
if you had let the minority make its own deal, and also how many
men there are who would have acted as you did. "16
Mter the deal was consummated, Hollerith took advantage of
his new freedom and wealth to build a home for his family in
Georgetown and to establish a small farm in Virginia. But he was
not yet prepared to give up his involvement with the Tabulating
Machine Company. He negotiated a contract to work for the
company as a consulting engineer for ten years at the large salary,
for the time, of $20,000 per year. The contract stipulated that
Hollerith "shall not be subject to the orders of any officer or other
person connected with the Company and shall be allowed to
perform his services in such manner and at such place as [he]
may decide." In return Hollerith agreed "to devote his best ener­
gies to the development" of new product features, to "assign to
the Company all inventions and improvements ... without fur­
ther expense to the Company," and to refrain from engineering
in "any business competing with business of the Company. "I7
The chairman of the board of the new company was George
Fairchild, who had played an important role in the series of
mergers that had created the International Time Recording Com­
pany and now had created the Computing-Tabulating-Recording
Company. A member of the U.S. Congress since 1906, Fairchild
was not expected to take an active part in the business. IS The first
Copyrighted Material
28 Building IBM

president was an associate of Flint who served only one month to

get the company started. Fairchild reluctantly took over as acting
president until April 1912 when Frank N. Kondolf, formerly chief
operating officer of the International Time Recording Company,
was chosen.
In May 1914 Thomas J. Watson was hired as general manager
of CTR, becoming president in March 1915. Nine years later, in
February 1924, Watson changed the name of the company to
International Business Machines Corporation, a name that had
grown in favor following its adoption in 1917 by CTR's Canadian
subsidiary.19
Watson had the greatest interest in the products of the Tabu­
lating Machine Company, which he had observed in operation
long before he was approached to manage CTR. Products of the
Computing Scales Company interested him the least, and under
his leadership this part of the business failed to prosper. In 1935
the scales business was sold to the Hobart Manufacturing Com­
pany in exchange for shares of Hobart stock. Much to Watson's
chagrin, the scales business almost immediately began making
money after it was sold.20
Although the time equipment business was reasonably
profitable, it became an ever smaller segment of the company's
business and was finally sold to the Simplex Time Recorder Com­
pany in October 1958.21 Long before then the product line initi­
ated by Hollerith dominated all other sources of IBM's revenue
combined. Thus the two businesses that were merged with Hol­
lerith's in 1911 can be viewed primarily as catalysts for a process
that provided the Tabulating Machine Company with financial
backing and experienced management at a critical time.22
It is traditional in IBM to honor Watson by equating the found­
ing of the company with his arrival as general manager of CTR. 23
From a historical perspective, however, the founding of IBM
might better be dated to the founding of Hollerith's business.
Although other candidates for the "founding event" exist, win­
ning the contract in late 1889 to process data from the upcoming
census literally put Hollerith in business.

Copyrighted Material
3
Watson: A Man with a Mission

"I only want a gentleman's salar y and a part of the profits when
1 am able to make the company a success," Watson told the
directors of the Computing-Tabulating-Recording Company
(CTR) when queried about the compensation he would expect if
hired as general manager. Asked why he had left the National
Cash Register Company (NCR), Watson responded, "Because Mr.
Patterson asked for my resignation." These straightforward an­
swers, combined with Watson's outstanding record as a salesman
and man ager, satisfied the directors, and they hired him as gen­
eral manager of CTR beginning in May 1914.1
Being dismissed by the idiosyncratic John H. Patterson of NCR
was in some ways a tribute. Patterson had a reputation for firing
those who had risen far enough to challenge his authority. In
spite of his forced departure from NCR, Watson never com­
plained or lost his respect for the man about whom he would
later say, "Nearly everything I know about building a business
comes from Mr. Patterson." The father of modern salesmanship,
as Patterson has been dubbed, made his salesmen memorize
standardized sales pitches, motivated them at emotionally
charged meetings, and set challenging goals.2
Patterson had created arguably the best sales organization of
the time, but from him and NCR there was far more than sales­
manship to learn about business. In the mid-1880s the company
had begun to establish a vertically integrated structure in which
all facets of its business were centrally managed. At that time
manufacturers of office equipment typically sold their products
through independent sales agents. Remunerated primarily by
commissions on sales, these agents generally had little incentive
or capability to provide advertising and technical support for
salesmen or credit and other services for customers. Manufactur­
ers were also likely to. contract Qut much of the manufacturing
r..;opyngnred Matenal
30 Building IBM

Figure 3.1. Thomas J. Watson

Watson is shown (top) in his office soon after he was appointed general man­
ager of the CTR in 1914. He was photographed as presid ent (bottom, left)
in 1920 with board chairman George W. Fairchild. Thousands of copies of
the photo-po rtrait of IBM p resi dent Watson. created by Yousuf Karsh in 1948
(bottom, right), were displayed throughout company buildings for many
years.

Copyrighted Material
 Watson: A Man with a Mission 31

process and to hire independent engineers to devise improved

products. By placing all of these functions under one central
management, NCR was able to expand its size substantially and
dominate its industr y. Vertical integration could provide in­
creased profits because of the interdependence of mass produc­
tion and mass marketing and the economies of scale available in
both.
The successful methods of NCR and of the Remington Type­
writer Company (which had created an integrated business struc­
ture at about the same time) were soon followed by others in the
office equipment business. A. B. Dick & Company, developers of
the mimeograph machine, and the Burroughs Adding Machine
Company had integrated business structures by about 1900.3
Thomas]. Watson, starting with CTR in 1914, was to become on e
of the more successful followers. At NCR he had learned just
about everything there was to know about structuring and man­
aging such an enterprise.

A Gifted Salesman

Even before he joined NCR in the Buffalo office in 1895, Watson

had demonstrated his natural gift for salesmanship. He had come
to Buffalo to seek his fortune after spending two years selling
organs, pianos, and sewing machines for a local businessman in
the small town of Painted Post, about twenty miles west of Elmira,
New York. At his father's insistence, he had also enhanced his
basic eighth-grade education by studying business and accounting
for a year at the Miller School of Commerce in Elmira. Watson
had lived and worked in and around Painted Post all his young
life. He regarded it as his hometown although he was born in the
nearby village of East Campbell on 17 Februar y 1874. His father,
a man of Scottish and Irish descent who derived only a modest
living from farming and lumbering, was ambitious for his son and
urged him to leave Painted Post to improve his prospects.4
His first job in Buffalo was much like his job in Painted Post,
selling sewing machines off the back of a wagon. One day he
followed the custom of many of his cronies by going into a
roadside saloon to celebrate a sale. He had too much to drink,
and when the bar closed, he found that his horse, buggy, and
merchandise had been stolen. His employer dismissed him and
dunned him for the lost property. Word got around, and Watson
had difficulty findin� another job.5 He was reduced to sleeping
-Copyrightea Material
32 Building IBM

on a pile of sponges in the basement of a drugstore and having

the only suit he owned cleaned and pressed while he waited in a
back room.
His next job was as an assistant to a man selling shares of stock
in the building and loan business. Investors paid for the stock in
installments, and the salesmen kept the first installment as com­
mission. Watson made so much money that he decided to go into
business for himself by opening a butcher shop. His employer in
the stock business turned out to be a cheat, who abruptly left
town with the result that Watson was forced to sell his butcher
shop to pay his debts.6
Desperate for a job, young Watson talked his way into one at
the Buffalo office of the National Cash Register Company. It was
October 1895; he was twenty-one years old. If he could prove
himself, he was told, he would become a regular salesman. At
NCR he studied the Primer, reportedly the first fully structured
sales pitch in America and the brainchild of the company's presi­
dent, John Patterson. Watson learned how to use the canned
pitch, how to var y it, and how to combine it with his own natural
enthusiasm and sincerity. He succeeded so well that, at age twenty­
five, he was put in charge of sales for the Rochester office. The
sales performance of the office, as measured against its sales
quota, was near the bottom for NCR offices when he took over.
Three months later, it ranked near the top.
In October 1903 he was summoned to the head office in
Dayton, Ohio, to meet the fabled president of NCR. Watson was
now twenty-nine years old. The proposition Patterson put before
him was challenging. The company had produced such reliable,
long-lasting machines that the market for secondhand machines
was encroaching seriously on the market for new ones. Having a
near monopoly of the market for new machines, Patterson had
decided to expand the company's revenues by entering the sec­
ondhand market. The proposed business was to be financed by
NCR, but this relationship was to be kept secret.
Watson was put in charge of the entire operation. As described
by his biographers: "It was rough business. Watson established
stores next to successful competitors, copied their successes, dis­
carded their failures, hired their salesmen, undersold them, even­
tually put them out of business. His experiences during those
years Watson always regretted and rarely mentioned. " 7 Not until
three years after the surreptitious activity was begun did NCR
publicly announce that Thomas J. Watson was in charge of its
Copyrighted Material
 Watson: A Man with a Mission 33

secondhand business. One and a half years later, in August 1908,

Watson was named assistant sales manager, and the secondhand
business was gradually absorbed into the regular sales offices. In
effect he was the sales manager because the nominal sales man­
ager was spending most of his time on other company matters.
In May 1910 Watson became sales manager in name as well as
fact.
During his years in the secondhand business, Watson learned
a great deal about the inner workings of cash registers. Far more
than in selling new equipment, secondhand equipment sales
depended on a salesman's ability to explain how the mechanical
structure of the machine provided superior function and dura­
bility to those machines he was selling against. Because of this
knowledge, Patterson asked Watson to attend meetings of his
Future Demands Department, which had responsibility for antici­
pating future customer demands and translating them into prod­
ucts. Through this activity Watson gained considerable
experience and responsibility for NCR's engineering effort at the
same time that he was head of the sales organization. In Patter­
son's view it was more important for the people responsible for
engineering to know what was needed than it was to know how
to achieve it. 8

A Lesson in Antitrust Law

The manner of entering and running the secondhand business

is indicative of the vigor with which Patterson sought to preserve
the cash register business for the only company he believed
deserved to have it. It is perhaps not surprising therefore that in
Februar y 1912 Patterson and twenty-nine other officials of NCR,
including Watson, were indicted and brought to trial in the
Federal Court for the Southern District of Ohio for violation of
the Sherman Antitrust Act. According to the prosecuting attor­
neys, the company had used "unfair, ?ppressive, tortuous, illegal,
and unlawful means" to create its near monopoly of the industr y.
Reportedly NCR had over 90 percent of the cash register business.
One year after the suit was filed, the jur y reached its verdict. All
but one of the defendants was guilty. Watson, Patterson, and one
9
other were given the maximum sentence of one year in jai1.
While the case was being appealed, tensions mounted between
Watson and Patterson. The latter perceived that Watson was over­
stepping his bounds and "d�veloping q�ite a following among the
Copynghted Marer/al
34 Building IBM

rank and file."lo Patterson made Watson's position increasingly

unpleasant until November 1913 when it was announced that
Watson had resigned.ll About to turn forty, out of work, and his
first child due in less than two months, Watson was nevertheless
ver y selective about his next position. He turned down several
opportunities before accepting the position of general manager
of CTR.12
Through his experiences he had become a man with a mission.
He wanted to be an entrepreneur like Patterson with a share in
his company's profits. Having been fired by Patterson, he wanted
his business to become even bigger than NCR. Although Watson
was convinced that the criminal charges against him would not
stand, he was remorseful about many of his past actions. He vowed
that he would never again take any business action that could be
considered illegal or unfair-nor would he allow any of his asso­
ciates to do so. And alcohol, which had cost him his first job in
Buffalo, would never be permitted at business functions for which
he was responsible.13
The impact on Watson of his youthful incident with alcohol
and the later criminal indictment can be inferred from the strong
stand he took on related issues throughout the rest of his life.
Almost more revealing is the fact that his older son was unaware
of either of these events until long after he had joined IBM. Even
then, he learned about them from others. 14 The events apparently
had been too painful for his father to discuss.
In March 1915 the appeals court set aside the earlier verdict
and granted a new trial. The decision indicated the government
would not be able to make the criminal charges stick. Watson was
jubilant. "This decision," he told a close friend, "has lifted a great
load off my shoulders. "15 Two days later the CTR board of direc­
tors elected him president of the company. When Watson had
been hired as general manager in May of the previous year, the
board had chosen not to appoint him president until he could
clear his name of criminal charges. With this accomplished, he
could now assert his leadership with full confidence.16

The Man

The Tabulating Machine Company had been the means by which

Hollerith proved the worth of his inventions. Its growth had been
driven by him, and it was also limited by him. Developing associ­
ates and working through them did not come easily for Hollerith.
Copyrighted Material
 Watson: A Man with a Mission 35

By contrast Watson was not trained as an engineer and had little

interest in personally creating new devices or ideas. His gift was
in recognizing and adopting the good ideas of others. He meas­
ured his own success by that of the organization he led, and
he knew that its success was dependent on the contribution of
others.
To drive home the importance of individual contributions,
Watson liked to stand before an easel at company meetings and
write a series of job titles, one under the other: "general man­
ager," "sales manager," "salesman," "factor y manager," "factor y
man," "office manager," and "office man." Then crossing out all
but the word "man" in each line, he would observe, "We have
different ideas, and different work, but when you come right
down to it, there is just one thing we have to deal with throughout
the whole organization-that is the MAN."
To be successful, Watson believed men had to be well trained
and well dressed, and feel good about themselves and their com­
pany. "You cannot be a success in any business without believing
that it is the greatest business in the world," he asserted. To help
create an image of success, he instituted an unwritten but rigidly
adhered-to dress code, held evangelical sales meetings, and had
songs written about the company's products, its employees, and
himself. He commissioned one of these songs to be written to the
stirring music of Hammerstein and Romberg's "Stouthearted
Men." The result was "Ever Onward," IBM's best-known company
song. "We're big, but bigger we will be," asserted the song that
hailed the corporation , its pioneers, and "that 'man of men,' our
sterling president. "17
Gifted as he was in dealing with people, Watson could not
achieve a good working relationship with Hollerith. Part of the
problem lay in the personal loss Hollerith suffered when he sold
his business. His sense of loss was intensified by Watson's very
different management style. Antagonism was particularly gener­
ated by Watson's emphasis on increasing the sales effort. Hol­
lerith believed his equipment should sell itself. In his view the
company should select from among potential customers only
those that would clearly benefit from the equipment and then
work closely with each customer to assure satisfaction.18 "It is a
good deal better never to have the machines put in than to have
them fail," he emphasized to Watson. 19
During the years after he sold his business, but prior to Wat­
son's arrival, Hollerith had vigorously pursued his inventive ac­
Copyrighted MateriSI
36 Building IBM

tivities on behalf of the company. His most important patent filed

during these years described a means for controlling tabulator
operations automatically using information stored in punched
cards. First used by the company's engineers in coupling a print­
ing mechanism with a tabulator, this patent represents an im­
portant first step in the evolution of punched-card systems
that performed complex information processing tasks and
mathematical computations long before the advent of electronic
computers.20
Whether primarily for reasons of health or out of personal
animosity, Hollerith substantially reduced his involvement with
the company soon after Watson arrived. Indeed Watson got so
little cooperation that he came to believe Hollerith wanted CTR
to fail so he could buy back his business at a low price.21 Despite
Hollerith's failure to help him, Watson was outwardly generous
in his praise of the man who had founded the company. ""ben
Hollerith died of heart failure in November 1929 at age sixty-nine,
Watson was prominent among those who attended the funera1.22
As Watson advanced in age and his business grew, he became
increasingly committed to the methods that had brought him
success. He was more aloof and less flexible. Regrettably it is his
later years that are best remembered, in part because no effort
was made to document his life until quite late. His official biog­
raphy, The Lengthening Shadow, was not published until more than
five years after he died.
Watson is of course destined to be remembered for dress code,
company songs, alcohol prohibitions, paternalistic policies,
authoritarian practices, and salesmanship. But as we shall learn
in later chapters, he should be remembered even more for his
ability to adapt to changing times. He successfully modified the
company's products and business methods to prosper during two
world wars and the Great Depression. He pioneered professional
careers for women in his industr y. And contrar y to popular my­
thology, he showed remarkable ingenuity and flexibility as he
readied the company for its post-World War II thrust into elec­
tronic computers.

Copyrighted Material
4
Building an Engineering Organization

It is ironic that Watson, who was hired for his sales and manage­
ment skills, found his greatest early challenge at CI'R to be in
product development. This is not surprising, however, since the
company consisted of three quasi-independent, high-technology
business units. To help meet the challenge, he asked the Guaranty
Trust Company for a $40,000 loan, $25,000 of which was to be
allocated to engineering studies. Three years earlier the Guaranty
Trust had provided $4 million to get CTR underway, but now it
refused to extend additional money until business improved. In
wha t Watson regarded as one of his finest sales pi tc hes , he re­
sponded, "Balance sheets reveal the past; this loan is for the
future." The Guaranty Trust reversed its decision and provided
the money.!

Competition Emerges

The genesis of Watson's problems in product development lay in

the lengthy battle between Hollerith and the Bureau of the Cen­
sus. Out of this battle had arisen the first competition in punched­
card equipment in the person of James Powers. According to the
records of the bureau, Powers was born in 1871 in Odessa, Russia,
and was educated at the Technical School of Odessa. He sub­
sequently helped fashion scientific instruments for the univer­
sity ' s physical laborator y before emigrating to the United States
in 1889. For eighteen years prior to j oining the Census Bureau
Machine Shop inJuly 1907, he held positions in which he worked
on improvements to a variety of devices including cash registers,
typewriters, and adding machines, and he had acquired a number
of patents. Although North hired him primarily to work On the
sorting machine, Powers made contributions in many areas. Re­
signing from the Bureau of the Census in April 1911, he estab-
Copyrighted Material
38 Building IBM

lished the Powers Accounting Machine Company to compete

directly with Hollerith.2
Powers's primar y competitive edge was equipment that pro­
vided a printed output. This made unnecessar y the manual re­
cording of numbers registered in dials or counters. Printed
output is a feature Hollerith had apparently considered for many
years, but failed to patent or implement. The reasons for this are
not clear. As early as 1899, for example, he had written to his
patent attorney proposing a printing feature on a tabulator.3 A
few years after the CTR merger, however, he is reported to have
"severely reprimanded" an engineer for attempting to attach a
printer to a tabulator.4 Hollerith's decision not to pursue the
printer attachment and Powers's decision to do so at the Bureau
of the Census gave Powers the competitive edge he needed to
challenge the better established Tabulating Machine Company.
By the time Watson took over CTR, Powers had developed a
range of machines for the commercial market. There were three
basic units: the "slide punch" based on a punch he had developed
earlier at the Bureau of the Census; a horizontal sorter that was
functionally identical to Hollerith's vertical sorter except that the
operator did not have to stoop or stretch to remove cards; and
the all-important printing tabulator. The Powers equipment used
the same size cards but had a mechanical rather than an electrical
hole-sensing method. Mechanical sensing was apparently faster
and more reliable as originally implemented. Powers also devised
a ver y attractive feature: linkages between machine parts were
localized in an interchangeable mechanical "connection box"
that made it possible to change rapidly from one application to
another.5
Changing applications on the original Hollerith tabulators had
required tedious rewiring-changing the interconnections
among electrical terminals. This burden was alleviated at first by
the provision of a plugboard whereby terminals were more easily
connected by plugwires. Later, to compete more effectively with
Powers, the plugboard was moved from the back of the machine
to a more accessible position at the front. The third improvement
was the introduction of the so-called automatic plugboard that
could be quickly removed from or inserted into the machine with
its plugwires in place. A customer might well have dozens of these
prewired plugboards stored on a rack, ready for use. Although
the plugboards were more than a foot on a side, and somewhat
cumbersome to handle, they were easier to insert and remove

Copyrighted Material
 Building an Engiru>ering Organization 39

than were the mechanical connection boxes of Powers. Plug­

boards also had the advantage that they could be wired or rewired
by the customer.6
Initially Watson lacked technical resources with which to meet
the challenge from Powers. Under Hollerith's direction the de­
sign and construction of machines had been carried out on a
contract basis by several firms. The Western Electric and Pratt and
Whitney companies had supplied most of the equipment used in
the 18 90 census. After that Hollerith sought out smaller compa­
nies with specialized skills, such as the Taft-Peirce Company of
Pawtucket, Rhode Island, and the Frederick Hart Manufactur­
ing Company in Poughkeepsie, New York. The Taft-Peirce Com­
pany had devised improved tabulating and sorting machines,
card-feeding mechanisms, and accumulators; the Poughkeepsie­
based company had primarily contributed designs for keyboard­
controlled punches.'
Even had Watson chosen to continue working with vendors, he
could not count on Hollerith to provide the necessary technical
guidance and coordination. After selling his business in 1911,
Hollerith increasingly found contentment with his family and his
farm in Virginia. A year before Watson was hired in May 1914,
Hollerith had resigned from the CTR board to avoid the incon­
venience of traveling to New York. Three months after Watson's
arrival, Hollerith asked to resign from the Tabulating Machine
Company's board as well. Watson reacted quickly. Citing the value
of Hollerith's advice, Watson urged him to remain. But Hollerith
was not deterred from his decision. His resignation was accepted
with regret in December 1914.8

Hiring Creative Engineers

Aside from Hollerith, the only senior technical person in the

Tabulating Machine Company was Eugene A. Ford. His labora­
tory was located in Uxbridge, Massachusetts, not far from the
Taft-Peirce plant with which he had worked closely in the past.
Within months of taking over CTR, Watson prevailed on Ford to
move to New York City to establish the company's first full-time
engineering department for the development of tabulating ma­
chines. Bringing three draftsmen with him, Ford soon had activi­
ties under way in a huge room Watson rented on the top floor of
a twelve-stor y building on Sixth Avenue and Thirty-first Street. A
model shop was equipped and ten model makers were hired.9
COpyrighted Material
40 Building IBM

Watson selected this location near Pennsylvania Station, hoping

that Hollerith would upon occasion make the train trip from
Washington to contribute to the development effort-but this
hope was never realized.
The first senior technical person to be recruited by Ford was
Clair D. Lake, a machine designer employed to design automo­
biles in the Locomobile plant in Bridgeport, Connecticut. Lake
had no previous experience with tabulating or time-recording
equipment when he joined CTR in 1915. His only education past
eighth grade was in manual training, but he was highly recom­
mended to Ford by a mutual friend. In the following year Watson
hired Fred M. Carroll, an experienced inventor who had pre­
viously worked on cash register designs for NCR.
Watson assigned both Lake and Carroll the task of inde­
pendently developing printing tabulators that would meet the
competition from Powers. Carroll pursued the design of a wholly
new tabulating machine that incorporated a novel rotary printer,
whereas Lake chose to build a more conventional printer that
could be attached to an existing tabulator. Watson selected Lake's
design to keep development and manufacturing costs low and to
reduce the time required to develop the product. Even so the
printing tabulator was not introduced until 1920, partly because
of difficulty in achieving good reliability and partly because of the
interruption of World War I. The military establishment was
apparently well satisfied with nonprinting tabulators because of
their simplicity and reliability.
During 1917 Watson took two significant actions concerning
the company's research and development activities. First, he
closed the Sixth Avenue shop in New York City and moved Lake
and Carroll with seven others to Endicott, New York, to facilitate
interaction with the Time Recording Division's development
group. This move also brought tabulating equipment develop­
ment and manufacturing together in Endicott, where manufac­
turing had been consolidated soon after CTR was established. 10
Second, he hired James W. Bryce as supervising engineer of the
Time Recording Division in Endicott. II
Of all the people hired by Watson, none contributed more to
the company's success than James Bryce. Born in New York City
in September 1880 to parents of Scottish descent, he has been
described as taciturn, humble, dry-humored, and intensely proud
of his ancestry. He attended the city's public schools, but spent
much of his spare time with his brother performing physics and
Copyrighted Material
 Building an Engineering Organization 41

Figure 4.1. James W. Bryce

Shown addressing an IBM group in 1939. James Bryce had been hired by
Wat�n as supervising engineer of the Time Recording Division in 1917 and
appointed chief engineer of the CTR in 1922. He served as Watson ' s pri­
mary technical adviser until shortly before his death in 1949. By then he had
acquired more than five hundred U.S. and foreign patents.

Copyrighted Material
42 Building IBM

chemistry experiments at home. When he was eighteen, he con­

tracted to do the bell wiring in a new apartment house.
Bryce studied mechanical engineering at the College of the
City of New York, but he quit after three years to take a position
as a draftsman and designer. In another three years he joined an
inventor with whom he designed what may have been the first
front-wheel-drive automobile. In 1904 he was employed by Harry
W. Goss, an engineer and patent designer, with whom he soon
formed a partnership. One year later the firm of Goss and Bryce
contracted to assign all patent rights for its time recording devel­
opments to the International Time Recording Company-an ac­
tivity that later brought Bryce to the attention of Watson who
determined to hire him.
Shortly after he joined CTR, Bryce initiated development of a
self-regulating electric time system. It was a major factor in the
wide adoption of the company's master-clock-controlled time sys­
tem. In an era of direct-current electric power systems-before
60-cycle alternating current was used to keep clocks synchro­
nized-this innovative system caused clocks and time recorders
throughout a building to register the same time. Prominently
mounted in the rooms and halls of m ajor buildings, IBM clocks
became a familiar sight throughout the country. Bryce also
helped develop a system that printed a record of when and by
whom any door in a building was locked or unlocked. The system
additionally recorded the time and the initials of watchmen as
they punched into each station while making their rounds.
Bryce was also interested in card-controlled machines, and he
filed for a number of patents on combinations of time recording
and punched-card equipment. Even before joining CTR, he had
invented a punched-hole-reading, elapsed-time recording ma­
chine. Through interferences declared by the U.S. Patent Office
between his patents and those of Hollerith and an independent
inventor, J. Royden Peirce, Bryce became familiar with Peirce's
patents and urged Watson to acquire rights to them.12

Obtaining Crucial Patents

The importance of patents had been brought forcibly to Watson's

attention several times before he hired Bryce. Within three
months of joining CTR, for example, he had been approached
by representatives of the Powers Accounting Machine Company
requesting a licensing agreement to use some of Hollerith's pat-
Copyrighted Material
 Building an Engineering Organization 43

ents. "You are in a position to put us in business or put us out of

business," Watson recalled them say ing.13
Watson, who was still facing criminal antitrust charges for his
activities at NCR, chose to set terms that would permit Powers to
remain in business. Nevertheless the licensing fees were substan­
tial. Set at 25 percent of gross machine rentals and 18 percent of
gross receipts from the sale of cards, these fees contributed to the
Powers company going into receivership during the recession of
1921-1922. The licensing agreement also limited Powers to the
use of Hollerith's patents for mechanical equipment only, thus
forestalling any shift by Powers from mechanical implementations
to electrical ones that proved to be far more versatile in the long
run.14
The CTR's upper hand in patents was severely challenged in
January 1917 when an interference was declared by the U.S.
Patent Office between patent applications of Hollerith and Pow­
ers. The issue involved the use of "stop cards" versus "automatic
group control." Hollerith had designed his tabulators to stop
whenever a stop card was encountered. These were inserted be­
tween card groups in the deck of cards by the operator to indicate
when selected subtotals should be copied and accumulators reset
to zero.
An alternative means, known as automatic group control, had
also been conceived by Hollerith and described in a patent ap­
plication filed in March 1914, two months before Watson joined
CTR. As described in the patent, two sets of wire brushes were
used to read each card twice. The second reading of a card
occurred at the same time as the first reading of the subsequent
card. Whenever the identification fields in adjacent cards failed
to match, the tabulator paused to signal that the last card of a
group had been read.15 Believing the stop-card method was too
cumbersome for a printing tabulator, Clair Lake proposed to use
Hollerith's automatic-control feature. The plan was jeopardized,
however, by an interference with a related patent of Powers.
Just as the interference proceedings appeared to be going in
Hollerith's favor, the patent examiner called attention to patents
of C. A. Tripp and J. R. Peirce that also had claims pertaining to
automatic control. Interferences with these patents were declared
in 1921, and the process of determining who had priority in the
claims progressed simultaneously with efforts to negotiate a set­
tlement. It was a serious situation. According to the patent attor­
�
neys retained by CT%� y�R�8���,ilrnce at the claims in the
44 Building IBM

Tripp application ... to reach the conclusion that whoever pre­

vailed in these interferences would be in possession of the claims
which would completely dominate the automatic printing tabula­
tor field. "16
In 1921 a license agreement with royalty payments based on
usage was negotiated with Tripp. Problems in accurately deter­
mining royalties led to a new contract in Januar y 1924 in which
CTR made a one-time payment of $212,500 plus some interest.
The payment corresponded to nearly 10 percent of the com­
pany's earnings or 2 percent of its gross revenue for the year. Early
in 1922 Watson also obtained Peirce's patents. Simultaneously he
acquired the laboratory and services of Peirce and some of his
co-workers, who by then had been designing and building inno­
vative punched-card machines for the Metropolitan Life Insur­
ance Company. The price paid for Peirce's company and patents
was $261, 000-even more than was paid to TrippP
Also in 1922 the patent licensing agreement with the Powers
company was modified to reduce by half the fees paid to CTR.
This helped return the Powers company to profitability. Small
royalties were better for CTR than none at all, which would have
resulted had the Powers company been forced out of business.
Furthermore some competition was necessary to forestall anti­
trust action by the government.18

A Drop in Revenue

The decline in patent licensing revenue and the huge price paid
for Peirce's company and patent,> came at a difficult time for the
company. Its three major divisions had projected continuing reve­
nue increases for 1921, but a severe (albeit brief) economic
recession had caused a downturn in all three. Particularly hard
hit was the Computing Scales Division. Having projected a dou­
bling of sales during the year, its revenues actually dropped by
half. A strike at the Dayton plant was a contributing cause, but
severe competition and a less-than-satisfactory product line were
primary factors. The only part of the business whose revenues
held reasonably steady was tabulating machines, largely because
its equipment was rented rather than sold. To offset the huge
companywide drop in revenue from $14 million to $9 million,
Watson cut his own salary and that of other executives by 10
percent. A similar reduction was made in the wages of factory

Copyrighted Material
 Building an Engineering Organization 45

workers. Many employees were laid off, particularly those with the
greatest skills and highest wages. 19
Seeking ways to avoid such a catastrophe in the future, Watson
made particular note of the stabilizing benefits of renting rather
than selling products. He had always believed tabulating equip­
ment had the greatest potential, but his emphasis on that part of
the business was now reaffirmed. To overcome problems of poor
engineering in the scales product line and to assure top-quality
engineering in tabulating equipment, he appointed James Bryce
of the Time Recording Division as chief engineer for all of CTR.
Watson also created a Future Demands Committee (discussed in
chapter 9) to provide Bryce with guidance in the selection of
product development objectives.

Managing Engineers

"Mr. James W. Bryce has been appointed Chief Engineer of the

Computing-Tabulating-Recording Company," Watson advised his
key executives in June 1922. "He will be in direct control of all
phases of engineering and development work, reporting directly
to me.» Bryce's office was moved from Endicott to CTR he adq uar­
ters at 50 Broad Street, New York City. At the same time, the
Peirce facilities at 37 East 25th Street were enlarged to provide
for expanded engineering activity and a research laborator y. Car­
roll and his group were moved from Endicott into these facilities
so that he could work more closely with Peirce and Bryce on
longer-term research and development projects.20 Convinced that
Hollerith's equipment provided the best basis for company
growth, Watson spent more than ten times as much on high-risk
development projects for it as he did for the other product lines.21
Appointment of a chief engineer was intended to supplement,
rather than replace, Watson's own involvement with the com­
pany's inventors. Excerpts of correspondence between Ford and
Watson in February 1924 typify that involvement. After assuring
Watson that all product requirements were "perfectly satisfied" by
his horizontal sorter, Ford asserted: "I have been perfecting the
mechanical details relating to the new rotary card feeding, the
brush shifting mechanism, and the electrical connections and
now have it practically just as I should like to see it manufac­
tured . . . . I think applications for patents covering the new sorter
and alternative constructions should be immediately made. The

Copyrighted Material
46 Building IBM

fundamental principle of direct sorting, without the intervention

of tripping mechanism that all sorters have heretofore used, can
be applied to a mechanical machine, and the Powers Company
should be forestalled along these lines. I should like the pleasure
of demonstrating the new sorter to you and your Board of
Directors. "22
Watson responded promptly: "It is a great pleasure for me to
congratulate you upon the outcome of your experiments and
development'> in the building of a horizontal sorter.I have heard
something of it from Mr. Bryce and am looking forward with a
great deal of anticipation to seeing the exhibition of the com­
pleted models now under construction.... The matter of patent­
ing your new machine has been taken care of as the work
progressed and you may be sure that it will all be covered. "23
Some two years later, after the sorter had demonstrated its
utility and reliability in customer use, Ford received a $5,000
bonus from the board of directors. In expressing his appreciation
to Watson, he wrote: "I thank you still more and am made happier
by the kind expression of approval of my efforts in behalf of our
Company. At times I have felt discouraged when my recommen­
dations have been turned down and opinions overruled but now
that I know you have not looked upon my defects as evidence of
incompetence, 1 shall strive harder tojustify your good opinion. "24
Ford's reference to "opinions overruled" and his intent to strive
harder to justify Watson's good opinion give a glimpse of the
authority and influence Watson wielded. As a manager of the
Endicott laborator y observed some years later, "Many, many
times, the engineers have been reasonably satisfied with a ma­
chine development, and it is only because of the dissatisfaction
and the criticism which Mr. Watson has presented that the engi­
neers renewed their efforts and came out with something far
superior. "25
Handwritten letters to Watson expressing appreciation for sal­
ary increases or other favors were common. "1 am proud to have
some part in this important work and glad to remember that this
could not have been, had you not some confidence in my fitness
for the job," Carroll wrote Watson in April 1925. "It is a matter
of pride and ambition with me to make this undertaking a great
success, but it would not be human to under-value that part of
your kind letter which refers to an increase in salary. Coming as
it does at a time when our expenses will be much greater, Mrs.
Carroll and I doubly appreciate the favor. "26
Copyrighted Material
 Building an Engineering Organization 47

So it was that Watson involved himself in the company's prod­

uct development activities from his first days at CTR. He hired
and promoted key technical people and took an interest in their
assignm�nts. As an experienced salesman, he was sensitive to the
needs of customers, and he became adept at expressing these
needs in terms inventors understood. Often he gave similar tasks
to two or three inventors and then selected the result he liked
best.

Internal Competition

The first of a long series of such competitive assignments was the

previously mentioned one given to Lake and Carroll to develop
a printing tabulator.27 Mter Watson selected Lake's approach,
Carroll turned his attention to other challenges. His most impor­
tant contribution was the rotary punched-card press. More com­
monly called the Carroll Press, it manufactured punched cards
in one continuous process involving printing, cutting, and corner
trimming. One young engineer recalls how impressive it was "for
those of us who were working for sixteen dollars a week to see a
machine printing a thousand cards a minute that sold for a dollar
a thousand. In sixteen minutes it made as much money for the
corporation as I earned in a week. "28
In the case of the printing tabulator, Lake and Carroll were
aware of each other's assignment, but competing assignments
were frequently made by Watson without informing the inventors.
This policy increased the pressure to excel and the sense of
urgency. Inventors never knew whether Watson might have as­
signed the same product objectives to someone else. Intense
rivalry developed, and inventors jealously guarded their work.
Those who worked for an inventor were careful not to be seen
talking with people from other departments lest they be sus­
pected of passing secrets. Aside from the obvious disadvantage of
not sharing ideas internally, the system of internal competition
seems to have served the company well.
Egged on by Watson, an intense rivalry developed between
Lake's group in Endicott and Peirce's group in New York City,
following the latter's move in 1927 to an expanded facility on
Varick Street. Both groups were attempting to respond to Wat­
son's request for a way to store more information on each
punched card. This was required by many accounting applica­
tions. A forty-five-column card was then standard for IBM prod-
Copyrighted Material
48 BUilding IBM

ucts, as well as for those of Remington Rand, Inc., which had

acquired the Powers Accounting Machine Company in 1927.
Working in secrecy from each other, the Lake and Peirce
groups took different approaches. To enhance the value of their
solutions. each group intended to provide additional functions.
Peirce was placing his emphasis on developing machines capable
of handling alphabetic as well as numeric information. His ap­
proach was to increase storage capacity by using a multihole.
combinational code for each character, while continuing to use
the standard card format with hole locations arranged in an array
of 45 columns by 1 0 rows. With the proposed multihole (binary)
code, 6 adjacent hole locations in each column provided 64
possible combinations-enough to store any one of the 26 letters
of the alphabet or of the 10 digits and have 28 additional combi­
nations for special symbols. The remainin g 4 locations for
pun ched holes per column provided 16 possible combinations.
This was 6 more than needed to store any one of the 10 digits.
By changing from the standard code, in which the location of
1 hole per column provided storage for only 1 digit, Peirce's
group had provided the ability to store twice as many characters,
half of which could be alphanumeric.
Meanwhile Lake's group was taking a more standard engineer­
ing approach. His group chose to use smaller holes so that each
card could accommodate more columns. This required the de­
velopment of new punches and readers, but (unlike Peirce's
approach) the coding of the holes and the underlying numerical
operations were not affected.
Mter reviewing the competitive designs, Br yce selected Lake's
simpler appr oach. It could be implemented more rapidly and
would retain the long-established machine timing methods for
reading numbers on cards. Another factor in the decision was
possible patent interference with Powers relating to the multihole
card punch. To Peirce's utter amazement and disappointment,
the ability to store alphabetic information in addition to numbers
was judged to have little value to customers.29

The IBM Card

In 1928 IBM introduced the Lake-developed 8O-column card

system. It offered nearly twice the storage capacity per card as the
previous system. The most obvious feature of the new card was its
rectangular holes. Rectan g ular instead of round holes had been
Copyrighted Material
 Building an Engineering Organization 49

w M 0 .. �, �. . "
'" 0 12 : 0 12 .... ,.. ,. .- '" ., '"

10 15 I. " : I ",
�.
..
"
.. i� <0,
-,
M'
,.
�
..
. ..
"

Ch 20 21 '" 30:
,.
MO I. 2 N .,
"
.. '"
"
..
,.
.. .
..
::;

6 Jp 15 40 45 'I,."
50: 1� AJit W
MI : ' ..
15 3 F ' "'.- � -, -
:
In 55 "
1� "
'co\:
60 65 70 : on �, �, w,.
Wd • 10 I. : • 10 _. � " " �

75 80 85 90 95- Un 17. r 5 13 i: ,,-

..
.. ,
"
"
'"
-..
"
, ..
og.."
En 01< 0 • : 17 ,,- _.,
Un 15 0 US Un En US ... En " "'

�
01 '''' 5 :01 12 NG 20' "''-
3 G, 5, G, S, '"
,

No ,
,
0 13 • wm ...
fw S. GE W. S. CE W. .. OH

tJ
I I. Po: 5 ""
S, ,.. CF Hu N- CF Hu ..

2
,
15 AI • Po D. F, D' F,

.. 3 I. 10 Un , 10 01 ". B. 01 ". Bo S,

0123456789 AS CDEFGH JJ KUANOPQ RSTUVVIXYZ

I j 'j"
I

J
O T ON
1 H G P SI I
""'-- C IN
P UN
II
1 2
1 1 1 1�

J
PUNCHING POSITION
1

�'�,������!���� . ��\,::�:::���Gu:��;:)�!��I)��:���!
�; I
I ,. I _
, 1111111111111 111111111111111111111111.11111
"
� 1:: 222111222222 22 222222222222 222212212222222222
13 ] J n)JlllllJllln II 1 llJllllllllJlllll13111131111Jl

,n II 44444444444444444444444444444.

Sji5SJ jSS5SSSSSSS!iSSSSS 555 55i5 5:; 555555555555 555555 555555

I
G�':;666r. 6tii66�6S�"6E6666S 066" 1i666 666661i6666666lilililililililililili666661i

11111111 1111111111111111111 111111 11111 111111111111111111111111111111

aeSdil8888 888Sa.ae88a88.aSS.88 S8888a. 8B8888 888ge3888888888888888888888888

I : � � : � � � � � �,,?::.�.:,! � � �! �, � �J�' �\!�,! �! � �'Il!! � �,! !!�'I!! � !�,�,���?}�?���,�,�,��,}:�;�,;,�!�,� ,�?'!? ��,�,! !�,�. �!

Figure 4.2. Punched Cards

Top: The punched card used in the U.S. Census of 1900 (unlike the card
used in 1890) had information printed on the surface to facilitate visual read­
ing of punched information. In both cases punched holes were round, and
the card size was 65/8 inches by 31/4 inches. Bottom: The IBM 80-column
card, with rectangular-shaped holes, replaced the then-standard 45-column
cards in 1928. By the early 1930s, the two additional rows shown here were
introduced to facilitate alphabetic coding. Holes in these rows were punched
using the 11 th and 12th punching positions of a 12-key punch. The 80-
column card and the earlier 45-column card were both 73/8 inches by 31/4
inches.

Copyrighted Material
50 Building IBM

proposed by Lake and Br yce because they provided better space

utilization for wire brushes that electrically detected the holes.
Cards with rectangular holes also proved to be mechanically
stronger, making the new design patentable. Perhaps most impor­
tant, the rectangular holes gave punched cards a distinctive ap­
pearance that could not be copied by others. Recognizing the
marketing value of this patented appearance, Watson began pro­
moting it as the "IBM card." It was not just a punched card
anymore.
To help customers convert their records to the new 80-column
cards, a "reproducer" was offered that punched 80-column cards
from existing decks of 45-column cards. T he value of being able
to reproduce cards became evident through this experience, and
reproducers to copy from either 80-column or 45-column cards
became standard products.30
Remington Rand countered IBM's 80-column card with a 90-
column card in 1930. Like the IBM card it was the same size and
shape as the 45-column card previously used by both companies.
T he 90-column card used a multihole code that was similar, but
not identical, to the one Peirce had proposed at IBM. Physically
the card had 45 columns and 12 rows, but it was logically divided
into two sets of 45 columns of 6 rows each , one set above the
other. Using one-to-two holes to code a digit and two-to-three
holes to code an alphabetic character, there were more than
enough possible combinations to provide for full alphanumeric
coding . Remington Rand's 90-column card thus provided the
customer with more storage capacity for numbers than the IBM
card, and it offered alphabetic capability as well.31
Peirce's engineering group at the IBM Varick Street laboratory
had already switched its efforts from developing its own multi­
hole-coded card to finding an acceptable way to provide alpha­
betic capability on the newly introduced 80-column card. But
when Remington Rand's announcement was made, the pressure
from IBM's sales organization for alphabetic capability was so
great that the company hastily introduced a machine from
Peirce's group that provided only a subset of the full alphabet. It
used two punched holes for each letter, one of which was in one
of the two rows above the ten rows used for numbers. Known as
the ATB (Alphabetic Tabulating Model B), it was offered to
customers beginning in 1931. Model C, offered soon after, pro­
vided full alphanumeric capability. A hole punched in one of the
top three rows (X, Y, and zero) combined with a hole in one of
Copyrighted Material
 Building an Engineering Organization 51

the bottom nine rows was used to designate alphabetic characters.

This form of alphanumeric coding survived to the end of the
punched-card era.32
The 80-column and 90-column cards were roughly equivalent
in their capabilities. The slightly greater storage capacity of the
90-column card was offset by the greater ease with which opera­
tors could visually read information on the 80-column card-es­
pecially the numbers, which were still represented by a single hole
for each digit. Using today's terminology. one would say the IBM
card was more user friendly.
Competition between these cards was influenced by a new
factor: they were incompatible. No longer could cards prepared
by an IBM system be processed by a Remington Rand system, or
vice versa. Not only did each installation now have to chose one
system or the other. but communication between two installations
was possible only if they used the same system. The benefits of
compatibility among installations within an organization-and
even among those in different organizations-led inexorably to
greater domination of the market by IBM.
Only a major improvement in machine function and reliability
and in the service provided by Remin g to n Rand could have saved
it from continued decline. But this did not happen: IBM main­
tained its generally superior reputation for product reliability and
service; in advanced function, where its conservatively structured
product line was often behind the competition, IBM pushed
forward with numerous product enhancements.
Significant among new products introduced in 1928 was the
Type IV tabulator. Designed by Clair Lake's group. it offered
direct subtraction capability. Negative numbers were identified by
a hole punched in an appropriate column in row 11 of the
8O-column card. When this so-called "X-punch " was detected, the
tabulator carried out the indicated subtraction automatically. Pre­
viously negative numbers, called "credits, " and positive ones,
called "debits," were punched in separate card fields and sepa­
rately summed by the tabulator. The sums were then manually
subtracted. (Alternatively, the complements of negative numbers
were manually determined and then key-punched in the same
field as the positive numbers so as to achieve subtraction through
addition of the complements.)
The company's first product to perform multiplication, the
Type 600 multiplying punch, was added to the line in 1931. Based

Copyrighted Material
52 Building IBM

on pioneering work of Bryce and others, it could take two factors

from a single punched card, multiply them, and punch the result
into a blank field of the same card. Two years later it was super­
seded by the Type 601 electric multiplier that, among other
improvements, could perform "cross footing,» a term used by
bookkeepers for summing a table of numbers by row rather than
by column.33

Copyrighted Material
5
Responding to the Great Depression

The largest ever one-day drop in New York stock market prices
occurred on 24 October 1929, a day that became known as Black
Thursday. Following years of unbridled speculation and business
excesses, the drop triggered-but was not the root cause of-the
most serious business recession ever experienced in the United
States. By 1932 one-third of the labor force was unemployed, and
the value of shares on the New York Stock Exchange had dropped
to about one-tenth of the peak value in 1929.
The depression was disastrous for most businesses. The reve­
nues of NCR, for example, dropped from $58 million in 1929 to
$16 million in 1932} The impact was less severe for IBM because
of increased marketing efforts and because its equipment was
rented rather than sold. Unless returned, equipment on rental
continued to provide revenue. The company's revenues dropped
from $18 million in 1929 to $17 million in 1932-less than 10
percent (appendix A).
Taking advantage of the relatively strong financial condition of
IBM during the depression, Watson consolidated its market posi­
tion and strengthened its technical capabilities. Numerous prod­
uct improvements were made, culminating in the Type 285
Numeric Printing Tabulator introduced in 1933 and the Type 405
Alphabetic Accounting Machine introduced in 1934. Both could
tabulate 150 cards per minute or tabulate and print from cards
at 80 cards per minute. The more expensive Type 405 provided
greater versatility in its control features and could print alpha­
betic as well as numeric information. It was the company's flag­
ship product until after World War 11.2 Even more important than
specific product improvements, however, was the strengthening
of research, manufacturing, and employee education activities.

Copyrighted Material
54 Building IBM

PRINT
UNIT

CARD
STACKER

READING
SWITCHES TABLE

SUMMARY
PUNCH CABLE
CONTROL CONNECTOR
PANEL

Figure 5.1. Type 405 Alphabetic Accounting Machine

Top: Arrows point to key functional elements of the Type 405 Alphabetic Ac­
co u n ting Machine. which offered considerable functional versatility through
its removable (automatic) plugboard. labeled "control panel." Bottom: Pic­
tured here when it was in troduced in 1934 is the Type 405 Alphabetic Ac­
counting Machine (center) with a key punch (left) and a horizontal sorter
(right) .

Copyrighted Material
 &sponding to the Great Depression 55

Coupling Engineering to Manufacturing

In 1933 Watson moved most of IBM's research and engineering

activities to Endicott, New York, where the company's primary
manufacturing operations were already located. A new building
known as the North Street Laborator y was constructed to house
the engineering activities of the manufacturing and product de­
velopment organizations. Close proximity of these two activities
permitted the sharing of common resources and was expected to
facilitate the transfer of new products from development to
manufacturing.
The reception room of the North Street Laboratory was known
as the Hall of Products. It shared the first floor with the manu­
facturing engineering department. On the next two floors, seven
of Watson's highest-ranking inventors-among them Carroll,
Ford, and Lake-occupied seven of the eight corner offices. Each
had his own cadre of draftsmen and technical support people.
Elsewhere in the building was a variety of common service func­
tions, such as the blueprint room, machine shop, tool crib, elec­
trical shop, industrial design shop, and the patent attorney's
office. A technical library was located in the penthouse. Unusual
for the time, the building was fully air-conditioned. 3
At the ground-breaking ceremony for the new laboratory in
July 1932, Watson asserted, ''We have realized from experience
that the future of our business largely depends on the efforts,
brain, and ability of our engineering department.. ..That is why
today we are breaking ground for this new building, which will
be devoted entirely to research and engineering work. "4
Even with manufacturing and engineering consolidated in En­
dicott, the number of IBM employees was small compared to that
of the major employer in town, the Endicott:Johnson Shoe Com­
pany, whose employees were mostly first- or second-generation
immigrants from eastern Europe. To help them adapt to the new
country, their employer provided them with company housing, a
hospital, a cafeteria, and a variety of recreational facilities. Initially
IBM provided none of these, but its employees enjoyed the use
of facilities offered to city residents for a small fee by the shoe
company.5
Convinced that the economy would soon recover, Watson pro­
duced tabulating equipment and stored it in warehouses during
the depression. It was a policy that won him high marks among

Copyrighted Material
56 Building IBM

Figure 5.2.North Street Laboratory

Ground was broken for the North Street Laboratory (top) in July 1932. One
year later it was occupied by about 250 employees, who carried out most of
IBM's research and development activities. Photographed in 1937 is the elec­
trical laboratory (bottom) within the North Street Laboratory.

Copyrighted Material
 &sponding to the Great Depression 57

his employees. Broader vindication of his policy came with pas­

sage of the Social Security Act in 1935. The record-keeping and
data-processing requirements of this law created a huge demand
for punched-card equipment. Many businesses as well as govern­
ment agencies installed equipment just to carry out tasks man­
dated by the law, and IBM machines built and stored during the
depression were brought out to help satisfy the needs.6 All Social
Security checks for unemployed and retired workers were printed
on IBM punched cards and mailed to the recipients. For those
who received these benefit checks, the punched-card era had
happily arrived. For IBM the depression was over.

Emphasis on Education

In May 1932 Watson formally established an Education Depart­

ment to manage the company's many educational activities for
employees and customers. An IBM schoolhouse devoted entirely
to the education of employees was built in Endicott in 1933. For
the first time, the company offered a formal engineering educa­
tion program. The first three engineers were graduated in 1934.7
Previously, newly hired technical employees had been assigned
to experienced engineers for a training period that was similar to
an apprenticeship. Several of the company's leading inventors got
their start this way, but as the company grew and its products
became more diverse and complex, the apprenticeship form of
training was judged to be inadequate for engineers. As the new
program evolved, the curriculum included electrical and me­
chanical operating principles of the company's machines, tempo­
rar y assignments in development laboratories or factories, field
assignments to service machines in use, experience in sales appli­
cations, and lectures by the company's inventors describing the
underlying concepts of the machines. The entire course lasted
from six to fifteen months.s
The expanded education program for the manufacturing or­
ganization offered courses in blueprint reading, shop mathemat­
ics, measuring instruments and techniques, modern milling and
grinding practice, and electricity. Courses were typically sched­
uled one evening a week, after work, and lasted for thirty-six
weeks. Taking courses was not mandatory. Nevertheless in 1937,
approximately half of the 3,200 factor y employees in Endicott
were enrolled in fifty-two classes, studying twenty-four subjects.

Copyrighted Material
58 Building IBM

Particularly important was the tool-making apprentice pro­

gram. By the mid-1930s the company was hiring approximately
fifteen candidates for this program each year, selecting high
school graduates with high scholastic standings in such subjects
as drawing, shop practice, science, and mathematics. "Character,
personality and evidences of ambition" were all used as the basis
for selection. A new apprentice was placed on probation for six
months. If he demonstrated good mechanical ability, he was
enrolled in the course. The first year and a half was spent in the
apprentice machine shop learning the fundamentals of tool mak­
ing. This training was supplemented by practical experience in
the heat-treatment department, the physical laboratory, and the
machine repair department. The apprentice typically spent an­
other year and a half in the tool room before completing his
apprenticeship.
There was also a "supervisor's school" that all factory supervi­
sors and executives were required to attend. It was divided into
three sections to serve the special needs of executives, supervi­
sors, and foremen from the various departments. Deeply commit­
ted to growth from within the company, Watson believed there
was "no saturation point in education."g
In 1935 Watson established an educational program that must
have surprised even him. It began with a speech he presented
before the Institute of Women's Professional Relations. When
members of the audience said they were distressed by the limited
opportunities for women, Watson expressed surprise. "I would be
willing to hire twenty women immediately," he advised them.
Inundated by applications, Watson challenged his personnel de­
partment to create an appropriate program.
The resulting plan was to train women in the operation of IBM
equipment so they could assist salesmen in analyzing the needs
of potential customers and in developing a system solution. Once
equipment was installed, they were to teach the customer's em­
ployees how to use it. The first group of trainees consisted of
twenty-five women selected from top colleges such as Bryn Mawr,
Cornell, Smith, and Wellesley. It was an emotional and intellectu­
ally stimulating experience. They were presented with flowers and
attended a welcoming formal dinner dance hosted by Mr. and
Mrs. Watson during the first week. More social events followed,
and the young women delighted in writing and singing their own
songs about IBM and its leaders. The women were trained at the

Copyrighted Material
 Responding to the Great Depression 59

Figure 5.3. The FU'St Systems Service Women

This graduation pho tograph taken in front of the IBM schoolhouse in Endi­
.

cot t in 1935. shows the twenty-five coll ege graduates who attended the first
class for system s service women together wi th their three instructo rs .

Endicott schoolhouse for three months, taking essentially the

same classes as the seventy-five male sales trainees. Mter gradu­
ation they were designated systems service women and assigned
to branch offices.
The program provided IBM with an invaluable source of talent
for its expanding business, helped it meet wartime challenges
when many men were drafted into militar y service, and pio­
neered the entry of women into professional positions in the data
processing industry. In November 1943 Ruth M. Leach, a gradu­
ate of this program, became IBM's first female corporate vice
president. She had been manager of the Systems Service Depart­
ment since 1942.
In describing the opportunity to women in the first training
class, one vice president told them it represented a new concept
in business, "which emanated from the powerful brain and keen
intellect of our president, Mr. Watson." Another said, "I am sure
that you will meet this responsibility with energy and intelligence
and that you will make an outstanding success of this work,
thereby opening to a large number of college women, who will

Copyrighted Material
60 Building IBM

follow you, an important and productive field of endeavor which

did not exist before. "10

Research for Current Needs

Research in Endicott emphasized immediate product needs

rather than fundamental studies. The electrical laboratory, for
example, was primarily responsible for approving electrical fea­
tures of products under development. It also carried out research
for product improvements and devised test equipment. Consid­
erable research had been done in evaluating insulated wires to
find those most suited for electrical interconnections and for
winding electromagnets. Similarly the metallurgy and chemical
laboratories carried out studies of materials and chemicals
planned for use in new products or in production equipment and
processes, and research in manufacturing engineering was lim­
ited to improving processes such as drilling, grinding, turn­
ing, hobbing, die pressing, welding, heat-treating, plating, and
assembly. II
The paper-testing laborator y was particularly important. Its job
was to analyze incoming paper stock-or proposed new types of
stock-for the lucrative business of manufacturing and selling
punched cards. By 1937 IBM had 32 Carroll presses at work in
Endicott printing, cutting, and stacking 5 to 10 million standard
and special cards each day. The company derived less than 10
percent of its revenues, but more than 20 percent of its profits,
from the sale of these cards.
A significant problem was caused by occasional inclusions of
conductive material, such as carbon specks, in the card stock. A
conducting inclusion in a card could produce a false reading by
permitting electricity to flow through it almost as if the inclusion
were a punched hole. A unique testing method had been devised
to find these inclusions. As the roll of paper stock entered the
machine that cut it into strips, the sheet was passed between a
metal plate and wire brushes to which 220 volts was applied.
Whenever a conductive inclusion permitted current to pass
through the paper, a thyratron-tube detector activated an auto­
matic ink dauber that stained the defective portion of the paper
with a red dye. Mter the paper strips were run through the
Carroll presses, visual inspection was used to locate and reject
cards with red ink on the back sides.12

Copyrighted Material
 Responding to the Great Depression 61

0mlity and Productitm Ctmtrol

"Believing that tools are the basis of quality, IBM provides for
tooling as it plans its products," a trade publication reported in
the mid-1930s. Although IBM was not entirely alone in this ap­
proach, the publication indicated that its emphasis and methods
were well ahead of those in most other companies.13 Each part of
a new product was subjected to a manufacturability analysis dur­
ing the design stage to determine if changes were needed in
order "to conform to modern shop methods." The product itself
was subsequently "scrutinized by a tool analyzer" to determine
what manufacturing tooling was required. Prior to the depression
IBM purchased most of its manufacturing tools, but by the mid-
1930s it was making all of its special tooling. Only standard tools
such as drills, reamers, and milling cutters were purchased.14
To ensure that quality was built into each product, an inde­
pendent quality control d epar tm ent was established under the
factory comptroller. Its task was to inspect purchased parts and
materials, tools, subassemblies, and completed products to be
sure they met specifications. Emphasis was placed on inspection
of individual parts and subassemblies so as to avoid the large scrap
and rework costs associated with problems discovered after final
assembly. Although the statistical models now used for sample
testing were not then available, inspectors were authorized to
move simple parts directly to stock whenever "experience on
periodical checks" indicated inspection of each item was not
warranted.
The factory comptroller was also assigned responsibility for the
operation of a pioneering production planning and control sys­
tem that made extensive use of accounting and tabulating equip­
ment. The system included i nventory control and used
machine-written records and requisitions for all except raw ma­
terial and tool requisitions. Punched cards followed and recorded
the progress of all parts and subassemblies from inventory to final
product. 15
Standard punched-card equipment was used as much as possi­
ble, not only to support m anufac turing operations but also to
provide an applications research facility. New uses for punched­
card equipment in manufacturing could be tested and improved
before they were made available to customers. It was a fertile
source of new ideas because IBM's production control problems

Copyrighted Material
62 Building IBM

were particularly challenging. Each inventor had his own ma­

chine shop and had little incentive to use parts already in pro­
duction. Once Watson was shown a working model, he was likely
to want it placed in production so rapidly that there was in­
sufficient time to convert parts specified by the inventor to stan­
dard ones.
Each new product thus added many new parts to be cataloged
and tracked during production and throughout the product's life.
Parts ranged from mundane items such as studs, clips, screws, and
bolts to complex assemblies. Subsequent model releases might
employ a higher percentage of standard parts, but each release
added yet more parts to the overall inventory because of func­
tional enhancements and because both models now had to be
serviced in the field. The policy of renting machines and continu­
ally refurbishing and upgrading old machines exacerbated the
problem.I6

Methods Research

During the years following the stock market crash of 1929, IBM
salesmen needed considerable ingenuity just to forestall the re­
turn of rented equipment. A return to clerical processing of data
was attractive for many customers in an economy in which wage
rates and business volumes were decreasing. Nevertheless Watson
continued hiring salesmen. Many people of superior quality were
available. New salesmen were schooled in the company's products
and their applications before being sent into the field. Among
those hired in 1929 was Felix J. Wesley. Mter graduating from
sales school the following year, he went to work in New York City
as an assistant salesman. His primar y duty was to render technical
assistance to salesmen whose accounts were in jeopardy. "Only
through technical tricks was it possible to hold a customer's rental
at a minimum figure and yet accomplish the required tasks."l7
In September 1931 Wesley obtained a transfer to the newly
formed Methods Research Department at corporate headquar­
ters. The department's mission was to formalize the acquisition
and centralization of data-processing "methods know-how" and to
investigate field requirements for new applications and new ma­
chines. Such know-how had been created in a relatively unstruc­
tured manner over the years by individual salesmen, servicemen,
and customers. Often there were several different solutions for
similar problems. Electrical interconnections and plugboard con-
Copyrighted Material
 Responding to the Great Depression 63

trol of IBM's punched-card machines provided greater versatility

than was available in Remington Rand's mechanically controlled
machines and encouraged innovative data-processing methods.
Ingenious plugboard wiring arrangements were distributed by
the new department in a news bulletin called "Pointers." Most of
the wiring schemes were devised and contributed by branch office
people, but some were provided directly by customers. By collect­
ing and disseminating methods information and introducing
some standardization, the new department contributed sig­
nificantly to IBM's success in retaining its customers during the
depression. IS
By 1937 the growing business resulting from President Franklin
D.. Roosevelt's New Deal programs encouraged even greater em­
phasis on standardization and dissemination of know-how. To
accommodate this, the Methods Research Department was en­
larged. Its name was changed to Commercial Research and
Wesley became its manager. The department was also given re­
sponsibility for specifying desired machine functions, consult­
ing with engineers about new products, and writing product
manuals. 19
These responsibilities were not easy to fulfill, in part because
IBM's key engineers reported directly to Watson who thought of
them as his inventors. If a disagreement between an inventor and
Commercial Research was escalated to Watson, he was likely to
side with the inventor. Indeed there were cases in which an
inventor complained to Watson that members of the Commercial
Research Department were "bothering" him, and Watson showed
his support by ordering them to "let up" on the inventor.2o Never­
theless under Wesley's guidance, the department got many of its
ideas introduced in new products.21
Some ideas that originated in the department were quite so­
phisticated . Others were seemingly obvious, simple to implement,
and yet critical to product success-as in the case of a change
made to collators used in the Social Security system. The collator
was used to merge new information, punched in cards of a
"secondary" deck, into a "primary" deck of similar but older
information. Both decks were arranged in order of Social Security
number. The collator was designed to compare the Social Security
number of the next card in the secondary deck with the next
card in the primary deck to determine when a card from the
secondary deck should be inserted. Unfortunately, if a Social
Security number was incorr.ectlv plln!:herl, the information on the
c..;opyngntea MarenaT
64 Building IBM

new card would be assigned to the wrong person forever. There

was no provision for avoiding such an error by checking that the
names on the two cards matched.
The comparing mechanism that made the high, low, or equal
comparison needed for collating was designed to read only nu­
meric information. W hat was needed, ideally, was another com­
paring mechanism capable of handling alphabetic characters,
which were stored in punched cards by two holes (rather than
one hole) per column-one hole always being either in the O's
row or in one of the two rows above that. The methods expert
assigned to the problem quickly recognized that a comparison of
two names based solely on the holes punched in the rows nor­
mally used for numbers would be almost as good as a comparison
that used holes in the upper rows as well. This insight, combined
with knowledge of the mechanical and electrical operation of the
collator, permitted the methods expert to correct the problem
with only minor changes.22

Hiring Good People

Crucial to the success of Wesley's department was the quality of

its people. Their backgrounds were diverse, encompassing sales,
service, engineering, manufacturing, and ex-customers. "To­
gether this group worries each problem as a dog a bone-or
better perhaps , as a lot of dogs, one bone. Sometimes the ex-sales­
man is vanquished by the engineer, and sometimes the ex-cus­
tomer-service man has a bit of an awkward time at the hands of
the ex-customer," Wesley asserted. "But eventually all differences
are resolved in a joint opinion, which is the survival of the fittest
opinions. This, briefly, is the department's working platform-on
ever y problem that comes within its range, to seek out all the
evidence, to examine all possible opinions and to recommend,
among them, the fittest for survival." This description of his
department's activity was one Wesley thought "might pass muster
as a definition of'research."'23
Among those who joined Wesley's department in 1938 was
Stephen W. Dunwell. His uncle, who had developed the moving
assembly line for mass production of automobiles at the Ford
Motor Company, served as his father figure and mentor following
the death of his own father when he was ten. 24 Dunwell's IBM
career began in Endicott in 1933 on a work-study program with

Copyrighted Material
 Responding to the Greal Depression 65

Antioch College. During this program he alternately spent ten

weeks at college and ten weeks at work.
His first assignment was as an errand boy, obtaining missing
parts for the production groups in the plant. His second assign­
ment was in production control, and his third in wiring machines
in final assembly. So adept was he at the third task that he was
given "special engineering" work in which he began with a fully
assembled and tested machine, partially disassembled it, and then
wired in special features. Learning of the young man's talents, a
manager from the development laboratory offered him a full­
time engineering job. To accept the job, he would have to give
up returning to college for his senior year. "I thought on that
about ten seconds before saying, "Yes,''' Steve Dunwell recalls. "A
college degree didn't seem so important in those days."25
His job in the laborator y was to prepare retrofit kits with
instructions for installing new features to machines already in the
field. So long as he handled requests promptly, no one cared how
he spent his time. "I had access to everything in the laborator y,"
Dunwell recalls. "I could place an order to do anything and
nobody would question it. . . . So 1 did a certain amount of
experimenting ." As a result, he was selected to build a variety of
special devices, including a calculation control switch proposed
by Wallace J. Eckert at Columbia University (see chap. 6).
Having a strong interest in electronics, he was also given the
assignment to put together a display for Watson to show how
electronic devices were used, or being considered for use, in the
company's products. This was 1937. The only uses planned for
electronics involved photocell detection of marks or holes in
cards, and the only product already using a photocell was the
Social Security bill-feed machine. Each time a line entry was made
on a ledger card, a hole was punched to indicate that the line
had been used. The photocell detected the holes as the ledger
card advanced and thus determined when the first unused line
had been reached. This product, one proposed product, and
several electronic test devices used in production became Dun­
well's small display for Watson.26

Copyrighted Material
6
Support for Academic Research

Br yce and Watson became aware, during the late 1920s, that the
company's equipment was being used increasingly for scientific
calculations. Particularly noteworthy was the work of Leslie J.
Comrie, a New Zealander. who had studied mathematics and
astronomy at Cambridge University and taught college briefly in
the United States before joining the Nautical Almanac Office of
the Royal Naval College in England in 1926. By 1928 Comrie had
begun reporting on the use of Hollerith machines for the con­
struction of mathematical and astronomical tables. I
Mathematicians in the United States had also begun to use the
equipment to reduce the labor and errors in lengthy computa­
tions. A paper published in 1928 by a professor at Iowa State
College (now a university), for example, reported on the use of
Hollerith equipment for correlating various attributes. He had,
for example, correlated the college grades of students to their
high school grades, and in another study the "cost per hundred
pounds gain in swine as related to season, feeding system, average
daily gain, and initial weight."2

Educational Testing Studies

In the same year these papers were published, Professor Ben­

jamin D. Wood, head of the Bureau of Collegiate Educational
Research at Columbia University, wrote to leaders of ten office
machine companies, seeking help in creating an automatic means
for scoring and analyzing the results from large-scale testing
programs. Some did not respond. Others expressed interest but
offered no help. Only Watson showed sufficient interest to meet
with Wood.
The appointment was set for one hour. A secretar y, assigned to
advise Watson when tIubJ}�M�'Afc!tW�/mbsequently waved away
68 Building IBM

every hour on the hour. Excited by Wood's enthusiasm and ideas,

Watson kept him talking the entire afternoon. Soon after, he
appointed Wood an IBM consultant and ordered three truckloads
of IBM equipment sent to a location in the basement of Hamilton
Hall, where the Columbia University Statistical Bureau began
operations in June 1929.3 Watson donated the equipment because
he saw in Wood's request an opportunity to support an important
educational activity and, at the same time, to achieve a means for
keeping abreast of this new field.
Soon after this equipment was installed, Wood approached
Watson with two more requests. The first was for modified tabu­
lating equipment to carry out certain calculations more
efficiently. This request was quickly fulfilled when, in December
1929, Watson donated to the Columbia University Statistical Bu­
reau a one-of-a-kind difference tabulator that had been designed
for Wood by Bryce and others. It was equipped with ten accumu­
lators of ten digits each that could be operated in the standard
mode of conventional tabulators. But by changing the manner in
which the plugwires were connected so as to pair accumulators,
it was also possible to follow the normal card cycle with a special
cycle in which the content of one accumulator was summed into
an accumulator with which it was paired. Among the probems
devised to test the so-called Columbia Machine was calculating
the fifth powers of integers by the method of progressive totaling.4
Wood's second request was received by Watson just as enthusi­
astically as the first, but it turned out to be far more difficult to
fulfill. What Wood wanted was an automatic means "to reduce the
cost, inconvenience, and errors of scoring objective tests." It was
a need of which he had become acutely aware in the spring of
1928 while serving as technical adviser to a study conducted by
the Carnegie Foundation for the Advancement of Teaching. Dur­
ing this study 200,000 tests had been administered and graded by
hand. Wood joined others in a study of the test scoring problem,
finally concluding that "the design and construction of a satisfac­
tory scoring machine could be accomplished, if at all, only with
the aid of talented and experienced engineers and with the
facilities of a large and well-organized research laboratory." Upon
learning of the need, Watson agreed to make the solution of the
problem "a major project of the engineering staff and research
laboratory" at IBM.
Seven years passed before a satisfactory solution was found.
'The record of the succeeding long years of expensive research,
Copyrighted Material
 Suppurl fur Academic Research 69

of costly models laboriously designed, built, and abandoned, of

high hopes and discouraging disappointments, would make an
interesting story in itself," Benjamin Wood recalled, "but the
essential point of that story is that Mr. Watson and his engineers
carried on in the face of repeated disappointments and mounting
costs during a major economic depression until the problem was
solved in a way that goes far beyond the fondest hopes that any
of us entertained when we first appealed to Mr. Watson to under­
take the task."5

An Innovative Science Teacher

Wood failed to note that Watson would not have been able to
provide the solution as soon as he did-or perhaps at all-had it
not been for the ingenuity and persistence of Reynold B.
Johnson, a former high school science teacher and part-time
inventor from Ironwood, Michigan. Johnson had received his
bachelor's degree from the University of Minnesota in 1929.
Prior to joining IBM, Johnson had begun work on a similar
problem on his own. Noting that many textbooks printed in the
1920s had multiple choice tests in them, he decided to devise a
way to grade student answers more easily. He began by creating
an answer sheet from which students tore out tabs from the
partially perforated paper sheets. The sheets, with the tabs thus
removed at points corresponding to the selected answers, were
each placed over a board containing many wires that were used
to complete electrical circuits and turn on a light wherever a hole
appeared at the correct place in the paper. Since each of the
lights so energized drew equal current, a meter in the circuit read
directly the number of correct answers.
His first model was built with the help of two children who,
having been caught stealing a radio from the school, had been
assigned by the judge to assist Johnson after school. Johnson next
had an improved model built by a firm in Minneapolis. A com­
pany that published tests ordered twenty-five of these. The device
also received publicity in 1932 through an article written for the
local paper by a young woman who was then Johnson's teaching
associate and who subsequently became his wife. Her article was
picked up by the Associated Press, and thus came to the attention
of IBM. The idea was not pursued by the company, however, in
part because Watson had already assigned his own inventors the
task of building a tesLscori1l.2: m<\<;.hjn�· 1
copyngmea Malena
70 Building IBM

Meanwhile Johnson decided to attempt to detect pencil marks

on separate answer sheets using their electrical conductivity. He
had learned that pencil marks conducted electricity from the
then-common prank of short-circuiting the ignition on automo­
biles by making a pencil mark down the porous side of the
porcelain spark plug insulator. Using the conductivity of pencil
marks worked well enough that Johnson revised his equipment
and again attempted to sell his invention.
Failing to interest IBM, Remington Rand, or any other com­
pany-and being now jobless because of the depression-he and
his wife went on the road to sell silk hosiery. When they returned
home, they found a message from IBM asking them to come to
New York. A model he had shipped to the company had been
given to Benjamin Wood at Columbia University for evaluation.
Wood's evaluation was positive.
Johnson and his wife were provided $50 per week expenses to
stay in New York City during the six weeks in which he helped
complete the evaluation. The conclusion of the IBM engineers,
which the board of directors endorsed, was that Johnson's inven­
tion was incomplete. Among other problems, pencil marks ap­
peared to be unreliable conductors. Johnson was told that the
company would like to see the invention again if it could be made
more reliable. When Wood learned of this, he telephoned Wat­
son, who was on vacation, and urged him to reverse the decision.6
Watson did, and Johnson was hired two days later in September
1934. He was paid $4,000 for his patent and a salary of $4,000
per year.7
One year later, after considerable engineering work, the test
scoring machine was placed in experimental operation in the
IBM school and the public schools in Endicott, New York. Then
followed more trial uses as well as an evaluation by the Columbia
University Bureau of Collegiate Educational Research. According
to these evaluations, the machine could score tests about thirteen
times faster than trained clerks and with substantially fewer er­
rors . "I have received a very encouraging letter from Dr. Ben D.
Wood, of Columbia, as to the results accomplished by your ma­
chine," Watson wrote Johnson in March 1936, "and I wish to take
this opportunity to let you know that 1 am personally very much
interested in the work which you are carrying on."8
Students using the equipment were provided with soft lead
pencils and advised: "Blacken each narrow space between the pair
of [vertical] red lines numbered the same as the answers in the
Copyrighted Material
 Support for Academic Research 71

test that you consider correct. Make your mark as long as the red
lines. Use the special pencil only and move the point up and
down firmly two or three times." An important feature of the
machine was its ability to determine a student's score based either
on the number of correct answers or on the number of correct
answers minus the number of wrong answers-or minus any
fraction of the number of wrong answers. Lacking digital cir­
cuitry, Johnson had implemented this arithmetic function by
passing the electric signal from wrong answers through an electric
meter in the opposite direction to that from correct answers.9
T he system was announced for rent as the IBM Type 805
International Test Scoring Machine in 1937. Within five years,
rental revenues were approaching $100,000 per year and the sale
of supplies for it produced nearly twice that amount. Because this
was only about three-tenths of one percent of the company's
revenue, the Test Scoring Machine was generally considered to
be a minor product. But Watson's primary interest with this
product was to help education. In that sense it was very successful.
The test scoring method it introduced was widely adopted and is
still in use today, albeit with dramatically improved electronic
sensing and calculating means. \0
Johnson subsequently adapted the concept to more conven­
tional and profitable punched-card applications. But his most
significant contributions were achieved two decades later when
he and his colleagues created the world's first magnetic disk
storage devices for computers (see chap. 16).

Automatic Computation

Watson's support of Wood's educational research and statistical

bureau at Columbia University attracted many to the field of
automatic computation. Particularly important among these was
Wallace J. Eckert. With a bachelor's degree from Amherst in 1926,
he joined the Columbia University faculty as an assistant in as­
tronomy. Mter taking courses at Columbia, the University of
Chicago, and Yale, he received his doctorate from Yale in 1932.1l
Returning to Columbia, he became deeply involved with the
Statistical Bureau, using IBM punched-card equipment to per­
form mathematical computations in the field of astronomy.
Like Wood before him, Eckert recognized that his computa­
tional work could be speeded by specialized equipment. Watson
obliged by having ��erhlJn#MJ a Calculation Control
72 Building IBM

Switch designed by Eckert. Operational in 1936 this switch per­

mitted a dozen different modes of operation of an intercon­
nected system consisting of a Type 601 electric multiplier, a
printing tabulator, and a summary punch. In describing the
benefits of the control switch, Eckert noted that lengthy punched­
card computation processes were normally set up so that one
numerical operation could be done on many cards before shifting
to the next operation. But in numerical integration, for example,
"different arithmetical operations must be done in succession
for the first step before proceeding to the next." Solving this
type of problem was greatly speeded by the Calculation Con­
trol Switch, which permitted the standard electromechanical
punched-card equipment to perform a sequence of computation
steps between reading in a number and punching out the result.
To provide simplicity of construction, cards were moved man­
ually from one unit to another when necessary and the shaft
was rotated manually. Nevertheless this cam-sequenced calcula­
tor was a pioneering early step toward stored-program digital
computers.
As described by Wallace Eckert in his classic book, Punched Card
Methods in Scientific Computation: 'The calculation control switch
contains a row of electric contacts each of which is operated by
a rotating cam. The cam is a circular fiber disk which is notched
at various points around the circumference. A series of about
twenty of these disks are attached to a common shaft to form a
sort of player piano roll. When this roll is rotated from one
position to the next, the various contacts open and close accord­
ing to the notches in the disks. The circuits from the contacts are
used to operate the various control switches on the tabulator and
multiplier, and also a number of multicontact relays which effec­
tively change the wiring of the plugboards. "12
The computational facilities established by Eckert began as part
of the Columbia University Statistical Bureau and were later ad­
ministered as part of the Department of Astronomy. In 1937 these
facilities were named the Thomas J. Watson Astronomical Com­
puting Bureau and were made available to other astronomers
through the joint sponsorship of the Columbia University, IBM,
and the American Astronomical Society. 13
In the same year that IBM made its equipment widely available
to astronomers for mathematical computations, the Monroe Cal­
culating Machine Company rejected the opportunity to design

Copyrighted Material
 Support for Academic RPsearch 73

and build an electromechanical supercalculator envisioned by

Howard H. Aiken, a graduate student in physics at Harvard
University. The purpose of Aiken's proposal was to provide scien­
tists and mathematicians with an innovative sequence calculator
capable of a broad range of functions: performing basic arithme­
tic on positive and negative numbers in any desired sequence;
supplying and utilizing a wide variety of transcendental functions
(e.g., trigonometric, elliptic, Bessel, probability, etc.); providing
for storage of intermediate results; and carrying out a long series
of operations without operator intervention.14
The person Aiken approached was George C. Chase, the chief
engineer of the Monroe Company. A pioneer in the development
of key-driven calculating machines, Chase had begun his inven­
tive work more than ten years before joining the company in
1917. In 1932 the Monroe Company was awarded the John Price
Wetherill Medal by the Franklin Institute in recognition of its
implementation in mechanical desk-top calculators of "the four
basic rules of arithmetic: addition, subtraction, multiplication,
and division as attained through Mr. Chase's inventions." Accord­
ing to Chase, Aiken's proposal for the construction of an electro­
mechanical machine was quite general and not restricted to any
specific type of mechanism. He urged support of the project in
spite of the large anticipated costs because he believed the expe­
rience of developing such a sophisticated calculator would be
invaluable to the company's future.15
But top management of the Monroe Calculating Machine Com­
pany rejected the proposal. Chase then urged Aiken to contact
IBM through an intermediar y at the Harvard Business School.
Aiken did so, following which he wrote to Bryce in November
1937, saying, "Professor Brown of the Graduate School of Busi­
ness Administration tells me that he has discussed with you my
interest in automatic calculating machinery for use in computing
physical problems. If possible I should like to come to New York
early next week to call on you and discuss this matter in some
detail. "16
Br yce and Chase had much in common. Both were chief engi­
neers who had joined their companies in 1917, and both had
previously been established inventors. Bryce, however, enjoyed
greater management support and had no trouble convincing
Watson of the merits of the project.

Copyrighted Material
74 Building IBM

The Mark I Computer

At Bryce's suggestion, Aiken received education in the operation

of IBM equipment similar to that given to the company's custom­
ers and servicemenP Bryce believed such training was essential.
He wanted to be sure that Aiken had some understanding of the
capabilities and limitations of the devices available for implement­
ing his ideas. Following this Aiken made his first visit to IBM's
Endicott laborator y in February 1938. It was the beginning of a
cooperative effort that produced the IBM Automatic Sequence
Controlled Calculator (ASCC) , which became better known as the
Mark I computer, in part because Aiken followed it with Mark II,
III, and IV at Harvard. In February 1939 Bryce advised Lake,
whom he had put in charge of the project, that Watson had given
his approval for building the proposed machine at an estimated
cost of $100, 000. Its final cost was more than twice that amount.
Aiken spent the summers of 1939 and 1940 in Endicott, but he
was unable to spend much time on the project after assuming his
duties as a Naval Reserve officer in April 1941. The work of Lake
and others in Endicott was also interrupted by higher priority
projects of World War II. Nevertheless the Mark I was sufficiently
completed in January 1943 to solve practical mathematical prob­
lems that required multiplication, division, addition, subtraction,
and the computation of logarithms and antilogarithms. In Febru­
ary of the following year, the machine was completed and shipped
to Harvard. Men from Lake's department installed it, tested it,
and improved its operation for several months prior to its formal
presentation to Harvard in August 1944.18
The Mark I was 51 feet long , 8 feet high, contained 530 miles
of wire, and weighed five tons. It is the largest electromechanical
calculator ever built and the first machine able to exploit the
ideas proposed one hundred years earlier by the British mathe­
matician and inventor, Charles Babbage.19 Although Aiken and
Bryce were both aware of Babbage's earlier work, it appears to
have had minimal influence on their efforts.2o A key design fea­
ture of the Mark I contributed by Aiken was its ability to perform
many mathematical operations in a sequence defined by instruc­
tions punched in a paper tape. The "paper tape" was made of
IBM card stock to provide sufficient width and ruggedness to
supply instructions at the required speed. Computation with an
accuracy of up to 23 digits was provided by 72 accumulators, each

Copyrighted Material
 Support for Academic Research 75

Figure 6.1. Mark I and Its Creators

Clair D. Lake (top, center) and Howard H. Aiken (top, right) are pictured
in front of the Mark I, for which they served as chief engineer and chief ar­

chi tect respectively. Francis E. Hamilton (top, left), who provided daily su­
,

pervision for the proj ect, later led the development of the IBM 650, whi ch
became the most widely used computer of the 1950s. T he Mark I is shown
(bottom) soon after IBM presented it to Harvard in August 1944. At the
r igh t are the typewriters, card feeds, and card pu n ch To the left of these are
.

the paper-tape storage units, then the multiplying-dividing unit and counters
used in computing logarithmic and trigonometric functions, then storage
coun te rs and (at far left) dial switches.
,

Copyrighted Material
76 Building IBM

consisting of 24 electromechanical counter wheels, one of which

was used to represent the algebraic sign.
Design and construction of the Mark I was a major engineering
feat. Howard Aiken, who initiated the project, specified the func­
tions of the computer such as the algorithms by which complex
functions were computed. In today's parlance it would be said
that he served as chief architect. Clair Lake, IBM's best hands-on
engineer and most prolific inventor after Bryce, served as chief
engineer. Numerous inventions, which had been modified or
specifically devised for the machine, were incorporated in it.
These included improved relays and counters and means for
high-speed multiplication and division invented by Bryce, Dick­
inson, Lake, and others.21 Bryce, whom Aiken called "a very astute
inventor" and "a valuable senior adviser," played both of those
roles throughout the project. "If you started down a road that
didn't look very practical," Aiken said of Bryce, "he could put his
finger on it just about that quick. "22 With Aiken frequently inac­
cessible because of war-time activities, Bryce found himself mak­
ing many decisions that otherwise would have been handled by
Aiken.
IBM paid the entire cost of building the machine, approxi­
mately $200,000, and donated an additional $ 100,000 to Harvard
to cover costs of operation. All patents associated with the project
were assigned to IBM. The primary patent for the calculator was
filed in 1945 and issued in 1952 with 127 claims. The inventors
were listed as Clair D. Lake, Howard H. Aiken, and two associates
of Lake.23
Watson's support of academic research helped the company
establish itself in the new and vital field of automatic computa­
tion. It brought corporate leaders in contact with members of the
academic community who provided new insights and later helped
in the hiring of technical people for the post-World War II thrust
into electronics. By 1934 it had already led to the hiring of Rey
Johnson, who became one of the most creative research leaders
of the post-World War II era. It also attracted Wallace Eckert to
the field of automatic computation and paved the way for his
decision in 1946 to join IBM to establish the company's first
laboratory devoted to automatic computation and pure science.

Copyrighted Material
7
Research for Patents and Devices

Watson frequently intermixed the titles of engineer, scientist, and

inventor when referring to his key technical people. Although
this may reveal some confusion on his part, it also reflects the fact
that each of these terms properly applied to activities of his chief
engineer,james Bryce. Not only was Bryce a trained engineer and
an accomplished inventor; he had the inquiring mind of a good
scientist. He was able to perform the functions of chief scientist
and research director at the same time that he served as chief
engineer. There was no written job description for Bryce, but the
record speaks for itself. As chief engineer he devised products to
satisfy customer needs, and he made sure the company had rights
to all necessary patents. As chief scientist he was alert to techno­
logical advances and scientific discoveries that might affect the
company, and he initiated appropriate research projects.
Patents were Br yce's principal concern. If the company did not
own critical patents, it had to acquire rights to them or undertake
the costly and risky process of engineering around them. Bryce
worked closely with Watson to gain the rights to many patents.
Purchase of the Peirce company and its patents was the most
visible action, but numerous other patents were purchased. In­
cluded in the company's early patent portfolio for tabulating
equipment , for example, were fifty-nine patents that had been
issued during Watson's first ten years with CTR. Only twenty-six
of these had been issued to the company's employees. The re­
maining thirty-three had been purchased from others.}

A Prodigious Inventor

Increasing the number of patents created by the company's own

inventors was a major thrust of Bryce, who was himself a prodi-
gious inventor. Durin�his first twenty-five years with the company,
'e;opyngnted Matenat
78 Building IBM

he acq uired V. S. patents at an average rate of nine per year. Over

half were improvements to tab ulating equi pment . one-third to
time recording, and many of the remainder related to computing
scales.2 Bryce was honored as one of the ten " greatest l iving
inventors" by the V.S. Patent Office during its centennial celebra­
tion in 1936. By the time he died in 1949 at age sixty eight he
- ,

was credited with more than five hundred V.S. and foreign pat­
ents. Next to Bryce, the company s maJ or early inventors were
'

Lake, Carroll, and Ford. Peirce places wel l among these, although
many of his inventions were made before he joined CTR.3 (See
appendix B.)
Some of these patents related to inventions used in current
products and were a necessary part of product developm ent .

Others covered inventions for products not yet conceived and can
thus be thought of as resulting from the company s research '

activities. To stimulate new ideas, Bryce initiated the de velopment

of a "universal tabulator" with arithmetic and control features not
avail ab le in current or planned products. The resulting jumbo
Machine" was designed, constructed, and subjected to numerous
engineeri ng tests in End icott during the mid-1920s. It was not
released as a produ c t because its manufacturing costs were judged
to be too high. Many of its capab il ities , however, especially its
control features, were covered by patents issued to Bryce, Lake,
and their associates, and a num ber of these capabilities appeared
in later product... 4 .

A particularly significant aspect of Bryce's work was his self­

taught knowledge of elec trical phenomena and applica tio ns. At
IBM he continuall y sought ways to increase the use of electrical
devices and to enhanc e the benefits of electrical control of equip­
ment. The value of this em phasis was no t un iversally recognized,
as evidenced by Powers's dec ision to use mechanical im plemen ­

tations to compete with Hollerith. Even inside IBM, many inven­

tors favored mechanical solutions. Royden Peirce, for example,
who was brought into the comp any in 1922 primarily to obtain
his patents, strongly favored mechanical devices and resisted us­
ing electromechanical relays in any of his de signs until well into
the 1930s.5 Another key inventor who favored mechanical devices
was Frederick L. Fuller whom Watson hired from NCR to develop
,

a key op e rated p roof machine for use by banks. Th is hig h ly suc­

-

cessful product was strictly mechanical. Having worked on NCR

cash registers that used cylindrical tubing (one tube inside an­
other) to control their operation, Ful ler always referre d to the
CopYrighted Material
 RPseaTch jar Patents and Devices 79

Figure 7.1. Jumbo Machine

Work on the Jumbo Mac h i ne , also known as the Universal Tabulator, was in­
itiated by Bryce and carried out by Fred Carrol l , Clair Lake. and others be­
ginning about 1925. The back of the machin e is shown (top) with printing
mechanism at the u pper left. A gate is swung open ( bot tom ) to reveal the in­
ternal wiring. A key feature was the use of a large "universal" accumulator
that could be used to st ore the separate sums resulting from adding num­
bers in separate fields rather than allocating an individual acc u mul a tor to
each field. As was customary at IBM, the J umbo Machine was not revealed
outside the company even after many of its innovative features had been in­
corporated in products.

Copyrighted Material
80 Building IBM

electrical control wiring of IBM machines as the "tubing." He

could think of electrical control only in mechanical terms. 6
In 1926 Bryce formally organized and headed the IBM Patent
Department. Its functions were to help inventors obtain patents,
to maintain records of the company's patents, and to determine
if the company's products were likely to infringe on patents held
by others.7 Preferring to leave administrative matters to others,
Bryce had his former assistant appointed head of the department
in 1930 so he could "devote his entire time to invention and
development work."8 With Watson's enthusiastic support .. Bryce
began looking for an assistant to help him create patentable ideas.
In 1932 he hired A. Halsey Dickinson. Bryce had become ac­
quainted with Dickinson eleven years earlier in Endicott, quite by
chance, when he rented a garage for his car from the young man's
father.
Inspired by the well-publicized work of Charles Steinmetz at
General Electric, Dickinson had enrolled in the electrical engi­
neering department Steinmetz headed at Union College in
Schenectady, New York. There Dickinson obtained his bachelor's
degree. Expecting to enter the public utility field, he obtained a
master's degree from the Massachusetts Institute of Technology
in fuel and gas engineering in 1930. Unable to find work in his
chosen field during the depression, he contacted Bryce who hired
him. Dickinson was Bryce's first assistant in an activity that be­
came known as the Patent Development Department. Its function
was to create and patent new ideas.
Dickinson's job, for several years, was to begin with Bryce's
hastily drawn sketches and discussions and create working em­
bodiments that could be patented. "As time went on, I had some
of my own ideas," Dickinson recalls, "but it proved to be that most
of them had been done before. Nevertheless, Mr. Bryce gave very
considerable encouragement." According to Dickinson. Bryce's
management style was very informal even after the department
grew in size:

Bryce held no meetings. There were no deadli n e s , and his se c uring of

a certain number of patents had no bearing in his thinking whatsoever .

He plowed the field every day. By that I mean he came around to see
us or other people. Sometimes he would talk about the business. Some­
times he would talk about personal thin gs that he was doing . If he had
an idea or was thi n king about something, he would discuss it. Usually
he would make a sketch 01- a draw in g , which was sufficient to establish

Copyrighted Material
 Research fOT Patents and Devices 81

what he was thinking about or what he wanted to do. Since he explained

it to me, I would witness it.9

Much of the initial work pertained to enhancements for the

Type 601 multiplying machine, which had been introduced in
1933. Bryce invented numerous ways to handle more complex
arithmetic functions-for example, utilizing three factors rather
than just two. Bryce himself placed little emphasis on implement­
ing these inventions in products. His primary purpose was to
provide the company with patent protection-that is, freedom to
use any idea its engineers might need in future products.

Pioneering in Electronics

Bryce's interests carried him well beyond extensions of existing

technologies to entirely new ways of doing things. Even before he
joined CTR he had studied electron tube devices to see how they
might be applied in practice. At IBM he was continually on the
lookout for cost-effective uses for electronics. By 1936 serious
inventive activity was underway in Halsey Dickinson's laboratory
in Bryce's Patent Development Department. Describing this work
to Watson early in 1939, Bryce said, ''We have been carrying on
an investigation in connection with the development of comput­
ing devices which do not. employ the usual adding wheels, but
instead use electronic effects and employs tubes similar to those
used in radio work. Since these devices operate approximately
with the speed of light, it will be possible to compute thousands
of items per second. Patent work only and some few experiments
have been done so far, but it appears to look very promising for
certain fields of work where extreme speed is either desirable or
necessary. "10
In January 1940 Dickinson made what is believed to be the first
patent filing for an electronic computer circuit. The circuit was
capable of adding and subtracting multiple-digit numbers and
displaying the result on a cathode-ray tube. The next earliest
patent filings are believed to be two by NCR workers later that
same year and two by RCA workers more than two and a half
years later. (See appendix C.) The inventions were quite different
from one another even though they all related to electron tube
circuits for performing arithmetic. Dickinson, for example, used
a sawtooth oscillator whose phase, relative to the controlling
clock, indicated the decimal digit stored in that oscillator. The
Copyrighted Material
82 Building IBM

early work at NCR used ten thyratrons (gas-filled tubes) con­

nected in a ring for digit representation. Progressing from addi­
tion only, the NCR group had demonstrated subtraction and
multiplication by 1942. The RCA effort was apparently directed
toward electronic fire control for antiaircraft guns and involved
both analog and digital circuits.l1
Another engineer with early interest in electronics was Ralph
L. Palmer, who had joined IBM in January 1932 with a bachelor's
degree in electrical engineering from Union College. Like Dick­
inson before him, Palmer was attracted to electrical engineering
and Union College by General Electric's advertising campaign
that characterized Steinmetz as the wizard behind its advanced
electrical products. Palmer took all the math and physics courses
he could. He thought "Maxwell's electromagnetic theory of light
was one of the most beautiful things that you could run across."
Because of the depression, he was unable to get a job with GE,
RCA, or other companies that were known for their work in
electronics. So he returned to Union College to begin graduate
work.
When IBM offered him a job working on electromechanical
equipm en t he eagerly accepted. Assigned to the electrical labo­
,

ratory in Endicott, his primary mission was to support product

development work by measuring the electrical characteristics of
new devices and by offering suggestions for their improvement.
As the first engineer in Endicott with training in electrical and
electronic engineering, he taught a number of courses for the
other engineers. He also began building specialized test equip­
ment to measure the mechanical and electrical characteristics of
relays and other devices. Among these was an early oscilloscope
built to display device characteristics on a cathode-ray tube. The
necessary electronic components were not available in Endicott,
so he purchased them at radio supply stores in New York City
during a business trip to IBM's headquarters. The managers in
Endicott had so little understanding of the value of an oscillo­
scope that Palmer had to use his own money to pay for the
components. When they saw what an oscilloscope could do, how­
ever, their attitudes changed. In 1937 Palmer was promoted to
super visor of the electrical laboratory.12
At Bryce's urging, Palmer undertook numerous small research
projects in such areas as photodetection of printed spots on cards
and in the storage of information in magnetic media on cards,
tapes, and drums. The type of research Bryce encouraged was
CopYrighted Material
 Research for Patents and Devices 83

tolerated rather than supported in Endicott. "There was a tremen­

dous gap between that kind of experimental work in the labora­
tor y and the product," Palmer recalled. By and large IBM
engineers believed in achieving product design objectives in the
most conservative manner possible.13

Building Electronic Calculators

By early 1941 Palmer had become sufficiently enthusiastic about

the possibilities of using vacuum tube circuits for high-speed
counters that he assigned the task of building one to Byron E.
Phelps, a member of his department. Phelps had received his
electrical engineering degree from Union College in 1935, just
three years after Palmer. The immediate stimulus for Palmer's
suggestion was a 1939 article on counters published in the peri­
odical Electronics by a professor at the University of Illinois. The
counters made use of bistable vacuum tube circuits similar to ones
proposed two decades earlier. Originally called "triggers," bistable
vacuum tube circuits were soon more commonly known as "flip­
flops."14
Because flip-flops have only two stable states, usually described
as "on" or "off," their initial use in counters was to represent
binary numbers, in which each digit is either a 0 or a 1 rather
than ranging from 0 to 9. These binary counters, with one flip­
flop per binar y digit, found growing favor for scientific applica­
tions, for example, in radiation particle counters in which each
particle caused an electrical pulse that advanced the counter by
one. Binary arithmetic was little known outside academic and
scientific communities, however, and was completely foreign to
accounting practice. What Phelps needed was a circuit with ten
stable states to replace the mechanical counting wheel in which
each of ten rotational positions was used to represent one of the
ten decimal digits.
The most obvious way to replace a decimal counting wheel with
flip-flops was to use ten flip-flops arranged in a ring. Only one of
the ten was "on" at one time, and they were connected so that an
input pulse turned the "on" flip-flop "off' and the next one in
the ring "on." This simplistic approach was selected a year or so
later by the designers of the ENIAC, the world's first large-scale,
programmed, electronic computer.15
Wishing to make his design more economical for a commercial
product, Phelps devised a fliQ;flQO _cir c\lit using only one vacuum
r.;opyngnfea'Matenal
84 Ruilding IBM

tube (a twin triode), and he sought ways to represent the decimal

digits 0 through 9 with fewer than ten flip-flops . He believed this
must be possible because in binary notation 8 digits can be
represented by only three flip-flops, and 16 digits by four. In
December 1941 he hit upon a method of interconnecting five
tubes so that a sequence of nine pulses caused four of the inter­
connected tubes to progress from the O-state to the 9-state, but
the 10-state was blocked by the fifth tube so that the tenth pulse
caused the first four tubes to return to the O-state. Simultaneously,
it generated a carry pulse that could be transmitted to the set of
tubes representing the next higher order digit. Although Phelps
didn't think of it in those terms at the time, he had invented the
subsequently much used "binary-coded decimal" repre­
sentation.16 He had also devised an electronic circuit for repre­
senting decimal numbers that required half as many vacuum
tubes as the circuit subsequently designed for ENIAC.
He and Palmer began planning for a complete electronic mul­
tiplier based on the new circuit. With the help of one technician,
construction began before the end of 1941. Almost immediately,
Phelps recalls, "there appeared to be a commercial demand for
a "cross-footing" keypunch-that is, a keypunch that could add
or subtract two or more numbers as they were being punched
into a tabulating card. and automatically punch the result in
another field of the same card. Such a cross-footing punch, using
modified binary counters identical to those in the multiplier and
combined to form a 10-position accumulator, was built at the
same time as the multiplier and actually finished first." Com­
pleted in March 1942, the unit's test results were good enough
that a second cross-footing unit was completed with improved
components intended to meet formal product test requirements.
The multiplier, which was larger and more complex, was com­
pleted by the end of the year. It was capable of multiplying two
6-digit numbers and placing the product in a 12-digit accumula­
tor. Numbers read from a card could be multiplied and the
product punched in the same card at speeds greater than needed
by the fastest card-handling equipment. By December 1942 the
electronic multiplier routinely executed five hundred random
multiplications without error. The multiplication process con­
sisted of repetitive additions of the number in the multiplicand
register into the product accumulator.17
Because of Palmer's departure for the navy early in 1943 and
Watson's decision that company resources were to be devoted to
Copyrighted Material
 Research for Patents and Devices 85

Figure 7.2. Electronic Arithmetic Circuits

Top: Decimal digit counter built in the IBM Endicott laboratory near t he
1
end of 1941. Only 4 12 tubes were needed to sto re one decimal digit; two dig­
its co u ld be stored in this 9-tube unit. The sma ll box of ten neon bulbs is
shown plugged in to a test socket to reveal which of ten digi ts was stored in
the back row of tubes. Bottom: The electronic multiplier pictu red here with
a card reader-punch in December 1942 was able to read two numbers from
a card, multiply them together electronically, and punch the resultant prod­
uct into the same card at a rate of 50 cards per minute. These electronic
circ uits were the basis of circuits used in the IBM Type 603 Electronic Multi­
plier, the first commercial product to offer electronic arithmetic capability.

Copyrighted Material
86 Building IBM

the war effort, the electronic computing circuits developed in the

Endicott laboratory were not considered for use in products until
near the end of World War II. They were then used in the
company's first electronic product as discussed in chapter 10.

Contacting Outside Researchers

Bryce's primary research thrust was to increase the number of

patents generated from within the company. He recognized, how­
ever, that many crucial ideas and patents would always be created
by others, especially in areas new to the business. Accordingly he
and other top technical leaders were continually on the alert for
new developments.
It was in this context that the Endicott laboratory manager
wrote to John W. Atanasoff in May 1942 concerning Atanasoff's
effort to develop an electronic computing machine. "When your
development work has proceeded to a point where you feel that
it is proper for representatives of our Company to look over your
machine, we would appreciate an opportunity to do so," the
laboratory manager wrote. "Naturally, we do not wish to accept
any confidential disclosures. "18 The Endicott laborator y manager
had apparently learned of this project earlier that year when
Atanasoff requested electrical contact brushes from IBM.
More than four years earlier Atanasoff had become interested
in devising an electronic means for solving mathematical equa­
tions encountered in his research work. As a professor at Iowa
State College (later Iowa State University), he had very limited
financial or technical resources. He was , however, able to obtain
the help of a graduate student, Clifford E. Berry, to carry out
some preliminary circuit design studies. To reduce the complexity
of their vacuum tube circuits, they elected to use binary arithme­
tic. For Atanasoff this was an easy decision. A book given to him
by his mother had introduced him to number systems in bases
other than ten during his childhood.
By December 1939 Atanasoff and Berry had enough
confidence in their results to begin constructing a special-pur­
pose electronic computer. A grant of $650 from the school in
1940 left only $200 for materials after paying Berry the modest
stipend of $450. Significant progress was made in 1941 when a
$5,000 grant was received. The electronic computing device was
still not fully operational in 1942 when Atanasoff and Berry left
r
for wartime assignments. Likt the first .o the two I BM electronic
copyngntea Malena
 RRsearrh for Pall'nls and Dl'7.Iices 87

arithmetic units completed in 1942, the Atanasoff-Berry com­

puter was designed only for addition and subtraction because that
was all the application required. Its arithmetic unit contained
about 300 vacuum tubes. Approximately the same number were
used for system control and regeneration of information stored
capacitively on two rotating drums, each of which could hold up
to 30 numbers of 50 bits each.19
No response was received by IBM to its inquiry, presumably
because of the departure of Atanasoff and Berry to their wartime
assignments. Thus one can only speculate as to what type of
cooperative effort, if any, might have resulted. Furthermore, their
pioneering electronic computing machine was never fully com­
pleted, bringing it perilously close to having no impact whatso­
ever on future developments. It was saved from this fate only
because Atanasoff showed John W. Mauchly the partially com­
pleted computer and a written description of it in June 1941.
Fourteen months later at the Moore School of the University
of Pennsylvania, Mauchly proposed building what became the
ENIAC. Exacdy what Mauchly learned from his discussions with
Atanasoff has been a matter of dispute, but there can be litde
doubt that these discussions helped Mauchly formulate his belief
that an operational electronic computer could be built. For a time
he even considered constructing an Atanasoff-type computer at
the Moore School. Gradually his plans changed, however, and
ENIAC was dramatically larger, faster, and quite different in de­
sign from the Atanasoff-Berry computer.20

Copyrighted Material
8
World War II Activities

The decade of the 1930s was satisfying for Thomas Watson. He

had strengthened the company's R&D activities and its product
line during the early depression years, and he had watched these
efforts bear fruit as IBM's data processing equipment became
essential administrative tools for governmental programs. Reve­
nues, which had stagnated in the $17 to $19 million range during
the first half of the decade, doubled to $38 million during the
second half after passage of the Social Security Act in 1935. The
number of employees followed a similar pattern, changing little
during the first half and then nearly doubling to 11,315 by the
end of the 1930s. (See appendix A.) The company's leadership
over its nearest rival, Remington Rand, had increased from eight­
to tenfold when measured in number of units shipped.' President
of IBM since 1915, Watson was gaining ever more recognition as
a successful businessman and civic leader.
Like many others he believed Hitler would stop short of a major
war once his initial goals were achieved. It seemed self-evident
that far more could be accomplished through peace than
through war. Coining the phrase "world peace through world
trade," Watson vigorously promoted the concept as president of
the International Chamber of Commerce, a post to which he was
elected in 1937. He changed his view, however, following Hitler's
invasion of Poland in September 1939 and the entry of France
and Great Britain into the war. Soon Watson was devoting his
efforts to winning the war as fer vently as he had been to main­
taining the peace.2

Organizing for War

Watson placed all IBM facilities at the disposal of the government

�
in July 1940, just ten emJY ¥lW JWc}}
srlijltional Emergency Proc-
90 Building IBM

lamation found IBM to be a potential supplier of war materiel.

Manufacturing space was expanded with new facilities in Endicott
and Poughkeepsie, New York, in Washington, D.C., and in Cali­
fornia. To operate these plants, the number of manufacturing
employees was increased from 4,712 in 1940 to a peak of 12,238
in 1943.
The largest new facility was in Poughkeepsie, where the com­
pany previously had no facilities at all. A new IBM subsidiary, the
Munitions Manufacturing Corporation, was created in March
1941. Frederick H. M. Hart, of IBM's longtime supplier, the
Frederick Hart Manufacturing Company of Poughkeepsie, was
elected president. That April the new subsidiary purchased 215
acres of property adjacent to the Hudson River from R. U. De­
lapenha and Company. On the property were two small buildings
previously used in processing specialty foods. Just two months
after the December 1941 bombing of Pearl Harbor brought the
United States into World War II, a new 140,00O-square-foot manu­
facturing facility had been completed; over 250 people 'had been
hired; manufacturing tools had been installed; and the first prod­
uct, a 20-millimeter aircraft cannon, began coming off the assem­
bly lines.3
By the end of the war, thirty-two different ordnance items had
been mass-produced at IBM's manufacturing sites. These in­
cluded aircraft and naval fire-control instruments, 90-millimeter
antiaircraft gun directors and prediction units, Browning auto­
matic rifles, 3O-caliber carbines, supercharger impellers, and
bombsights.4 Not wanting to profit from wartime production,
Watson took the unusual step of freezing his own compensation
at the 1939 level. He also voluntarily limited corporate profits on
munitions to less than 1.5 percent, setting aside that amount as
a special fund for the benefit of widows and for the education of
children of employees killed in service. Watson was proud of the
wartime record of IBM. It never missed a delivery date and never
had parts rejected because of poor quality.5
Immediately after the United States entered the war, Watson
instructed his senior engineers and inventors to devote their
entire effort to wartime needs. Normal engineering duties were
to be handled by their subordinates, who were also to be assigned
to wartime projects whenever possible. At that time approximately
350 ofIBM's 10,000 domestic employees worked in the company's
research and development laboratories.6 Ninety-nine separate re-

Copyrighted Material
 Wrffld War II Activities 91

search and development projects were undertaken for the mili­

tary during the war.7
Watson also terminated the Commercial Research Department
at corporate headquarters so its members would be available for
assignments in support of the war effort. For example, the de­
partment's manager, F. J. Wesley, who had previously devised the
army's personnel accounting system, became a technical adviser
to IBM's Washington Federal office to assist in developing new
punched-card procedures for the military services and govern­
ment agencies.
Among the challenges was that of writing checks for payment
of dependency allowances to families of service personnel as
mandated by the Dependency Allowance Act of 1942. Wesley
worked with IBM engineers to develop a check-writing collator
specifically for this task . Simultaneously he provided technical
guidance in establishing the U.S. Office of Dependency Benefits
(ODB), which came to have the largest single installation of
punched-card machines of World War II. Inducted into the army
in August 1942 with the rank of major, he directed the accounting
and check-writing aspects of the family allowance program of the
ODB.8
To coordinate the company 's wartime activities, Watson created
the Department of Logistics in February 1942, placing it under
the direction of Frederick W. Nichol, who had begun his IBM
career as Watson's secretar y in 1914. Nichol, who had served in
the army in World War I and had been placed in charge of IBM's
foreign activities in 1930, was now the company's vice president
and general manager. "Major Nichol is eminently fitted for this
important work by reason of broad executive experience in in­
dustry, combined with military knowledge," Watson asserted. "He
has intimate knowledge of foreign industrial methods and
resources. '>9
Munitions production and special research and development
projects were important, but so were the company's regular prod­
ucts for which 85 percent of the orders were on government
priority schedules. Emphasizing this point, Watson told the em­
ployees, "One of our machines just recently was given priority of
shipment by the Government over a certain munitions prod­
uct. ... Two of our machines were taken to England on a bomber
recently at the request of the British Government."10 At the IBM
school in Endicott, some 1,300 members of the armed services
were trained during &p��tiRJ �Gtelf5J of EAM equipment for
92 Building IBM

accounting and statistical control. Spoken as the three letters

of the alphabet, EAM stood for electric accounting machines
and had grown in favor for identifying all IBM punched-card
equipment.ll
"Production of American industry is going to startle the world
by the end of this year," Watson asserted. 'This is no time for us
to think in terms of'!' or 'We,' whether we are in the factory,
the offices, or the field force. We must think only in terms of
WE-All for victory." Then showing a remarkably perceptive view
of the distant postwar era, he said, "Mter we have won the war,
we must win the peace by establishing sound economic policies
between nations so that raw materials, food and clothing, and
other comforts of life will be fairly distributed throughout the
entire world. "12

Mobile Machine Rewrd Units

Early in 1940 the Adjutant General's Office (AGO) became inter­

ested in the possibility of developing mobile punched-card pro­
cessing units to travel with combat units for the preparation of
personnel status records, reports of change, casualty lists, and
strength summaries. Following preliminary discussions in the lec­
ture rooms of the IBM schoolhouse in Endicott, twelve IBM
systems engineers and consultants were dispatched to Washington
in June to assist the War Department in determining its require­
ments. It was quickly determined that mobile units could not be
successful unless their record-keeping methods were compatible
with those of the entire AGO. All parts of the AGO record-keep­
ing system needed to be modernized.
The first phase of the modernization effort began with the
development of punched-card procedures for preparing and
processing the desired reports. Machine record units for imple­
menting these procedures were then established in the nine corps
areas and in the headquarters commands at Fort Knox and
Bolling Field beginning in November 1940. Similar AGO instal­
lations were soon operational in Panama, Puerto Rico, Hawaii,
and the Philippine Islands as well.
The second phase began in July 1941 when three IBM engi­
neers in Endicott were assigned the task of initiating the design
of a mobile EAM unit. A truck was purchased with a twenty-foot
van body, and special rubber mountings for punched-card equip­
ment were devised. Preliminary tests at the army's proving
Copyrighted Material
 World War II Activities 93

ground in Aberdeen, Maryland, were followed by more severe

tests of the mobile unit during militar y maneuvers in Louisiana.
The equipment performed well even under adverse road and
weather conditions. Requests for radio operators, truck drivers.
and other specialists were handled expeditiously by machine
processing of cards in the mobile unit.
A second and larger mobile unh using a thirty-foot trailer van
with a 21/2-ton tractor was designed to overcome limitations re­
vealed in the smaller unit. Less than two months after the army
successfully completed tests of the larger unit. the Japanese at­
tacked Pearl Harbor. A number of mobile units were immediately
ordered by the Adjutant General's Office. The Endicott Engineer­
ing Laboratory built and delivered sixteen by the spring of 1942.
A total of 274 had been built and delivered by the end of the war.
When U.S. troops landed in Tunisia, Sicily, Italy, and Nor­
mandy, the mobile machine record units were pulled up onto the
beaches along with guns, ammunition, ambulances, and other
mobile equipment. Frequently they followed the initial assault by
only a few days. Eager to learn the secrets of this new application
for data processing equipment, the German High Command
issued instructions for their capture. The American Fifth Army
in Italy, for example, came into possession of the orders of a
German commander instructing his troops to capture intact an
IBM mobile records unit with all of its machines, records, and
personnel. But none of the equipment or personnel was ever
captured.13

Teleprocessing with Punched Cards

Mobile record units plus modernized EAM facilities at fixed stra­

tegic locations solved many record-keeping problems for the mili­
tary. But for some information, the time required to transport
decks of punched cards from one location to another was too
long. The Army Air Corps had the most severe problem. Its
operations would involve continual rapid movement of planes,
flight crews, fuel, replacement parts, and other supplies through­
out the world.
"
Just prior to World War II, the Army Air Corps monitored its
stock by a system of machine-posting in duplicate of stock record
cards at stations and depots. Once a year the duplicates were sent
in to the Air Materiel Command headquarters at Wright Field,
Da yton, Ohio. Ther COWfrlgArSlfMID�RlIrlly transcribed to IBM
94 Building IBM

tabulating cards, which in turn were used for preparing reports

for purchasing and distributing new stock. Once-a-year balancing
of supplies against use had been satisfactory in peacetime, but
would be to tally inadequat e in a global war. To make inventory
balances available daily telegraphic transmission of data was en­
,

visaged, but such data could not be processed without manual

transfer from teletype tape to punched cards. The anticipated
volume of information was staggering. Transactions would ap­
proximate 40,000 per day from each depot for inventories esti­
mated at 300,000 items. A manually operated system seemed
doomed to failure.J4
The Army Air Corps approached IBM late in 1940, seeking a
machine that would read holes in telegraphic paper tape and
automatically transfer the data to punched cards, thus avoiding
the la borious and inaccurate manual trans cription Fortuitously .

the IBM Commercial Res earch Department had a lready made a

preliminary design of such a machine following inquiries by the
National Analine Company in the spring of that year. But business
considerations had impeded implementation. Among these con­
siderations was the type of relationship required between IBM
and the American Telephone and Telegrap h Company (AT&T),
which owned the telephone lines and telegraph equipment.15
With business issues now set aside for reasons of national secu­
rity the Endicott laboratory personnel obtained technical infor­
,

mation from AT&T concerning its equipment and had a

prototype machine completed within a few months.16 Initial op­
era tion was established between Wright Field and the Pacific
Overseas Air Service Command Headquarters in Oakland, Cali­
fornia, using an AT&T leased wire circuit. In August 1941 the
IBM Type 40 tape-to-card punch began coming off an Endicott
production line. Using this device, information transmitted over
telephone lines and punched into a teletype tape could be
punched automatically into IBM cards. In another six months the
Type 57 card-to-tape punch was also available for converting
information stored on punched cards into the same information
on teletype tapeP
With five punched-hole locations across its width, the standard
teletype tape offered thirty-two different combinations of
punched and unpunched holes-sufficient to encode the twenty­
six alphabetic characters, with six combinations left over. Three
of these were used for spacing, carriage returning, and paper

Copyrighted Material
 World War II Activities 95

feeding. Two more were used to signifY whether the subsequent

entries were alphabetic or numeric; this was needed because
there were not enough combinations in one row to accommodate
both alphabetic and numeric characters. At the standard spacing
of ten rows per inch, a single eight-inch-diameter teletype roll of
one thousand feet could accommodate the content of 1,500 fully
punched IBM cards.
Following the lead of the air corps, many other U.S. military
organizations were soon handling stock control at the depot level
using electric accounting machines in conjunction with a com­
pletely integrated system of telegraphic stock-balance control pro­
cedures based on IBM card-to-tape and tape-to-card machines .
This application of punched-card equipment may have been
IBM's greatest single contribution to the successful prosecution
of the war.18 But it was the innovative "radiotype" that won the
highest accolades from many military officers.

The Radiotype

"Every phase of our war effort was affected by your equipment,"

the chief of army communications told Watson after hostilities
had ended. "The setting up of overseas as well as domestic wireless
circuits with Radiotype, which sent and received messages auto­
matically and typed them out on the typewriter, was one of the
most important contributions. "19
The activity leading to this contribution began in 1931 when
Walter S. Lemmon, president of the Radio Industries Corpora­
tion, and his two associates built the first working modepo Born
in New York City, the thirty-five-year-old Lemmon had studied
electrical engineering at Columbia University and had already
acquired considerable wealth from his early patents on radio
technology.21 The Radiotype consisted of an electric typewriter
coupled with radio transmitting and receiving apparatus. Using
the equipment, a person could type a message on one typewriter
and cause the identical message to be typed automatically by a
second typewriter at a location many miles away. An early version,
described as a "crude model," was demonstrated to Watson, who
had learned of Lemmon through contacts at Columbia Univer­
sity. Enthusiastic about the concept, Watson agreed to support the
research needed to create a product. Thereupon Lemmon and
his associates joined IBM.

Copyrighted Material
96 Building IBM

By 1935 they had a relatively satisfactory model operating. Like

all subsequent Radiotype models, it used the IBM electric type­
writer for its keyboard and printing mechanism. (IBM had en­
tered the typewriter business in 1933 by purchasing the
Electromatic Typewriter Corporation of Rochester, New York.22 A
key motivation for B ryce and Watson was the possible use of
electric typewriter elements for input-output equipment and
small computing machines.)23 The Radiotype was designed for
shortwave radio hookups to facilitate interoffice communication
within large buildings or between factories, home office, and
branches. The first significant use of the system, however, was by
U.S. military forces during World War II.
Late in 1941 IBM loaned several Radiotype models to the Army
Signal Corps to test their applicability for communicating among
facilities in Washington, D.C., Dayton, Ohio, and Chicago, Illi­
nois. So successful were these tests that the Signal Corps ordered
units for its stations at San Francisco, Honolulu, Panama, and
Puerto Rico immediately after the country's entry into the war.
More orders followed as additional sites were added. The nerve
center for this growing network was the Army Signal Corps radio
station WAR, located in the Pentagon in Washington, D.C.
Prior Signal Corps practice had been to transmit messages by
radio using Morse code. The incoming messages automatically
activated a recorder that printed the dots and dashes on a con­
tinuous tape, but it was necessary for a specially trained operator
to translate the code and type the messages on a typewriter. With
Radiotype the messages were automatically typed and ready for
delivery the moment the transmission was finished. Later in the
war Radiotype units were coupled to electromechanical encryp­
tion and decryption equipment to provide almost instant trans­
mission and receipt of secure messages.24
Soon after the Commercial Research Department at corporate
headquarters was terminated, Steve Dunwell returned to Endicott
to work on the Radio type. It was just being put into production.
The first order was for thirty-eight systems. Dunwell undertook
the task of writing the manuals and hiring and training people
to maintain the systems . To avoid having to train people in elec­
tronics, he hired only those with first-class amateur radio operator
licenses. As soon as they completed the month-long training
course, the trainees were shipped out with the first Radiotype
units and a new class was begun. 25

Copyrighted Material
 World War II Activities 97

Figure 8.1. Radiotype at WAR

Top: Radiotype units are shown in operation at the Signal Corps radio station
WAR in the Pentagon in 1942. Bottom: The component parts of a Radiotype
unit are labeled. The availability of Radiotype at the beginning of the war
dramatically improved the communications capability of t he U.S. military.

Copyrighted Material
98 Building IBM

Deciphering Enemy Codes

While working on the Radiotype project, Dunwell came in contact

with Signal Corps officers who persuaded him and a fellow em­
ployee to volunteer for duty in the cryptographic service. They
reported for duty in Washington, D.C., in May 1942. Commis­
sioned as officers without any military training whatsoever, they
bought their uniforms, pinned on their lieutenant's bars, prac­
ticed saluting, and were ready to serve.
To carry out their assigned task of deciphering encrypted en­
emy messages, they ordered EAM equipment from IBM and in­
stalled it in the auditorium of a temporary facility obtained from
the Arlington Hall Junior College (for women) in Arlington,
Virginia. Their offices were on the stage of the auditorium. Later
they moved into two-story buildings built especially for receiving
and analyzing enemy messages. Activities involved the German
and Japanese languages. Security was tight. A double wire fence
surrounded the complex, and there were additional fenced-off
areas within it. The greatest fear was that the enemy would learn
how successful the group was and, as a result, adopt new encryp­
tion methods that might be harder to decode.26
Dunwell taught himself cryptography using Signal Corps edu­
cational materials assembled before the war. One common cryp­
tographic method used a code book and a key book. The code
book contained a list of words or phrases, and beside each was a
number selected to represent it. A message encoded in these
numbers could easily be decoded by anyone possessing the code
book. To make the encoded message more secure, a second
cryptographic step was taken. This involved the key book, which
contained a large number of random digits running in one long
string from the beginning to the end of the book. The sender
selected a small portion of the string as his "key" and "locked"
(or encrypted) the encoded message by adding the first digit in
the selected portion of the string to the first digit of the encoded
message, the next digit in the string to the next digit in the
message, and so on to the end of the message.
A message so encrypted could be decrypted (deciphered) by
anyone who possessed both the code book and the key book and
who also knew which key had been selected. The two books
occasionally came into enemy hands because they had to be
broadly distributed among friendly forces. Without the books,

Copyrighted Material
 World War II Activities 99

decipherers faced the daunting task of reconstructing them from

bits and pieces of information by painstaking processes that defy
easy description, but which could be facilitated by punched-card
equipment.
Even if one had a copy of the enemy's code and key books, the
task of determining what key had been used for a given message
was far from trivial. This task was also made easier by EAM
equipment. T he first step was to transfer all the numbers from
the enemy's captured key book onto punched cards in overlap­
ping strings of fifty digits each. These cards were then sorted into
numerical order. To decipher an intercepted message, the first
step was to guess what some of the words were, based on the
origin and context of the message. A guessed word was called
a crib.
By writing the digits from the enemy's code book correspond­
ing to one of these cribs, it could easily be determined what string
of digits, when added to those of the crib, would produce the
digits in the intercepted message. Then, using punched-card
equipment, a search was made for all keys containing this par­
ticular sequence of digits. Next, all the keys that hypothetically
might have been used were tried out against the next few words
in the message. If one of the keys appeared to be correct, it was
used to decipher the entire message. If none of the keys worked,
more guesses were made and the process repeated. Because hun­
dreds of guesses (cribs) might be needed to decipher a single
message, and because each of these guesses might produce
hundreds of plausible keys, the task of finding the correct
key would have been nearly impossible without automated
equipment.27
A deficiency of the automated process, as initially envisioned,
related to determining when a hypothesized key was correct.
Examination by clerks of the output of processing thousands of
hypothesized keys against the encrypted messages was far too
time consuming. T he solution was to assign a probability of oc­
currence to each word in the enemy's code book based on lin­
guistic frequency of use. The words of a correctly decoded
message segment would have a collective probability of occur­
rence dramatically higher than words produced by any incorrect
key. What was needed was a machine that could rapidly calculate
the collective probability of the words in a hypothetically decoded
message segment.

Copyrighted Material
100 Building IBM

Figure 8.2. Stephen W. and Julia DWlwell

Steve Dunwell is shown with his wife, Julia, at the March 1946 ceremony at
which he was awarded the Legion of Merit for his cryptographic work dur­
ing World War II. He and Julia spent their brief honeymoon at Watson's
suite at the IBM Country Club in Endicott in January 1943. In the late 19505
he headed Project Stretch, which produced the most powerful computer of
that era.

Copyrighted Material
 World War II Activities 101

A Trip to Remember

Dunwell's solution was to obtain a specially designed processor

built with electromechanical relays. He first studied the charac­
teristics of relays used in telephone switching systems and found
them to be too slow and cumbersome. Then he evaluated the
wire-contact relays developed by Clair Lake for the IBM-Harvard
Mark I and found them to be quite appropriate. Seeking final
approval from Watson to have this special equipment fabricated
by IBM, he stopped off in New York City on his way to Endicott.
Approval was given swiftly by Watson who said the Signal Corps
could have anything it needed.
T hen Watson made another decision that provided Dunwell
with a trip to Endicott he will never forget. Because of wartime
pressures, Dunwell had been allowed only one half day off for his
planned wedding. With his normal 61/2-day work week, there was
no opportunity for a honeymoon. So he had arranged to be
married on 18 Januar y 1943, one day before his trip to Endicott,
and to have his bride travel with him for their honeymoon.
Learning of this arrangement, Watson asked where they were
planning to stay. When Dunwell told him, Watson said, "That's
an awful place up there. Why don't you stay at my suite at the
IBM Co untry Club?" And so they did-probably the only couple
ever to spend their honeymoon in Watson's suite.
A unique feature of the equipment specified by Dunwell and
built by IBM was the function of its many plugboards. These
permitted the machine to be wired to recognize the sets of
code-book digits that corresponded to the most frequently used
words and to supply their probabilities. Cards were fed into the
machine so that each card that contained part of an encrypted
message was followed by one or more cards with a hypothesized
key. For each key the machine performed the decryption, as­
signed probabilities to the resultant words, and calculated the
probability that the message segment had been correctly de­
crypted. The relays were fast enough that the entire analysis could
be done between the reading of one card and the next-at the
rate of 150 cards per minute. The first unit worked exceptionally
well, and more units were put in service.
The process described here is only one example of the great
variety of decryption problems encountered and solved during
the war. The ready availability of standard IBM equipment and
the willingness of Watson to provide modified equipment made
Copyrighted Material
102 Building IBM

Figure 8.3. Cryptographic Equipment

A relay calculator specially built for cryptographic work by IBM is shown
(left) attached to an IBM Type 405 Alphabetic Accounting Machine (right).
The front gate of the relay calculator is swung open to reveal many rectangu­
lar-shaped rel ay units. These 4-pole, 6-pole, and 12-pole wire-contact relays
had been designed by Clair Lake for use in the IBM-Harvard Mark 1. Plug­
boards for controlling the calculator are behind the five rectangular covers
on the top, and the two covers on the bottom, of the large front gate.

it possible for the Signal Corps to solve new problems quickly. A

typical set-up used at the Signal Corps involved three sorters and
a collator with one operator. At times as many as twenty-five of
these set-ups would be working simultaneously on a single decryp­
tion problem.28
A critical limitation for the entire operation was the time it took
for keypunch operators to transcribe the intercepted messages.
That limitation became unacceptable "when out of the Pacific, a
full bomber load of intercepts was dumped on us at once," Dun­
well recalled. By the time they could have processed all of the
messages, the information would have lost its value. Following this
event, a teletype network was established throughout the Pacific
region. Intercepted messages were keyed in at local sites, trans­
mitted over telephone lines, and recorded on teletype tape in
Washington, D.C. The information on teletype tape was automat­
ically punched into IBM cards using the Type 40 tape-to-card

Copyrighted Material
 World War II Activities 103

punch, which had been placed into production only a few months
before the attack on Pearl Harbor. 29

Lusting for Electronics

As supervisor of the electrical laborator y in Endicott, Ralph Pal­

mer participated in many wartime projects. He had the opportu­
nity to go on battleships to observe fire-control systems
manufactured by IBM, and he worked on a sophisticated odo­
graph and a simulator for pilot training. He was disappointed,
however, on two counts. First, all of his projects used analog
rather than digital circuitry. Second, except for his own test
equipment, there was little opportunity to use electronics.
His frustration was heightened by the company's lack of inter­
est in using the electronic arithmetic circuits he and Phelps had
designed and built in 1942. Not only were these circuits judged
to be too fast for any EAM product then under consideration, but
there was deep concern that the company's patent department
lacked the expertise in electronics needed to ensure that product
designs would n ot infringe patents of others. Furthermore Wat­
son had terminated commercial product development in favor of
wartime projects.
Several of Palmer's friends had joined the navy's cryptographic
organization, the CommuniCations Supplementar y Activity­
Washington. This organization was often referred to simply as
Naval Communications or as "seesaw" because its initials were
CSAW. His colleagues told him about "the fantastic amount of
electronic equipment and the computing that was going on
there." In Washington on other business, Palmer arranged to
meet with leaders of the organization. Wartime security limited
the information Palmer could be given, but he was assured he
would do a lot of work with electronics. "Probably before I left, I
was sold," he recalls. "I almost immediately got orders to report
for dUty."3O

The Navy � Task

A primary task of the navy was to ensure that troops, munitions,

and other supplies could be shipped from the United States to
Great Britain and later to North Mrica and Europe. In 1942
German submarines had sunk more tonnage of ships than the

Copyrighted Material
104 Building IBM

Allies had built. During the first three months of 1943, shortly
before Palmer reported for service, the number of ships sunk
appeared to be doubling month by month: twenty-nine ships in
January, fifty in Februar y, and ninety-five in March. Critical to the
German success was their use of an encryption and decryption
device for sending messages to and from their submarines.
Known as the Enigma, this device contained three rotors, so wired
and mechanically interconnected that they converted letters
entered on the keyboard into an apparently random set of out­
put letters. Even if the Allies came in possession of an Enigma,
the Germans believed they would not be able to decipher mes­
sages unless they also knew what key had been chosen for the
transmission.31
The Allies' ability to protect convoys of merchant ships would
depend in part on their ability to read encrypted Enigma mes­
sages. Using espionage information and sophisticated mathemat­
ics, a Polish cryptographic group in the mid-1930s had
reconstructed an Enigma and modified it to run rapidly through
key settings to find the correct key for a given message. These
devices were called "bombes" because of their appearance and
the ticking noise they made.32 Under increasing military threat
from the Germans, the Poles delivered their equipment and
information to the French and British. The British refined the
methods, and the British Tabulating Machine Company pro­
duced numerous bombes to mechanize the decryption process.33
Late in 1942 the Germans introduced a four-rotor Enigma that
greatly increased the difficulty of finding solutions to encryptions.
The Americans, in close cooperation with the British, undertook
the task of designing and constructing larger and faster bombes
to handle the four-rotor Enigmas. Using a combination of vac­
uum tube and relay circuits, with mechanical rotors driven by a
high-speed motor, each new bombe was said to be equivalent to
six of the British-built bombes. Each unit was about 7 feet tall, 10
feet wide, and 2 feet deep and weighed 2.5 tons.
Using these machines, the Allied forces deciphered German
naval messages with increasing regularity and speed. By May 1943
the Germans were forced to move many of their submarines out
of the North Atlantic to safer areas because of "unbearable"
losses: one submarine lost for each 10,000 tons of Allied shipping,
versus one submarine lost for 100,000 tons only a few months
earlier. During the first three months of 1944, only three Allied

Copyrighted Material
 World War II Activities 105

merchant ships were sunk-at a cost to the Germans of thirty-six

submarines. The Battle of the Atlantic had been won.34
For Palmer, as for Dunwell, no time was wasted on military
training or indoctrination. Upon reporting for duty in May 1943,
he was sent directly to the Naval Computing Machine Laboratory,
operated in Dayton, Ohio, by the National Cash Register Com­
pany. Bombes and other top-secret electronic equipment were
designed and built there for CSAW. Assigned to a nearly com­
pleted project, Palmer soon came up with a design that would
require far fewer components. "Please don't push the thing too
hard," he was told. "It would upset our schedule. What we've got
may not be the best, but it will work." In developing special
wartime devices, of which only a few units would be made, swift
completion was more important than construction cost. The op­
posite had been true at IBM, where large production runs were
anticipated.
Palmer was excited by the opportunity to work on special-pur­
pose electronic equipment. Some units contained over a thou­
sand vacuum tubes and were far larger and more complex than
any he had seen before. His strong intellect, good education, and
earlier research in electronics quickly put him in a position of
leadership. Following a period in which he devised test equip­
ment and servicing methods, he became a technical executive
officer, with responsibility for coordinating cryptographic work at
Dayton with other military organizations. He learned how Dun­
well's group used modified EAM equipment to solve crypto­
graphic problems-often just as effectively as was accomplished
in the navy using a combination of special electronic equipment
and EAM. There was no doubt in his mind, however, about the
long-term potential of electronic circuits in data processing
equipment.
Toward the end of the war, Palmer began designing an ad­
vanced unit called the full selector-a cryptographic device in­
tended to provide more rapid access to information. A driving
force behind this project was Comm. Howard T. Engstrom, who
in peacetime had been a professor of mathematics at Yale. More
than any other person in CSAW, Engstrom impressed Palmer as
an outstanding technical leader. He particularly liked Engstrom's
view that anything they designed and built should be "stretching
the state-of-the-art in technology." Among the basic elements
Palmer set out to devise for the full selector was a magnetic drum

Copyrighted Material
106 Building IBM

storage unit based on research he had done in the IBM Endicott

laboratory. Magnetic drum storage devices would become critical
elements of postwar electronic computers, but Palmer's project
was barely begun when the war ended in the fall of 1945.35

A Commercial Venhtre

Already two men prominent in the hierarchy of CSAW had initi­

ated plans to continue CSAW activities as a commercial venture.
One was Howard Engstrom. The other was Lt. Comm. William C.
Norris, who in peacetime had been a sales engineer for Westing­
house and was to found the Control Data Corporation in 1957.
These two men were soon joined by Capt. Ralph I. Meader, who
had interviewed Palmer in Washington and had headed the Naval
Computing Machine Laboratory in Dayton. The idea of estab­
lishing a commercial company to continue the classified CSAW
work gained increasing support within the navy after NCR de­
clined to continue the activity and when it became evident that
key technical people were reluctant to accept civil service posi­
tions. Among those supporting the concept was James Forrestal,
then secretary of the navy and later the first secretar y of defense.
The navy could not legally help create a private company, how­
ever, and certainly could not guarantee that it would receive
government business.
The proposal was stymied by these constraints until late in 1945
when a successful investment banker and Annapolis graduate,
John E. Parker, was persuaded to raise money to start the com­
pany. In January 1946 the company was incorporated under the
name Engineering Research A.,sociates (ERA). Ownership was
equally divided between two groups: the founding technical
group headed by Engstrom, Norris, and Meader, and the investor
group headed by Parker. The equity investment was only $20,000,
although the investor group also provided a line of credit of
$200,000. Six months later the Navy Department. without com­
petitive bidding, awarded to ERA the first of a series of contracts
that assured its survival and made it the leading supplier of
large-scale electronic computers in the first few postwar years.36
Ralph Palmer was under considerable pressure to join the
group, but after speaking with Parker, he concluded the man
"didn't have the vision to really exploit the technology this group
had." Special equipment for the navy was the only business under

Copyrighted Material
 World War II Activities 107

consideration. To Palmer, the application of electronics in the

commercial data processing industry had greater promise.
Discharged from the navy in early 1946, Palmer returned to
IBM without any specific commitment concerning his job assign­
ment. Among the factors influencing his decision was that the
company had paid him 25 percent of his salary while in the navy.
The same had been done for all IBM employees in militar y
service. "I could have survived without it, 1 suppose, but the
thought behind it impressed me," Palmer observes. "This was Mr.
Watson's policy. 1 thought it was pretty unusual and pretty
good."'!>7

Copyrighted Material
9
Future Demands

"There is no business in the world which can hope to move

forward if it does not keep abreast of the times, l oo k into the
future and study the pro bable demands of the future," Watson
asserted as he broke ground for the North Street Laborator y in
1932.1
Use of the phrase "demands of the future to connote market
"

requirem ents for new or not yet concei ved pro ducts has often
- -

been considered peculiar to Watson. Like many th ings associated

with him, however, this phrase was acq uired while working for
John Patterson at NCR. To help assess future market requi re ­

ments and to review produc t proposals, Patterson had established

what he called the Future Demands Committee. It consisted of
key company executives, among them Watson. Another frequent
attendee was Charles F. Kettering, the company's leading inventor
and a person from whom Watson learned much about the pro­
cesses of invention and innovation.

Lessons from NCR

Slightly younger th an Watson Ke ttering had begun working for

,

NCR in 1904, immedia tely after graduating with a degree in

electrical engin eering from Ohio Sta te University. He resigned
five years later to found the Dayton Engineeri n g Laboratories
Company, better known by its acronym, Delco. Its first produc t ,

and the invention for which Kettering is best known, was the first
practical ele ctric starter for gasoline-powered automobiles. Before
Ket ter ing resi gned to found Delco, he had already built and
tested an operational electric starter with the hel p of other moon­
lighting engine er s from �CR. In 1916 Delco became a subsidiary
of General Motors, a company for which Kettering later served
as d irector of research from 1920 to 1947.2
Copyrighted Material
110 Building IBM

During his five years at NCR, Kettering made many improve­

ments to cash registers. These included the first electrically driven
cash register and a means for combining the output from all the
registers in a store to provide a continually updated record of
total sales . Concerning his NCR years, Kettering said: "I didn't
hang around much with other inventors or the executive fellows.
I lived with the sales gang. They had some real notion of what
people wanted."3 Top among these salespeople was Watson , who
is reported to have been "the first person outside of the Delco
group to enjoy a ride in the self-starting automobile."4
The rapid growth in customer demand for electrically driven
cash registers and other improvements devised by Kettering was
an important lesson for Watson. Good salesmanship could in­
crease the demand for existing products and influence customer
preferences, but research and engineering could create new
products with far more dramatic impact. A means for identifying
possible new product') of this type was crucial to a company's
success.
Lessons gained from Kettering and Patterson increased Wat­
son's respect for the specialized knowledge of engineers. He also
learned that he had the ability to bridge the knowledge gap
between engineers and salesmen and thus play an important role
in defining new products. Arriving at CTR in 1914, Watson took
on this role informally as he scrambled to create an engineering
organization. By the time he promoted James Br yce to chief
engineer and created a small research laboratory in 1922, he was
ready to address future demands more formally.

Forming a Committee

"In connection with the appointment of Mr. J. W. Br yce as Chief

Engineer of the Computing-Tabulating-Recording Company,
please note that a Future Demands Committee has been estab­
lished," Watson advised the division presidents and other key
executives. "We shall expect to have regular meetings of this
Committee, and as you are a member, due notices of these meet­
ings will be for warded to you in time to make your plans to be
present. "5
An important member of the committee was Otto E. Brait­
mayer, who had been hired as a fifteen-year-old office boy by
Hollerith in 1889 and stayed on to ser ve as secretar y and office
manager while obtaining a law degree at night.6 By the time
Copyrighted Material
 FulUTI' Demands 111

Watson arrived, Braitmayer was in his early forties and familiar

with all aspects of the business. Watson quickly adopted him as
his right-hand man. Working directly for Watson, Braitmeyer
negotiated contracts with customers and vendors, set and admin­
istered salaries and bonuses, gave out engineering assignments,
and documented product development objectives. By 1922 Brait­
mayer had become assistant general manager of the Tabulating
Machine Division.'
Another member was Clement Ehret, sales manager of the
Tabulating Machine Division.8 Ehret's contributions to the com­
mittee's work are typified by a 1927 letter to Watson: "T he at­
tached clippings refer to a $60,000 loss by the Waldorf Hotel
through the falsification of waiters' checks. I believe one of the
first things we should give attention to, as soon as our present
engineering program permits, is the restaurant machine. You will
recall you had in mind the use of the proposed Railroad Ticket
Printing device, with slugs, as a commissar y machine as well. We
appear to have a good basis to proceed on with this device which
would offer all food establishment.. protection against loss. ''9
Reference to Watson's ideas for new products was common in
discussions and correspondence with him. "In accordance with
your idea that it is important to study ways and means of getting
the primary data on tabulating cards as near the source of infor­
mation as possible, I have been giving some thought and study
to the problem of recording from meters, particularly electric,
gas, and water meters," Br yce wrote Watson in 1928. "We have a
sample of each of the above kinds of meters. It appears that while
it will require some nice design and ingenious work to fit a simple
recording device to these, there is nothing impossible about
it ... . The biggest problem in the whole matter I believe to be a
business one, and before recommending the expenditure of any
money in actual design or experiment, [ would like to submit
o
some figures and get your advice."l
Neither letter resulted in products at the time, but they are
indicative of the far-ranging ideas that were considered and aban­
doned along the road to products actually developed and mar­
keted. Watson's desire to capture data more directly from utility
meters was partially satisfied in the late 1930s, using pencil-mark
sensing technologies based on Rey Johnson's test-scoring equip­
ment (see chap. 6). Meter readers made pencil marks on cards,
and the marks were later converted automatically to conventional
punched holes by a mark-s�nsingdepr:oducing card punch.
copynghtecrmatenal
112 Building IBM

A Department and Its Demise

By 1929 the committee structure had become inadequate for the

task, and Watson established a Future Demands Department.
Headed by Clement Ehret, former sales vice president, the de­
partment was to study and recommend "the creation and devel­
opment of new products to cover new fields." A parallel line of
inquir y concerned the widening of usage for existing machines.
"Our facts will come from a first-hand study of industry and from
surveys and analyses made in the field," Ehret asserted.ll
With a background in sales and finance, Ehret was qualified
for some of these activities, but he was not qualified to replace
(and certainly not to compete with) Watson in working with the
company's engineers to bring about the "development of new
products." In recognition of these deficiencies in January 1934,
Watson appointed Ehret director of market research, a position
2
he held until his death in 1949.1
The Future Demands Department fell dormant for a six-year
period during which time Watson personally handled-as he in
fact always had-the assignments given to his inventors to develop
new products It was not a completely satisfactory arrangemen t
. .

Watson possessed neither the engineering background to under­

stand the technical problems nor a sufficiently detailed knowl­
edge of customer needs to fully specify new products. Filling this
void was an ad hoc responsibility of the Methods Research De­
partment, formed in 1931 to collect, disseminate, and develop
EAM methods.13 Sometimes the department learned of a pro­
posed product soon enough to influence the initial design. More
often, however, information became available only after Watson
had authorized production. Then followed a frenzied effort to
define and make those changes that were absolutely essential.
Many desirable changes were thus postponed for incorporation
in an "enhanced" product a year or so later.14

New Leadership

Although Watson had some appreciation of the problems he was

causing, he seemed unable to find anyone to whom he was will­
ing to entrust the future demands function. As he watched the
maturing of John C. McPherson, however, a man he had per­
sonally hired in February 1930, Watson increasingly believed he

Copyrighted Material
 Future Demands 113

Figure 9.1. John C. McPherson

John McPherson, IBM's director of engineering, is shown addressing a
group of customers in early 1948. He was IBM's principle technical contact
with government agencies during World War II, helped define and get an­
nounced the company's most profitable ever accounting machine (the Type
407), and was influential in getting the company into electronics in the early
postwar years.

Copyrighted Material
114 Building IBM

had found someone to whom he could entrust this critical func­

tion.
While at Princeton studying electrical engineering, McPherson
had spent his summers working for the chief electrical engineer
at the Pennsylvania Railroad shops in Altoona-a position ar­
ranged by his father, who was also an electrical engineer. The
McPherson home, like Watson's, was in Short Hills, New Jersey.
Home for Christmas during his first year at Princeton, McPherson
recalls, "Mr. Watson met me at a party at the Short Hills Club and
asked me ifl might be able to work for him at IBM." McPherson's
response was affirmative, and he joined the company soon after
obtaining a bachelor's degree in electrical engineering from
Princeton. Because of his experience in the railroad business, he
soon was assigned to the Transportation Department, which pro­
vided technical guidance to salesmen and their customers in
the transportation (especially railroad) industr y. Similar sup­
port groups existed for the insurance, wholesale stores, and oil
industries.
That May he attended IBM Sales Class 56. Upon graduation he
became a senior sales representative in the Philadelphia office
where he specialized in installations for the Pennsylvania and the
Reading railroads. Three years later he returned to New York as
assistant manager of the Transportation Department, a position
he held until Januar y 1940 when Watson had him promoted to
manager of the Future Demands Department. His new assign­
ment was to "cooperate with and coordinate the activities of the
Engineering, Patent, Market Research, and Commercial Research
departments. "15
McPherson's office at corporate headquarters was adjacent to
those of the commercial and market research groups. Among
their activities was the handling of customer requests for modified
equipment. The headquarters groups worked with customers and
with IBM engineers to define appropriate implementations and
provide price quotations.16
Working in parallel with these groups, McPherson understood
his task was "to think longer range, beyond the modification of
existing machines, to new ones we should create." It was a small
activity that did not grow beyond three people as long as he
headed it, but he enjoyed working with the engineers, patent
attorneys, and market research people. The work was intellectu­
ally stimulating and rewarding. Less than two years after his

Copyrighted Material
 Future Demands 115

promotion, the countr y had entered World War II. McPherson's

future demands activities were effectively terminated and re­
placed by wartime activities.

The Chief Engineer

In February 1943 Watson appointed John McPherson director of

engineering. Little explanation for this appointment was given,
but it was likely because of the significance of McPherson's role
as IBM's principal technical contact with various government
agencies. Since 1940, for example, McPherson had served as
contact for the U.S. Army Ordnance Corps Ballistic Research
Laborator y in Aberdeen, Maryland, where IBM equipment was
used for ballistics calculations. He arranged for Lake to build the
fastest-ever relay calculator for the Ballistics Laborator y. And he
was the person Dunwell and Palmer contacted for their special­
ized cr yptographic needs.17
Not announced when McPherson became director of engineer­
ing was Watson's plan to assume again full responsibility for
product decisions. IS The Future Demands Department fell dor­
mant (for the second time) until it was reactivated in 1947 as part
of the company's move into electronic computing.
As in his previous assignment, McPherson ostensibly reported
to Watson. Because those who wanted to see Watson had to wait,
possibly for weeks, McPherson submitted written reports from
time to time and busied himself making sure that development
projects moved ahead in the manner he believed Watson desired .
He dealt with sales and manufacturing executives at headquarters
and with key inventors in the development laboratories. He was
accepted as representing Watson when in the laboratories and as
representing the engineers when back at headquarters.
By his nature, McPherson functioned primarily as an analyst
and a facilitator. He tried to ensure that the best technical ap­
proaches were used to achieve each product development goal
established by Watson. It was a delicate task. He needed to know
what assignments each inventor had and how the inventor was
attempting to achieve them. Simultaneously he had to avoid
revealing any of the secrets of one inventor to another. To en­
courage a free exchange of information between himself and the
inventors, he adopted the policy of never filing for patents him­
self. It was a policy already in force for most technical staff people

Copyrighted Material
116 Building IBM

at corporate headquarters, with the notable exception of Bryce

and certain members of his Patent Development Department.
Concerning Watson's relationship with the company's inven­
tors, McPherson says: 'Watson challenged them to work on what
the business needed. He urged more than one engineer to work
on the same kind of project independently. Then he made a
choice." Concerning his own role as the first employee to hold
the title of director of engineering, McPherson says he always
perceived that in fact but not title, "Mr. Watson was the chief
engineer. "19

Copyrighted Material
10
Preparing for Peace

By mid-1943 the tide of battle had shifted in favor of the Allies.

The Germans had been defeated at Stalingrad in Februar y and
were retreating along much of the Russian front. Severe losses
had also forced them to withdraw many submarines from the
N orth Atlantic. The surrender of Italian forces in Tunisia in May
had opened the Mediterranean to All ied shipping and paved the
way for troop landings in S icily in July. In September, following
the resignati on of Benito Mussolini and the surrender of the new
Italian government, Allied forces landed on the I tal ian mainland
and began their advance toward Germany. The Japanese, who
had suffered their first ma jor naval defeat a year earl ier near the
island of Midway, were also on the defe n sive as U.S. troops began
working their way toward Japan, island by island.
In the cour se of three years, Watson had turned IBM into a
major supplier for the armed forces. From the end of 1940 to the
end of 1943, the company's fa ctor y floor space in the United
States had increased 140 percent to 1,891,000 square feet, its
domestic factor y workers had increased 160 percen t to 12,238,
and its total number of domestic emp l oyees had increased 100
percent to 17,262. Approxi mately two-thirds of the company's
produc tion consisted of munitions I Reven ues had nearly tripled,
.

although profits had been kept essentially flat by policies devised

by Watson to avoid profiting from the war effort. (See chap 8 .

an d appendix A.)
Now Watson faced the daunting task of continuing to support
the war effort for an indefinite time while simultaneously prepar­
ing to shift to a peacetime economy. He did not want the com­
pany to drop back to its prewar size. But how could he make
effective use of his vastly increased m anufacturing floor space?
And h ow could he pay for the cost of converting to peacetime

Copyrighted Material
118 Building IBM

production if the government and military organizations began

returning EAM equipment rented for wartime purposes?

Planningfur Product Improvements

In June 1943 Watson initiated a series of engineering meetings

intended to develop plans for improving the product line to meet
the anticipated challenges of the postwar era. Among the techni­
cal leaders in attendance were James Bryce, Halsey Dickinson,
Walter Lemmon, John McPherson, and the patent attorney John
Hayward. Other key executives attended, bringing the number to
about twenty people. Held twice in J�ne, the meetings continued
on a monthly basis to the end of the year and then somewhat less
regularly through 1944 and 1945.
At the first meeting, Watson said he hoped to retain all employ­
ees hired during the war, but he was even more strongly commit­
ted to IBM's employees in military service. "Twenty-three percent
of all the men in our organization are in the Service, and we hope
they will all come back to us, " he stated. "We will lose a few, but
the majority of them will come back, and they are going to have
jobs when they come back. We have got to create enough business
to take care of all of them, or else the rest of us will have to share
the work. That is as plainly as I can put it, because we are not
going to let any man or woman who has gone into the Service
suffer when they get back home. "
In an undisguised exhortation for greater effort, Watson said,
"It isn't necessary for anybody to make any sacrifice if we are
willing to work now and do the things we know how to do. " As
evidence that the old ways would not be good enough, he re­
minded them how quickly they had put new products into pro­
duction for the war. 'This war is changing everything," he
asserted. "Everybody who makes any progress in business is going
to work along different lines than they have ever worked before.
The people who do not change in time are going to be sitting on
the curbstone waiting for the parade to come by. "2
Watson then began a brainstorming session that was continued
in future meetings. Topics discussed included possible new prod­
ucts and improvements to old ones, the adaptation of special
wartime devices to peacetime, new technologies, and the need to
hire additional technically trained people even before the war
ended. On this latter subject, Watson had particularly strong
views. He chided the eBp _ h>M�hb9oking hard enough. He
 Preparing for Peace 119

read aloud newspaper ads placed by engineers seeking work: for

example, "Engineer,45, experimental or mechanical, electrical,
inventive and executive ability; can design and make complete
models, parts, and so forth; can do machine hand work. " Not all
those advertising for work would be good, he noted, but some
surely were. "Go to technical schools and hire people," he urged.3
To provide structure to the product development discussion,
McPherson had been asked to select the five most urgent engi­
neering projects. His choices, in priority order, were a wheel
printer to replace the type-bar printer, a new key punch, a mul­
tifunction EAM unit, machines for small businesses, and an im­
proved time recording device. The list provoked a lengthy
discussion, primarily between McPherson and Watson. Projects
were clarified and their priorities reset based on ease of accom­
plishment and value in the market.
The wheel printer, a long-time favorite of Watson, stayed in first
place. For second place, he suggested replacing the improved
punch with a machine for small businesses.As the discussion
progressed, he came back to this subject saying, "No, I am wrong
there. I would put the Electromatic typewriter second-a ver y
low-priced Electromatic typewriter, and redesign the present
model in plastic to make it at a lower cost as well." To emphasize
the point , he asserted that before long, "Every typewriter is going
to be electrical. .. . We pioneered this, now let's not let it get
away from US."4 Of particular interest to Watson was an explora­
tory project in Endicott that used an electric typewriter as the
basis of an accounting machine for small businesses.5
Watson's zest for new products to replace wartime revenues was
dampening his normal enthusiasm for longer-range projects.
Concerned by this, McPherson said, ''There is one observation 1
want to make in connection with the list. I feel that the electronic
research should go on regardless of the projects we are rushing
to current completion. We must attempt to use vacuum tube
circuits in business machines."
"Let's understand this does not mean these are the only things
we are going to work on. It means we will concentrate on these
five things as being the five most important things," Watson
responded."I want to put electronics in a class by itself because
1 think it is that important," McPherson persisted. "Can you have
something ready in six months that we can go out in the field
with?" Watson asked. "No," was the inevitable response. ''Well
" "6
then," said Watson, it8jJJ:AdMl8"�Jmfl9ftance.
120 Building IBM

Printer Research

Prior to these discussions, the wheel-printer project had been

managed in the manner of most of the company's research. A
single engineer, with minimum help, worked on it whenever he
found time from more pressing assignments. In this case H. S.
(Bud) Beattie had begun research on printing devices, leading to
the wheel printer, soon after he joined IBM in 1933.
Beattie had minimized the speed-limiting effect of the type-car­
rier inertia by mounting the type on the outside surface of a
rotating wheel rather than on the front surface of a long, verti­
cally oscillating bar as then used in IBM printers. He also devised
an ingenious method for printing in which the motion of the type
parallel to the surface of the paper was reduced to zero at the
moment of impact. The wheel was rotated at high speed initially
to bring the desired character near the print line shortly before
impact. Then the speed of rotation was slowed to bring the
character into precise position at the moment of impact. The slow
rotation speed was chosen to offset exactly the pivoting motion
that brought the type in contact with the paper.
Now after ten years of part-time research, Beattie's project
became part of an effort to provide an improved accounting
machine. The effort was slowed at first by wartime priorities and
then by more immediate postwar priorities. These delays were
frustrating for Beattie, but they also afforded him time to improve
his printer. Using a separate print wheel for each of 120 columns
across the paper, he was able to design a printer that achieved a
speed of 150 lines per minute with good print quality. Forty-eight
numerals, alphabetics, and special characters on each wheel gave
the printer outstanding function in addition to its 50 percent
greater speed than the fastest previous printer.7
In 1947 the proposed accounting machine with its improved
printer faced a crucial test. The estimated cost of tooling was so
high that the sales department recommended project termina­
tion. Watson sought the opinion of John McPherson, who had
worked closely with the development engineers. Mter McPherson
explained how its printing speed and other features could give
the customer up to eight times the performance of current ma­
chines, Watson said: "You go into the meeting tomorrow and tell
them that it can do eight times the work of the present machine.
At eight hundred dollars [monthly rental], it's a bargain."

Copyrighted Material
 Preparing for Peace 121

McPherson did as instructed, and the sales department with­

drew its objections. The machine was introduced in 1949 as the
IBM 407 Accounting Machine. Its market reception exceeded
expectations, resulting in revenues far larger than any of its
predecessors. The 407 is also noteworthy for being IBM's last
major electromechanical accounting machine product-and for
serving as the printer on the company's first large-scale electronic
computers.8
During the extensive period devoted to testing and improving
the 407, Beattie found time for research on other printing meth­
ods. A novel method devised by him used a mushroom-shaped
type element whose spherical outer surface held fifty-two raised
characters. The desired character was selected by rotating the
element about it" spherical center.9 First demonstrated by Beattie
in 1946, this mushroom-shaped device was the predecessor of the
type element used in the highly popular IBM Selectric typewriter,
announced in 1961, as well as in numerous text processors and
computer terminals introduced during the 1960s and 1970s. \0
Watson's decision in the June 1943 engineering meetings to
give the second highest priority to improving the electric type­
writer line was followed by the transfer of typewriter deve lopm en t
and manufacture from Rochester, New York, to Poughkeepsie in
1944. This paved the way for increasing research and develop­
ment on typewriters and for expanding production when the
Poughkeepsie plant shifted to peacetime products. It was a good
decision. Revenues from electric typewriter sales, service, and
supplies increased fivefold ($11 million to $5 7 million) during
the ten-year period from 1946 through 1955 . 11

Emphasis on Electronics

Stiffening German resistance by the fall of 1943 and continued

hard fighting in the Pacific caused Watson to recognize that the
end of the war was further away than he had perceived in June.
He still wanted new products for rapid release to production, but
he was now prepared to focus on longer-term projects as well. In
particular he was eager to respond to McPherson's recommenda­
tion that research in electronics be strengthened.
"Mr. Watson wants electronics used on the accounting machine
and on the printer," the summary of the October engineering
meeting records. "He wants Mr. Bryce to get out and visit the
Copyrighted Material
122 Building IBM

various colleges, find the most outstanding Professor on electron­

ics and get him for IBM. If he gets $10,000 with the college, we
would be willing to pay him as much as $25 ,000 and let him keep
in touch with his college, going back once a month to give
lectures, or perhaps as often as twice a month. "12
Responding to Watson's mandates on electronics was difficult.
All senior electronics engineers were still deeply involved in the
war effort. But events associated with the August 1944 dedication
of the IBM-Harvard Mark I stimulated Watson and Bryce to more
vigorous action. Just prior to the dedication, Howard Aiken of
Harvard University had issued a press release describing himself
as the sole inventor. The role of IBM was largely ignored. Aiken
was willing to acknowledge the intellectual contributions of
Bryce, but apparently he could not bring himself to acknowledge
the engineering contributions of Clair Lake and others.
Commenting on these issues in a letter to Aiken after the
dedication, Bryce wrote: "It was indeed very unfortunate that the
mistake was made in the original press statements given out,
identifying you as sole inventor of the machine. . . . While I
appreciate all that you said about me, I feel that injustice to these
other men, it would have been a gracious gesture on your part
and very much appreciated by them, if your letter to Mr. Lake
had contained an acknowledgement of sincere regret over such
unfortunate and erroneous publicity. "13
Watson himself was enraged by the press release and by Aiken's
deportment. Only reluctantly did Watson attend the dedication.
He had lavished hundreds of thousands of dollars and IBM's best
research and engineering talents in developing the Mark I, and
he had donated another hundred thousand dollars for its opera­
tion. Now Howard Aiken seemed determined to deprive IBM and
its technical people of any credit or benefit from the project.
Aiken also indicated he would design and build future machines
without IBM's help.14
Beyond the emotional pain, there was a very practical consid­
eration for Watson and Bryce. A primary reason for working with
Aiken had been to establish with Harvard University a mutually
beneficial relationship of the type the company had long enjoyed
with Columbia. To foster this relationship, Br yce had encouraged
the use of the company's most innovative arithmetic and storage
devices in the Mark I. He had viewed the Mark I as a machine
IBM would continually upgrade-thus providing a test bed for
new devices and COJroo��b JfA5�s as well as a facility in
 Preparing for Peac.e 123

which customer needs in scientific computation could be as­

sessed. Now the breakdown of cordial relations with Aiken had
dashed all hopes for future cooperation.15
To fill the void, Watson concluded that IBM should build its
own computational facility. He contacted Wallace Eckert with
whom he had enjoyed congenial cooperation at Columbia Uni­
versity. During the war Eckert had served as director of the
Nautical Almanac at the U.S. Naval Observatory. His first task at
the observatory had been to design an air almanac for the ex­
panding air force. Using equipment rapidly installed by IBM, the
observatory's personnel were able to compute and compile the
data, design the air almanac, and get it approved and into air
force training camps in less than a year.
With the war entering its final phase, Watson suggested Eckert
should consider peacetime employment. In particular he offered
him the opportunity to establish the company's first laboratory
devoted to "pure science." A major emphasis, consistent with
Eckert's own interests, was to be automatic scientific computation.
As part of the assignment, Eckert was to establish specifications
for a "supercalculator" Watson proposed to build. Eckert ac­
cepted the offer and joined IBM in March 1945 . 16
In describing the supercalculator project to his engineers, Wat­
son said its performance must eclipse the Mark I and any ma­
chine Aiken might build in the future. Recognizing that this
objective could not be achieved without significant innovation,
Bryce urged that a small electronic calculating machine be devel­
oped first. The task was assigned to Halsey Dickinson, who had
continued his research on electronic arithmetic circuits through­
out the war-even in his own basement when wartime pressures
made such work inappropriate at IBM. Byron Phelps was assigned
to the project, as was an engineer from Dickinson's department.
They also had the help of two techniciansP
Using circuits developed in Endicott by Palmer and Phelps at
the beginning of the war, the small engineering group had a
production-prototype electronic calculator operating in April
1946-0ne year after Phelps joined the project. The prototype
and the first four production models could perform both multi­
plication and division, but market planners concluded the divi­
sion capability being offered was too limited. Therefore the
capability was eliminated entirely before the machine was
demonstrated in September at the business show in New York
City. Copyrighted Material
124 Building IBM

Announced as the IBM 603 Electronic Multiplier, the machine

was the first electronic calculator ever placed in production. It
contained approximately 300 vacuum tubes. It could multiply two
6-digit numbers read from a card and then punch the 12-digit
product into the same card at a speed of 100 cards per minute.
Because of unexpectedly good reception for the electronic ma­
chine, a decision was made to limit production to one hundred
and, as quickly as possible, to develop a more versatile electronic
machine to replace it. The replacement machine was the highly
successful IBM 604 Electronic Calculating Punch.18 Exploratory
electronic projects encouraged by Br yce before the war, and
reestablished by him near the end of the war, had thus resulted
in the two most successful electronic calculating machines of the
immediate post-World War II era.

The SSEC: A Supercomputer

Meanwhile Watson was busily assembling a team to develop a

supercomputer. Engineering responsibility was assigned early in
1945 to Frank Hamilton, who had served as lead engineer under
Clair Lake for the Mark I project. In August, Robert R. (Rex)
Seeber, a 1932 Harvard graduate with an interest in machine
computation, joined Eckert's staff and was assigned to work with
Hamilton. As his involvement grew, Seeber became the chief
architect.
Previously Seeber had spent a year writing programs for the
Mark I, directly under navy commander Howard Aiken. It was not
a happy relationship. Their personalities clashed, and Aiken re­
fused to consider Seeber's proposal to endow the next Harvard
machine (Mark II) with provisions for modifying its instructions
during operation-a feature that later became standard in com­
puters. Seeber gladly joined Eckert, who offered him consider­
able freedom in defining the capabilities of Watson's proposed
machine.
Specifications for the supercomputer were developed by Ham­
ilton and Seeber and approved by Eckert in March 1946. Use of
electronic arithmetic circuits like those in the IBM 603 was key
to the high performance of the machine. To implement these
circuits, Byron Phelps was transferred back to Endicott after
finishing his work on the 603.19
Completed and tested in Endicott in the summer of 1947, the
supercomputer was d�iJ � a m�rflJJipped to the company's
 Preparing for Peace 125

Figure 10.1. The SSEC: A Supercomputer

Top: The SSEC (Selective Sequence Electronic Calculator) was dedicated in
January 1948. It occupied the periphery of a room 60 feet long and 30 feet
wide. Along the left wall are panels of vacuum tube circuits for card reading
and sequence control and thirty-six paper-tape readers for table lookup.
Along the back wall are the three paper-tape punches and the thirty paper­
tape readers used for information storage. Along the right wall are the elec­
tronic arithmetic and storage units. Card readers, card punches, and printers
can be seen in the center of the room, but the operator's console is ob­
scured by the pillars. Bottom: Standing behind the operator's console are
from left: Robert R Seeber, Wallace J. Eckert, Thomas J. Watson, and Frank
E. Hamilton. Eckert established specifications for the SSEC, and Seeber and
Hamilton served as the chief architect and chief engineer, respectively.

Copyrighted Material
126 Building IBM

headquarters at 590 Madison Avenue where it was reassembled,

tested, and publicly dedicated in January 1948 as the IBM Selec­
tive Sequence Electronic Calculator (SSEC). The SSEC contained
21,400 relays and 12,500 vacuum tubes. Its electronic arithmetic
circuits could perform 14-by-l4-digit decimal multiplication in 20
milliseconds, division in 33 milliseconds, and addition or subtrac­
tion of 19-digit numbers in only 0.3 milliseconds. The SSEC was
more than 250 times as fast as the Mark I, whose electromechani­
cal arithmetic unit required 6 seconds for multiplication and 16
seconds for division.
The SSEC could store up to 400,000 decimal digits in a hierar­
chical memory system consisting of 160 digits of high-speed elec­
tronic storage, 3,000 digits of intermediate-speed relay storage,
and the remainder in long paper tapes made of standard IBM
punched-card stock. About a quarter of the tape storage was in
continuous loops mounted on up to 36 tape readers. The loops
contained prepunched mathematical function tables, programs,
and input data. Three additional storage units used blank rolls
of paper tape. Consisting of a punching station followed by 10
reading stations, these units could store intermediate results for
later reference.20
As in later computers, the SSEC represented, moved, and
stored instructions in the same manner as data. Punched-card
machines had long been able to sense the sign of the amount in
an accumulator and take alternative action when it changed. For
the SSEC, Seeber extended the idea (albeit awkwardly) to permit
branching to a different subprogram, thus enabling the SSEC to
execute program loops. It was the first operational computer
capable of modifying its program based on intermediate results.
Although the SSEC was the most powerful computing machine
of the time, its influence on later computer designs was limited
by external developments (as discussed in chap. 11). It did, how­
ever, provide IBM with important patent coverage. and it played
a significant role in computer developments by facilitating the
early training of programmers inside and outside the company.
The programming staff established and directed by Seeber, for
example. included John Backus, who later developed FORTRAN,
and Edgar F. (Ted) Codd, who later became known as the "origi­
nator of the relational model for databases. "21
Dedicated by Watson "to the use of science throughout the
world," the SSEC was operated on a nonprofit basis, with no
charge for computations oertaiuio£'. to "pure science." Wallace
c.;0PYflgntea MMena,
 Preparing for Peace 127

Eckert was, of course, a major user. His accurate calculations of

the moon's positions for the period 1952-1971 became the world
standard when published in 1949. Fifteen years later his work
provided the basis for orbital calculations made to guide the first
landing on the moon.22

The Watson Labaratory

Wallace Eckert's decision to leave the prestigious position at the

Naval Observatory and join IBM in March 1945 was influenced
by many years of association with Watson. He knew that Watson
was genuinely interested in supporting scientific work and that
Watson believed a company that reaped the benefits of science
should contribute to science. Eckert was also pleased that Watson
urged him to maintain his own credentials as an astronomer so
he could retain-and his laboratory could gain-respect and
leadership in the scientific community. Eckert, incidentally, was
the first IBM employee with a doctorate.
Concerning Watson's desire that the laboratory be devoted to
"pure science," Eckert provided the following clarification: "By
pure science we mean scientific research where the problem is
dictated by the interest in the problem and not by external
considerations." Consistent with this objective, Eckert located at
Columbia University to facilitate interaction with faculty scientists.
Temporarily placed in Pupin Hall, the laboratory moved in No­
vember 1945 to a townhouse at 612 West 116th Street where space
could be provided for computing equipment, a library, reception
lounge, a small machine shop, and two to three dozen profession­
als. Officially designated the Watson Scientific Computing Labo­
ratory at Columbia University, it was commonly referred to as the
Watson Lab.23
Among the early employees was Llewellyn H. Thomas, a world­
renowned mathematician and theoretical physicist who is best
known for the Thomas-Fermi approximation for describing de­
generate electron gases and for the Thomas precession of elec­
tron spins. Thomas left Ohio State University to join the new
laboratory because of his interest in adapting mathematical prob­
lems to automatic calculating machines and the promised free­
dom to pursue other research topics of interest to him. At the
Watson Laboratory he continued to be a prolific contributor to
science while also finding time to involve himself in practical
engineering problems.
Copyrighted Material
128 Building IBM

Figure 10.2. Watson Laboratory

The Watson Laboratory was located in this five-story townhouse at 612 West
I 16th Street in New York City beginning in 1945. It was relocated to a larger
building on 612 West 115th Street in 1953. There it remained until 1970
when the laboratory's activities were transferred to the Thomas J. Watson Re·
search Center in Yorktown Heights, N.Y.

Copyrighted Material
 Preparing far Peace 129

To identify and attract Thomas and other top talent, Eckert

obtained the help of I. I. Rabi, a Nobel Prize-winning member
of the Columbia University physics department who had served
as associate director of the MIT Radiation Laborator y during the
war. Rabi was especially helpful in locating much needed special­
ists in electronics.24
Watson's view of the role of the new laboratory is conveyed by
the official news release when it was opened: "The research and
instructional resources of the laborator y will be made available
to scientists, universities, and research organizations in this coun­
try and abroad, and special cooperative arrangements will be
made with scholarly institutions. " In addition to their primary
activities of devising new methods of automatic computation and
applying them to scientific problems, laboratory personnel un­
dertook the task of educating others. A course tided "Three-Week
Course on Computing" started in 1947 to teach the use of
punched-card equipment for mathematical calculations �s ulti­
mately attended by approximately 1,600 people from twenty
countries. 25
Because IBM's first laborator y devoted to pure science was
established in the same y ear in which Vannevar Bush issued his
report to President Harr y S. Truman on Science, the Endless Fron­
tier, there is a tendency to link the two events more closely than
is warranted.26 The decision to create the Watson Laborator y in
New York City in 1945 appears to have been driven almost exclu­
sively by internal· considerations and competitive pressures, in­
cluding the loss of good relations with Howard Aiken at Harvard
and the excellent relationship Watson had enjoyed with Wallace
Eckert since long before World War II began.

Copyrighted Material
11
Government-Funded Competition

Watson did not become complacent after World War II, although
he had ample reason to do so. The company's gross revenues
dropped only about 15 percent during 1946, despite lost sales of
munitions. At the same time, profits jumped by 70 percent. Dur­
ing 1947 revenues returned to their peak level and profits were
nearly three times those of wartime. (See appendix A.) More
secure than ever was IBM's position as the leading supplier of
punched-card equipment for record keeping, accounting, and
other business applications.
Also strengthened during the war was the company's leadership
in digital computation methods and equipment. Some specialized
electronic devices for code breaking were in use during the war,
but no electronic computing devices were completed in time. The
ENIAC, for example, was not ready for test operation until late
in 1945. Thus the burden of wartime computational activities fell
primarily on IBM equipment. It was used to analyze airframe
designs, to predict artillery shell trajectories, to break enemy
codes, to support the atomic bomb development, and to perform
a range of other tasks that defy easy description. 1
Mter the war in September 1946, IBM introduced the indus­
try's first product to employ electronic digital computation, the
IBM 603 Electronic Multiplier. Early that same year, upon his
return from the navy, Ralph Palmer established an electronic
engineering effort in Poughkeepsie. His highest priority was to
develop a successor to the 603. The resulting IBM 604 Electronic
Calculating Punch was first shipped to customers in the fall of
1948. It provided speed and flexibility of operation unmatched
by any calculator in the marketplace for sometime.2
Watson's primary competitor was still Remington Rand. Its
punched-card equipment business was only one-tenth the size of
Copyrighted Material
132 Building IBM

IBM's, and it had no experience in electronic computing devices. 3

From where might any significant challenge come?

Sounding the Alarm

The primary postwar challenge was generated unexpectedly by

government actions directed by individuals who had been respon­
sible for harnessing advanced technologies during the war. Key
among these individuals was Vannevar Bush, a one-time MIT
professor whose differential analyzers had been used extensively
during the war for calculating ballistic firing tables.4 President of
the Carnegie Institution during the war, Bush also served as head
of the Office of Scientific Research and Development (OSRD).
His mission was to organize the country's technical resources for
military purposes. In fact but not title, Bush was the chief adviser
to President Roosevelt on matters of science, engineering, and
technology.
At war's end, Bush and his colleagues successfully sought to
preserve the country's leadership in technologies crucial to na­
tional security. As a result, federal money continued to be spent
in postwar years on research and development projects in univer­
sity and industrial laboratories. This eased the transition from war
to peace, contributed to the country's military preparedness, and
stimulated economic growth. Bush emphasized electronics-es­
pecially its application to automatic computation.5
Only gradually did IBM leaders become aware of these events,
many of which were shrouded in secrecy appropriate to national
security matters. John McPherson, with his technical expertise
and federal contacts, was the first to sense the government's new
peacetime role. By September 1946 his concern over the size and
number of government-funded projects in electronic computing
led him to write to Wallace Eckert, urging Eckert to reevaluate
the adequacy of his laboratory's program. 6 "In view of the vast
program which the government is now sponsoring," Eckert re­
sponded, "I feel that the present tempo at the Watson Laboratory
is not adequate for IBM, and a greatly accelerated program of
development of the various phases of the new electronic tech­
niques should be undertaken."7
In November the two of them took the unusual step of writing
a joint letter to Watson to advise him of their concerns. "As an
outgrowth of the wartime demand for high speed computing,"
the letter began, "the Natiooal Bureau of Standards has em-
-Copyngnrea Marenal
 Government-Funded Competition 133

barked on a program of development of electronic digital calcu­

lating devices with funds from the Departments of Commerce,
War and Navy. Part of this development will be carried on in the
Bureau, and part by means of development contracts with outside
individuals and organizations. These development contracts are
of such nature that they will be very attractive to anyone without
previous private experience or patents in the computing field;
but the patent provisions make it doubtful if IBM, which has led
in the field, can afford to participate in the program."
The army and navy were independently financing major re­
search projects on electronic computing at the University of
Pennsylvania, the Institute for Advanced Study, Harvard Univer­
sity, MIT, and other places according to the letter, which further
noted that "several commercial concerns in addition to IBM have
been developing electronic computing machines."
"Whereas before the war IBM was the only organization able
and willing to carry on large scale development of calculators,
such development is now taking place on a large scale," they
advised Watson. The letter provided disturbing information, but
Eckert and McPherson avoided recommending specific action.
Instead they observed, ''The complete answer to this situation is
not obvious since questions of basic policy are involved."8

The ENIAC

Among the many government-sponsored activities in digital com­

puters, a prqject at the Moore School of Electrical Engineering
of the University of Pennsylvania achieved the greatest long-term
impact. Sponsored by the Army Ballistic Research Laborator y, it
had the goal of building an electronic computing machine for
wartime ballistics calculations. Work began in the summer of
1943, and the resultant ENIAC (Electronic Numerical Integrator
and Computer) was ready for its first test runs in late 1945, some
months after the surrender of Japan. The ENlAC was formally
unveiled at a dedication ceremony at the Moore School in Feb­
ruar y 1946.
The senior member of the development team, John W.
Mauchly, had proposed the project. His interest in electronic
computation began well before his 1941 visit to Iowa State College
(discussed in chap. 7) during which he had lengthy discussions
with John Atanasoff about the latter's partially completed elec­
tronic computer.9 Sec.Dnd iJl corpmanq was J. Presper Eckert, Jr.,
c;opynghtea Matenal
134 Building IBM

Figure 11.1. The ENIAC

Left: Rear view of a small portion of the ENIAC reveals some of its 18,000
vacuum tubes. Right: Front view reveals some of the plug wires and switches
used to program the machine. (Courtesy of Smithsonian Institution.)

a twenty-two-year-old graduate student at the Moore School when

he first met Mauchly in 1941. Much of the credit for the success
of ENIAC belongs to Eckert, who was a brilliant and innovative
engineer. 10
The ENlAC was a huge room-sized machine. Its 18,000 vacuum
tubes made it the largest electronic computing device in exist­
ence. So many tubes also created a severe operating problem.
Seldom did a day go by without replacement of burned-out tubes.
Although a faulty tube could be replaced in minutes, locating it
often took hours. The ENIAC had 40 percent more vacuum tubes
and was somewhat faster in individual operations than IBM's
SSEC, which was completed two years later. For diverse calcula­
tions, however, the ENIAC was no match for the SSEC because it
lacked storage capacity and required time-consuming resetting of
many switches and the rewiring of large plugboards to shift from
one job to another. II
Operation of the ENIAC was described by one early user as
follows: "We had about 40 plugboards, each several feet in size.
A number of wires had to be plugged for each single instruction
Copyrighted Material
 Government-Funded Competition 135

of a problem, thousands of them each time a problem was to

begin a run; and this took several days to do and many more days
to check out. When that was finally accomplished, we would run
the problem as long as possible, i.e. as long as we had input data."
Reprogramming the machine for a new problem was typically
undertaken only after several weeks of operation. 12
The deficiencies of the ENlAC design became increasingly
evident to Eckert and Mauchly, even before John von Neumann
became involved in August 1944. A distinguished Hungarian
mathematician who emigrated from Germany to the United
States in 1930, von Neumann received a permanent appointment
three years later to the newly established Institute for Advanced
Study in Princeton, New Jersey. During the war he consulted for
numerous government organizations and became familiar with
the use of punched-card equipment for mathematical applica­
tions.IS Von Neumann shared the feeling that ENlAC's plug-wire
control and dearth of internal storage capacity were major
deficiencies.

Stored-Program Computers

Through the interactions of Eckert, Mauchly, von Neumann, and

others, a design emerged of a more advanced machine called the
EDVAC (Electronic Discrete Variable Automatic Computer) . The
generous support given the ENIAC project can be perceived from
the fact that the Army Ordnance Department in January 1945
granted a ten-month supplement of $105,000 just for research
and development on the EDVAC. Much of this money was spent
for developing a mercur y delay-line memory to provide the large­
capacity, high-speed internal storage needed to supply data and
instructions to the electronic arithmetic units. 14
While Eckert and Mauchly concentrated on electronic devices
and engineering aspects of the machine, von Neumann focused
on machine properties of direct concern to a programmer-on
what today is called architecture. Perhaps more than anyone else
to that time, he "gave a logical treatment to the subject, much as
if it were a conventional branch of logics or mathematics."15 By
separating the logic of machine operation from matters of prac­
tical hardware implementation, von Neumann was able to identify
computer design principles that outlived the hardware issues of
the day.

Copyrighted Material
136 Building IBM

Computers adhering to the design philosophy he espoused

came to be known, years later, as stored-program computers. Using
von Neumann's neurological terminology, the critical storage
device was called memory-and still later main memory to distin­
guish it from larger-capacity and slower storage devices that were
also attached to the central processing unit. In stored-program
computers, typically a single memory stores both instructions and
data so that program sequences can be made available as rapidly
as data. Furthermore, by treating instructions as data, the se­
quence of calculations can be altered automatically during a long
computation, based on intermediate results.16
The IBM designers of the SSEC had also begun their work in
1945, but they were constrained to use existing technologies in
order to complete their machine as quickly as possible. Thus the
sequence control of the SSEC made use of an internal memory
hierarchy consisting of high-speed vacuum tubes, intermediate­
speed electromechanical relays, and paper tape storage devices
offering large capacities at relatively low cost.
The SSEC was the first operational computer to satisfy the
modern definition of a stored-program computer. Although its
architecture lacked the elegance and simplicity achieved by the
EDVAC's designers, a patent based on the SSEC and filed in
Januar y 1949 provided IBM with primary patent rights on the
stored-program computer.17 All claims in this patent were sub­
sequently upheld during patent-interference proceedings with
Sperry Rand.ls
Designed and built under heavy pressure for rapid completion
and without benefit from the Moore School work or the broader
intellectual discussions led by John von Neumann, the SSEC
was in full operation in January 1948, well over a year before the
first computer based on the Moore School design. Ironically
the Moore School's EDVAC was not operational until late in
1951.19

Promoting the Concept

Long before the first EDVAC-type computer was completed, the

influence of EDVAC on future developments was assured by a
series of events. The first of these events was the distribution in
June 1945 of a "First Draft of a Report on the EDVAC," written
by von Neumann. Why this remarkable but unfinished report was
distributed with von �8 �RMJi�? Mlhe sole author, with no
P e
 Government-Funded Competition 137

credit to others, is surrounded by controversy. But its initial

impact is clear. It ended the productive and harmonious relation­
ship at the Moore School.
Von Neumann's use of neurological analogies and terms out­
raged Eckert because it enabled von Neumann to give unclass­
ified talks about work at the Moore School "without giving any
credit" to anyone else. "I was too young to know how to fight back
against this type of behavior," Eckert recalls.20 Meanwhile Eckert
and Mauchly were not able to talk about their work because it
was performed under a government security classification of
"confidential." Their first detailed engineering progress report
on the EDVAC (completed three months after von Neumann's
highly abstract, theoretical treatment) was available primarily to
project participants and contract administrators.21
Herman H. Goldstine, an army officer and mathematician who
served as the Ballistic Research Laboratory's representative to the
ENIAC project, says the First Draft document was intended as a
''working paper for use in clarifYing and coordinating the think­
ing of the group." It was initially distributed only to people closely
associated with the project. The content was soon deemed so
significant, however, that knowledge of the document spread, and
Goldstine gave copies to many who requested them. Thus, accord­
ing to Goldstine, the document was given broad distribution
without von Neumann's knowledge.22
The document was not only a significant technical contribu­
tion, but it also became an important tool in von Neumann's
effort to establish his own computer project at the Institute for
Advanced Study (lAS). Seeking support from a variety of sources,
he ultimately established his lAS Computer project with $300,000
provided in three equal parts by the institute itself and by the
ordnance departments of the army and navy. He also entered into
a joint research and development contract with RCA to develop
a computer memory. The RCA device, known as the selectron,
was based on a specially designed vacuum tube that stored infor­
mation internally in an array of metallic eyelets, the voltages of
which were set to represent logical Is or Os by the selective
application of electron beam bombardment and electric currents.
The selectron was expected to provide even faster access to stored
information than the delay-line memory being developed by Eck­
ert for the EDVAC.23
Just as important as von Neumann's conceptualization of the
stored-program comQ%pyr191\�eW MHf&mption of it. He attended
138 Building IBM

technical meetings, visited research laboratories, and presented

his ideas in private and public forums. Following his initiation of
the lAS Computer project, it became a focal point for the ex­
change of ideas on electronic computing. As a result of his activi­
ties, a significant degree of design standardization was achieved,
and numerous projects were initiated to design and build stored­
program computers. As reported by John McPherson following
discussions at the Aberdeen Proving Ground in September 1945,
"probably four organizations [are] building machines of this type
simultaneously-a machine at the Institute for Advanced Study
in Princeton, a machine at the University of Pennsylvania, a
machine in England, and probably one at the Naval Ordnance
Laboratory in Washington. "24
A particularly important event was an eight-week Moore School
course held in the summer of 1946 on the "Theory and Tech­
niques for Design of Electronic Digital Computers." Jointly spon­
sored by the Office of Naval Research and the Army Ordnance
Department, the purpose of the course was to disseminate infor­
mation about electronic digital computers and about the oppor­
tunities for government funding. Eckert, Mauchly, von Neumann,
and others presented lectures. Invitees included representatives
from Columbia University, Harvard University, the Institute for
Advanced Study, MIT, AT&T, Eastman-Kodak, General Electric,
IBM, and National Cash Register as well as from the National
Bureau of Standards and other government organizations.25
The invitation to IBM was withdrawn shortly before the course
began. The official reason given was that the company was not
among those receiving government funds to work on electronic
computing.26 A more cogent reason may well have related to
recent decisions of two key lecturers. Eckert and Mauchly had
decided to leave the Moore School and establish their own com­
pany in direct competition with IBM.27

Commercial Ventures

A new patent policy of the Moore School was a major factor in

the decision of Eckert and Mauchly to leave the University of
Pennsylvania. Henceforth, employees would be required to assign
to the university all patents resulting from work done while in its
employ. Similar policies were common in industry, where employ­
ees typically assigned their patents to the company that paid their
salaries and the othercHpVfigdre8 ��»131
 Government-Funded Competition 139

The primar y intent of the university was to maintain the "intel­

lectual purity" of the research it sponsored rather than having
research driven by a desire for personal gain. The policy was not,
however, acceptable to Eckert and Mauchly. They were motivated
by building useful computing devices-and making money-as
well as by the purely intellectual aspects of their work. Indeed
they had already explored potential funding for the construction
of machines for the Bureau of the Census. Refusing to sign the
patent release, the two men were forced to resign at the end of
March 1946. 28
In addition to the various opportunities arising from govern­
ment-funded projects, Eckert and Mauchly received an offer from
Thomas Watson. He agreed to establish a computing laborator y
for them at IBM. Eckert was very interested because it would have
permitted him to develop electronic computer technology inde­
pendent of institutional constraints, but Mauchly was reluctant.
He believed he and Eckert could establish their own very
profitable business. In the end Mauchly's view prevailed.29
In June 1946 they formed the Electronic Control Company,
later called the Eckert-Mauchly Computer Corporation. It was the
first company established specifically to design, manufacture, and
sell electronic stored-program computers. By the fall of 1949 the
company gave the appearance of a thriving organization. It had
134 employees and six contracts for its large, partially developed
computer, now called the UNIVAC (UNIVersal Automatic Com­
puter). In fact, however, the Eckert-Mauchly Computer Corpora­
tion was in deep trouble.30
Initial problems were created when the National Bureau of
Standards, which had broad responsibility for government fund­
ing of computer projects, received a less than enthusiastic evalu­
ation of the Eckert-Mauchly project from a highly respected
expert. As a result, only a study contract for $75,000 was initially
offered. A contract for $170,000 to complete the computer design
for the Census Bureau was not signed until June 1948, almost two
years later. Compounding and contributing to this financial prob­
lem, Eckert and Mauchly initially had little comprehension of the
difference between developing a commercial product and creat­
ing a one-of-a-kind device such as the ENIAC. They also had no
business experience.31
Running late on their government work and short of funds,
they chose to obtain additional money and development experi­
ence by entering intcco �Rp,RNPMlitD1f.jtCt to build a small com-
140 Building IBM

Figure 11.2. Mauchly, Groves, and Eckert with Computer Memory

John W. Mauchly (left) and.J. Presper Eckert. Jr. (right), are shown discuss­
ing UNIVAC's mercury delay-line memory and its test facility with Gen.
Leslie R. Groves (center). The memory is the large metallic cylinder jutting
out from the bottom of the rack. (Courtesy of Smithsonian Institution.)

Copyrighted Material
 Government-Funded Competition 141

puter (BINAC) for the Northrop Aircraft Company. Delivered in

September 1949, the small computer had taken three times
longer than the projected eight months to develop, had cost
almost three times the contracted price, and fell far short of
expectations in reliability and performance.32
Already behind schedule on all commitments and out of money
in the fall of 1948, the company achieved a temporary solution
when the American Totalisator Company agreed to provide ap­
proximately $500,000 for 40 percent ownership. The decision was
based on the belief that electronic devices migh t some day replace
the relay-operated totalisators used at race tracks. Death in a
private plane crash of the American Totalisator vice president
who had arranged the deal, combined with escalating develop­
ment problems, made further fund-raising necessar y in the fall
of 1949.33
Unable to get additional loans, Eckert and Mauchly sold their
business to Remington Rand in February 1950. The purchase
marked Remington Rand's entry into the field of electronic com­
puters. The Eckert-Mauchly Computer Corporation became a
wholly owned subsidiar y and functioned as an independent divi­
sion. In March 1951 the Census Bureau accepted the first
UNIVAC, and three more were delivered within eighteen
months.34
During the years in which Eckert and Mauchly were barely
surviving, Engineering Research Associates (ERA) was thriving.
Seriously undercapitalized, it nevertheless managed to grow with
the solicitous help of the National Security Agency, whose prede­
cessor organizations had helped create it. In fiscal year 1947
employment rose from 145 to 420 as ERA grew its staff to the size
needed to develop the top-secret, special-purpose electronic de­
vices needed for code breaking and other national security
activities.
In August 1947 ERA was given the task of designing a general­
purpose, stored-program computer to carry out many activities
previously handled by special-purpose equipment. Seven months
later the navy approved the design and authorized construction.
Code-named Atlas, after the mental giant in the comic strip,
"Barnaby," it was a stored-program computer similar to the ED­
VAC. Its main memor y was provided by a rotating drum with
magnetic recording media on the cylindrical surface-a technol­
ogy pioneered by ERA after the war. Atlas contained 2,700 vac­
uum tubes and 2,385 crystal diodes. It was delivered to the
Copyrighted Material
142 Building IBM

government in Washington, D.C., and successfully operated in

December 1950.
A modified version, called the ERA llOl, was cleared by the
navy for sale as a commercial product and announced in Decem­
ber 1951. No commercial orders were received, however, primar­
ily because the computer had very limited input-output facilities,
no operating or program manuals were available, and there was
no formal marketing effort. Furthermore the company lacked the
financial resources to compete in the commercial market.
John Parker, who had arranged ERA's initial financing­
$20,000 equity and a $200,000 line of credit-had long recog­
nized the constraints imposed by this meager funding.
Accordingly he had been in contact with potential buyers for
several years. In December 1951 he announced that an offer had
been accepted for the purchase of the company by Remington
Rand. Consummation of the deal was delayed until reviews by the
Federal Trade Commission and theJustice Department were com­
pleted. Approval was granted in May 1952. The value of ERA was
set at $1.7 million. Government sponsorship had thus helped
ERA shareholders receive an 85-fold return on their investment
in only about six years.35
The purchase of ERA, combined with the earlier purchase of
EMCC, made Remington Rand the undisputed leader in large­
scale electronic computers. Government funding of electronic
computer development had resulted in a massive concentration
of expertise in electronics and digital computing in Remington
Rand-a company that would continue to provide IBM with its
primary competition for another two decades.36

Provocative Terminology

While promoting the stored-program design concept, von Neu­

mann also promulgated his neurological terminology for parts of
the machine. The high-speed storage device, for example, be­
came known as "memory." Presper Eckert has suggested this
practice was merely a ploy by which von Neumann avoided gov­
ernment classification restrictions. More than likely, however, von
Neumann was attracted to a terminology that likened a machine
to a human being. In this he was not alone. Eckert and Mauchly,
themselves, chose to call their machine a computer-a word pri­
marily used at the time to designate a person who did computing
Copyrighted Material
 Government-Funded Competition 143

with the help of calculating equipment. By calling their ma­

chine a computer, Eckert and Mauchly were con juring up the
perception that ENIAC had the intellectual capabilities of a
person.
Prominent among those resisting such terminology was
Thomas Watson. He well understood the emotional and sales
impact of the term, but he was selling products, not ideas. The
public had often resisted new machinery that was difficult to
comprehend. frightening, or likely to replace people in the work­
place. What could be more frightening or likely to replace people
than an electronic "computer." Watson wanted to convey the
message that IBM's machines were not designed to replace peo­
ple. Rather they were designed to help people, by relieving them
of drudgery.
It was Watson's policy, and therefore IBM's, to refer to such
machines as "automatic calculators" or "sequence controlled cal­
culators." Thus the world's first large-scale digital computer, the
IBM-Har vard Mark I, was officially named the IBM Automatic
Sequence Controlled Calculator, and the first stored-program
computer was called the IBM Selective Sequence Electronic Cal­
culator. Even IBM's first electronic product designed and mar­
keted specifically to meet the computing needs of engineers and
scientists was called the Card Programmed Calculator rather than
the Card Programmed Computer.
The battle for naming computing equipment was not won by
IBM, however. The emotional appeal and simplicity of the word
computer could not be resisted. All that resulted from Watson's
effort was a common misperception that early machines that were
called calculators must have lacked some important quality of a
computer.

Copyrighted Material
12
IBM� Initial Response

"For IBM to keep its position in the business world. it seems to

me that we must somehow pass on that spirit of belief in our
enterprise and the will to go ahead. Our people must know and
believe that they cannot stand still. "\ So wrote James Bryce to
Watson in April 1 948. It was a time when tensions were mounting
between those advocating greater use of electronics in company
products and those who favored continuing with electromechani­
cal devices because they were fast enough for most applications.
more reliable. less expensive. and-above all-more familiar.
Bryce had worked more than thirty years with Watson to
strengthen the product line and to broaden its applications. With
Watson he had shared in the success of that endeavor. Having
promoted the use of electronics in test equipment and products
for many years . Bryce was particularly pleased to witness the 1946
introduction of the IBM 603 Electronic Multiplier. the first com­
mercial product to incorporate electronic arithmetic circuits. Too
ill to attend the January 1948 dedication of the SSEC, he could
nevertheless take pride in that event as well. Then in March 1 949,
Bryce died of a cerebral hemorrhage. He was sixty-eight years
01d.2
Attending the funeral at age seventy-five, Watson knew he had
little time left to achieve his own final goal-to transfer leadership
of IBM to his two sons. It was a goal he had given high priority
since the end of the war. The resulting transition profoundly
affected IBM's response to electronic computing, and it produced
one of the more stressful periods in the company's history.

Grooming 1Wo Sons

Watson's choice of positions for his sons was constrained by their

rather different per� �e�aMJlil-bilities. His older son,
146 Building IBM

Thomas J. Watson, Jr., had not been a good student. Imperfect

eye-hand coordination had adversely affected his reading and had
also prevented him from doing well at most sports. Overwhelmed
by his father's successes and frustrated by his own failures , Tom
Watson had suffered periods of depression in his youth. His
academic record in high school was poor, causing him to seek
recognition in the girls he dated, the expensive cars he drove,
and the pranks he played on others.
His first real success was in competitive rowing in which his
team qualified for the international regatta in Henley, England.
His scores on college admission tests were generally low, except
for physics in which he scored remarkably high. With the consid­
erable influence of his father and support from an understanding
dean, he was admitted to and graduated from Brown University.3
Joining IBM as a sales trainee in 1 937, Watson,Jr., enjoyed the
informality of being called Tom Watson, or simply Tom. By con­
trast his father was always addressed as Mr. Watson, even by his
closest associates. (Consistent with that practice, this book re­
serves the less formal name, Tom Watson, for the son.) Graduat­
ing from sales school in Endicott, the young Watson was
immediately given a prime sales territory in Manhattan. He did
well, but he could never get out from under his father's shadow.
In his own mind he believed his successes were orchestrated by
his father, whereas his failures were his own. Only in flying air­
planes, a hobby he took up as a freshman at Brown, did he
achieve the personal accomplishment he longed for. Part of the
triumph was overcoming his eye-hand coordination limitations.
Eager to escape his father's influence and wishing to fly military
airplanes, Watson considered joining the Army Air Corps, but
feared he would fail the eye exam. Learning that his civilian flying
experience would qualify him for the Air National Guard without
attending flight school or taking the eye exam, he signed up
immediately and soon obtained his commission as a second lieu­
tenant. When President Roosevelt mobilized the National Guard
in September 1940, Tom Watson had achieved his dream. He had
become a full-fledged military pilot.4
Wat'ion did a lot of flying, and he distinguished himself in a
variety of other assignments. He promoted the use of Link train­
ers to improve pilot training; he served as aide-de-camp to the
head of the First Air Force; and he helped establish a ferry route
to deliver airplanes to the Soviet Union . In the air corps he
became his own man. He achieved the reputation of a self-asscr-
Copyrighted Material
 IBM's Initial Response 147

tive, strong-willed young officer, capable of accomplishing

difficult assignments.5
Discharged from the Army Air Corps at the end of 1945, he
reported back for work at IBM on the first business day of 1946.
His initial assignment was as assistant to Charles A. Kirk, the
forty-one-year-old executive vice president who had previously
served with distinction as vice president for manufacturing. Kirk's
new task was to educate Tom Watson in IBM's business and
prepare him for a leadership position. The young Watson found
Kirk to be an aggressive executive who had a remarkable grasp
of the business and an ability to size up situations quickly and
make good decisions. He admired the way Kirk taught him the
business without any apparent concern that the two might some­
day compete for the top position. Watson by contrast was deeply
concerned that his mentor might stand between him and the
presidency of IBM.

Figure 12.1. Tom Watson meets Charlie Kirk

While still in the air corps, Tom Watson met Charles A. Kirk for the first
time during an unofficial stopover in Endicott. When Tom returned from ac­
tive duty, his father assigned Kirk to train him for a leadership position in
IBM.

Copyrighted Material
148 Building IBM

"All I could think of," Watson, Jr., recalled, was that "if Dad ever
got sick or died, Kirk was the logical successor."6 He therefore set
about to master all the skills Kirk had, to learn all that Kirk knew,
and to outperform him in every possible way. Conflicts between
the two became common. With young Watson's private track to
the top, it was not a fair match. In June 1 946, at age thirty-two,
Tom Watson was promoted to vice president, and in October he
was elected to the board of directors. In May of the following year,
Kirk suffered a coronary thrombosis and died while on a business
trip with Tom Watson in Europe. 7 In September 1949 Tom Wat­
son was promoted to executive vice president and J. George
Phillips, who had served as executive vice president following
Kirk's death, became president of IBM. Watson, Sr., still retained
ultimate authority as chairman and chief executive officer.s
In sharp contrast with his older brother, Arthur K. (Dick)
Watson was an excellent student. At Yale he majored in interna­
tional affairs and did particularly well in foreign languages. When
his senior-year studies were interrupted by the war, he applied for
European duty but was assigned to the Ordnance Corps in Aber­
deen, Maryland, as an instructor. Sent to the Philippines near the
end of the war, he was discharged as a major in 1947. Returning
immediately to Yale, he earned his bachelor's degree and joined
IBM as a sales trainee in the same year.
In 1948 Dick Watson joined his father and a group of executives
on a trip to Europe that led to the establishment of the IBM World
Trade Corporation in October 1949. T he senior Watson's plan
was to operate the company's international trade entirely through
this new corporation. A national company, wholly owned by the
World Trade Corporation, would be established in each country
where IBM did substantial business and would be managed and
staffed primarily by citizens of that country. The national compa­
nies were to provide flexibility for adapting to local conditions
and customs, while the World Trade Corporation would optimize
global operations and assure the quality of IBM products and
ser vices worldwide.
Having served as assistant to his father and translator for the
group on the European trip that helped create the World Trade
Corporation, Dick Watson was appointed a vice president and
director. Thus by the end of 1 949 (the year Bryce died) the
seventy-five-year-old Watson had placed his two sons, aged thirty­
five and thirty, in positions that virtually assured their leadership
of IBM.9
Copyng ' hted MatenaI
.
 IBM's Initial Response 149

Assignment in Electronics

While still under Kirk's tutelage, Tom Watson was introduced to

two pioneering electronic projects that provided direction to his
career. First was the ENIAC, which he and Kirk visited early in
1 946. Kirk proposed the visit because he was curious about the
machine and wanted to learn more about the patent situation.
During the visit, Presper Eckert advised them that he and
Mauchly were planning to obtain patents and go into competition
with IBM. At the time neither Tom Watson nor Charlie Kirk
believed that the huge machine posed any particular threat.
"I never stopped to think what would happen if the speed of
electronic circuits could be harnessed for commercial use," Wat­
son later admitted. Fortunately for him, others at IBM had already
given that possibility considerable thought. Within a few weeks
after seeing the ENIAC, he was brought into Bryce's laboratory
by his father to see the company's top-secret electronic project.
A product prototype with electronic arithmetic circuits was al­
ready in full operation. Unlike the enormous ENIAC, the IBM
prototype was a relatively small machine that nevertheless could
perform simple arithmetic with electronic speed.
Tom Watson was impressed. Here was an electronic machine
that could be used in IBM's business. He recalls urging that the
unit, soon named the IBM 603 Electronic Multiplier, be marketed
as "the world's first commercial electronic calculator" even if only
a few could be sold.lO The record is replete, however, with reasons
to believe that Watson, Sr., needed no urging from his son to
market the 603. As early as October 1 943, for example, he had
expressed concern that IBM was not doing enough in electronics
and urged his engineers to use electronics in new products then
being developed. II
By not emphasizing his own early role, Watson, Sr., provided
his son with ownership of the company's thrust into electronics.
The master salesman had made one of the more important sales
of his life. His son had chosen to be responsible for the company's
activities in electronics. The senior Watson could now concentrate
on traditional EAM products and a planned expansion of the
electric typewriter business.
With his career inextricably tied to electronics, Tom Watson
had no intention of letting anything stand in his way, including
his father. Those who witnessed the ensuing battles between the
two-or made the mis.take .oft2'ettin2' between them-have con-
/
c.;opyngfj; (fa Marena
150 Building IBM

eluded that Watson, Sr., resisted the entry into electronics. The
record fails to support that view, however. Rather it reveals a
father who enticed his son into taking responsibility for a new
technology that his technical advisers, James Bryce and John
McPherson, assured him had great potential. Having done so, the
older Watson had the unenviable task of balancing the company's
priorities and tempering his son's otherwise unconstrained drive
to make the company's thrust into electronics a resounding suc­
cess. In recalling his many battles with his father, Watson, Jr., has
said, "Electronics was the only major issue on which we didn't
fight. "12

Getting Started in Poughkeepsie

A key element of the senior Watson's postwar strategy was to

locate the company's more promising advanced development
activities in Poughkeepsie. The resulting new products would be
manufactured in the plant established there for wartime muni­
tions production. Strongly committed to expanding the market
for electric typewriters and to creating modified versions for use
in the EA.\1 line, he had moved the development and manufac­
ture of electric typewriters from Rochester to Poughkeepsie in
1944. 13
When Ralph Palmer returned from the navy in February 1946,
Watson had Charlie Kirk summon him to headquarters to discuss
establishing an electronic research and development laboratory
in Poughkeepsie. Watson knew the opportunity would appeal to
Palmer, who had done pioneering research on electronics in
Endicott before the war. Even before he returned from service,
Palmer had warned John McPherson and others that wartime
advances in electronics were going to put the company at a
serious disadvantage.
"I understand you think IBM is falling way behind in electron­
ics," Kirk began their discussion in a somewhat confrontational
manner. ''That's right," Palmer responded without hesitation.
"Why don't you go to Poughkeepsie," Kirk retorted, "and do
something about it?" The challenge was accepted. Within a mat­
ter of weeks after his discharge, Palmer transferred from Endicott
to Poughkeepsie to establish a laboratory devoted to electronics.14
He began by hiring electrical engineers and initiating explora­
tory studies on electronic arithmetic and storage <;ievices likely to

Copyrighted Material
 IBM's Initial Response 151

Figure 12.2. Ralph L. Palmer

Ralph Palmer is shown (left) at the head table of a dinner honoring a sales
trainee class in Endicott, soon after he was assigned to establish electronic
development facilitie s in Poughkeepsie in 1946.

have value in future product". With Tom Watson's strong encour­

agement, Palmer soon made his primary task the development
of a greatly enhanced replacement for the not-yet·announced 603
Electronic Multiplier. Subjected to closer scrutiny by marketing
people, the limitations of the 603 had become increasingly evi­
dent.15 Nevertheless the first production lot of fifty was sold
almost immediately. A decision was then made to limit production
to one hundred and to introduce the enhanced product as soon
as Palmer could develop it.16
The resulting IBM 604 Electronic Calculating Punch was deliv·
ered to customers beginning in the fall of 1948. It was a major
advance over the 603 and provided computational capabilities
unmatched by any qther commercial product. The 604 offered
fifty decimal digits of internal electronic storage, twice that of the
603. It could perform addition, subtraction, multiplication, and
division with electronic speed. In its most basic version , it could
execute up to twenty plugboard·controlled program steps be­
tween reading data from a card and punching out the result. Up

Copyrighted Material
152 Building IBM

to sixty program steps were possible with optional features. Indi­

vidual program steps could be omitted selectively based on input
data or computational results.
The 604 contained over 1,400 vacuum tubes versus 300 in its
predecessor. The circuit designs were essentially the same as those
devised by Palmer and Phelps before the war, but greater com­
pactness and reduced power consumption was achieved by use of
miniature vacuum tubes developed during the war. Recognizing
the need for tubes with different characteristics and greater reli­
ability than those used in radios, Palmer established a vacuum­
tube research facility. It was located in a small building,
affectionately known as the "pickle factory" because it had been
used as a specialty food processing plant before the war. Here
tubes could be designed and built in sufficient quantity to
influence major suppliers such as General Electric, RCA, and
Sylvania to meet IBM's special requirementsP
Palmer also devised an innovative scheme in which each vac­
uum tube and its closely associated resistors and capacitors were
assembled in a pluggable unit that further increased circuit pack­
aging density. Iii The pluggable unit could be fully tested before
insertion, thus facilitating manufacture. It also aided field service.
By swapping pluggable units within a machine, rather than indi­
vidual vacuum tubes or other electronic components, a service­
man could more quickly identify and replace defective units.19
The rental for the 604 was under $600 per month, well within
the range of sophisticated electromechanical tabulating equip­
ment. So successful was the 604 that annual production peaked
at more than 1 ,000 machines, consuming over 1.5 million vacuum
tubes per year. A total of 5,600 machines were built, installed, and
serviced during a ten-year period.2o

A Commercial Electronic Computer

The success of the IBM 604 was due largely to its ability to reduce
the number of card passes required by accounting tasks and
business analyses, as compared to the number needed in EAM
configurations without a 604. It was also attractive, however, to
the growing number of customers who wanted to use card equip­
ment for scientific and engineering computations. To satisfy the
needs of such customers with the heaviest requirements, the 604
was soon coupled with other EAM equipment in the first IBM
product designed spee lfpyH�1elJ
�mnMtation centers. Formally
 IBM's Initial Response 153

announced as the IBM Card-Programmed Electronic Calculator

in May 1949, it was universally known simply as the CPC.
The CPC contained fewer than one-tenth as many vacuum
tubes as the ENIAC, but it offered more flexibility and ease of
use. Whereas the ENIAC's computational sequences could be
changed only by laborious rewiring, the CPC's sequences con­
sisted of programmed instructions in a deck of punched cards
that could be changed or modified with relative ease. The CPC
executed instructions directly from these cards rather than load­
ing them into memor y as was subsequently done by UNIVAC and
other stored-program electronic computers. A single CPC instruc­
tion could initiate rather lengthy computations by calling on
subroutines prewired in a plugboard.
The genesis of the CPC can be traced to a cooperative effort
between IBM and one of its customers, Northrop Aircraft, Inc.,
of Hawthorne, California. Among Northrop's many uses for EAM
equipment was the calculation of guided-missile trajectories. In­
itially these calculations were made using a standard IBM 601
electric multiplier and 405 accounting machine. Because many
steps in the calculation were dependent on the previous step, the
process was slowed by the need to move card decks continually
back and forth between the two machines.
Without any help from IBM, the engineers at Northrop
speeded this process by interconnecting the two machines so that
results of computations completed by one machine could be
transmitted directly through a cable to the other without moving
cards between them. 12 This arrangement provided considerable
improvement. It was woefully inadequate for truly lengthy inte­
grations, however, even when time-saving methods developed by
Wallace Eckert of the IBM Watson Laboratory were employed.
Desperate for a better solution, the Northrop group arranged
for a discussion of their needs at IBM headquarters in the winter
of 1 947-1948. They had already concluded the solution should
involve the IBM 603 Electronic Multiplier. (The far superior 604
was not yet available. ) As described by a member of the Northrop
team, 'The problem was then taken to IBM Headquarters in New
York where Mr. J. C. McPherson went over the problem require­
ments in detail and agreed with our conclusion with respect to
machine requirements. The machine was built. In fact, we re­
ceived much more than we expected. . . . We received a machine
comparable in programming technique to current large com-
. .1<:NI fir " � 2
puter deSlgn-a poor�i, �nre fNVft!l tenal
154 Building IBM

McPherson '!\ ability to understand the Northrop requirements

and respond rapidly to them was aided by his knowledge of work
at the Watson Laboratory. Here Wallace Eckert had been working
since 1946 on what he called "a baby sequence calculator with
instructions on punched cards." It consisted of a relay calculator
connected with an accounting machine and a specially designed
control box in a configuration somewhat similar to that provided
to Northrop. Eckert appears to have made no attempt to initiate
a product development effort himself. His research activities at
the Watson Laborator y, however, were widely disseminated
through conferences and publications and may have helped
stimulate the Northrop request, although there is no direct evi­
dence for that.23
By the end of 1948, IBM had received more than a dozen
requests for configurations similar to Northrop's, and the reacti­
vated Future Demands Department had undertaken the specifica­
tion of a standard product. Leading the effort was Steve Dunwell,
who had returned from the Army Signal Corps in June 1946 and
had been responsible for specifying the characteristics of the
recently introduced IBM 604.24
Dunwell began by replacing the 603 Electronic Multiplier in
the configuration by the faster and more versatile 604. He also
replaced the venerable 405 accounting machine with a recently
enhanced version called the 402. As an optional feature he in­
cluded up to three Type 941 auxiliary storage units, each of which
provided electromechanical storage for sixteen numbers of ten
decimal digits each and their sign. Preparation for a calculation
on the CPC began by writing the sequence of operations to be
performed on a "planning chart." Numbers to be entered into
the computation at each step were specified by their location
either on a punched card or in storage locations within the CPC.
This process was similar to that used in preparing a problem for
computation by hand except that, with the CPC, the information
on the planning chart was punched into a deck of cards.
Then, under control of the punched cards, data were routed
over one or the other of two channels to the 604, where desired
operations were carried out. These operations could include ad­
dition, subtraction, multiplication, division, and more complex
functions such as square root, which could be prewired into the
604 plugboard and called for by instructions in the program deck.
Using modern terminology one might say the prewired plug­
board provided the � J9'\
6fflf fe� IM#amnaiP put cards provided the
 IBM's Initial Response 155

program and data. Computation results were routed over another

channel to either storage, card punch, or printer.25
Customer deliveries of the CPC began in late 1949, at which
time more than twenty had been ordered by premier companies
in the airframe industry and by government agencies and labo­
ratories. More than a year would pass before the first UNIVAC
was accepted by the Census Bureau. Tom Watson was ecstatic.
Nearly 700 CPC systems were delivered during the first half of the
1950s compared to only 14 UNIVACs during the same period.26
There is no simple way to compare the computational capabili­
ties of these very different machines. The UNIVAC could execute
a typical mix of arithmetic functions about twice as rapidly as the
CPC. The UNIVAC also had an internal memory capacity more
than one hundred times greater than the CPC, and it was
equipped with magnetic tape storage devices that made it possible
to move information in and out of memory rapidly. These storage
devices helped keep the arithmetic circuits of the UNIVAC busy
doing useful work a greater fraction of the time than was gener­
ally possible with the CPC. The UNIVAC's advantage was prob­
lem-dependent, however, and could range from no advantage at
all to as much as two orders of magnitude. In practice a UNIVAC
could handle large computational problems that were impractical
for a CPC. 27
In the long run the relative computational capabilities of these
machines were less important than the number installed. During
the first half of the 1950s, the installed UNIVACs satisfied com­
putational needs at fourteen customer locations, whereas the
installed CPCs satisfied the needs at fifty times that many. The
company was rapidly creating an infrastructure of knowledgeable
customers, salesmen, and servicemen for electronic computers.
Contrary to widely held perceptions, IBM had the fastest start in
the market for electronic computing capability. It did not have to
catch up as most accounts of this era suggest.

The UNWAC Threat

The real threat of UNIVAC to IBM was not in the relatively small
niche market for computing but rather in the far larger market for
information processing equipment. Here IBM's punched cards had
long reigned supreme as the means for storing business records
such as payroll, inventory, personnel, insurance, or census data.
Information stored o��mH� cessed in a cost-effective
156 Building IBM

manner using the sorting, collating, and simple arithmetic opera­

tions provided by punched-card equipment.
The use of magnetic tape, as proposed by Eckert and Mauchly,
to handle many punched-card functions was a major innovation
and the primary threat posed by UNIVAC. Initially Watson, Sr.,
failed to understand this threat because he and his engineering
staff believed that data stored sequentially on long magnetic tapes
would have limited utility. It was important to be able to enter
information in one sequence and then rearrange it into an­
other-for example, to enter information in the sequence in
which it was received and then order it alphabetically. Hollerith
had recognized this when he chose cards over paper tape. Watson
initially believed punched cards would have the same advantage
over magnetic tape.
An important insight of Eckert and Mauchly was that UNIVAC
would be able to read into memory a sequence of information
from magnetic tape, rearrange the information in memory, and
then write the rearranged information on another tape. Many
passes back and forth between tapes might be needed to rear­
range all information, but this could be done automatically under
program control. Magnetic tapes could thus replace punched
cards for this important function. The sorting of information on
tapes and the "sort" programs to perform it were crucial to the
success of early business-oriented computers.
The failure of many at the time-and perhaps even more
today-to grasp the overarching significance of the use of mag­
netic tapes on UNIVAC is partly the result of an unfortunate
misnomer: all electronic information processing machines are
called computers even though they are seldom used for comput­
ing. Although ENIAC was the first large-scale electronic com­
puter, it lacked the memor y and storage facilities necessary for
information processing. In contrast, UNIVAC and other elec­
tronic stored-program computers equipped with magnetic tape storage
were effective information processing machines. Because they
were also superior to computers of the ENIAC design for com­
puting, ENIAC had no progeny.

Applied Science Representatives

In responding to the emergence of large-scale electronic comput­

ers, IBM initially focused on their application to engineering and
scientific computations. Success in tht: Il14rketplace would require
c;opyngnred Matenal
 IBM's Initial Response 157

more than equipment. Computation methods suitable for elec­

tronic computers would have to be devised, and customers and
salesmen would have to be trained.
The Watson Scientific Computing Laboratory at Columbia Uni­
versity played a key role in developing methods and creating
knowledgeable customers for automatic computation. Beginning
in 1 945 Wallace Eckert and others from the Watson Laboratory
taught courses in computing machinery and numerical methods
as part of the regular Columbia University graduate school pro­
gram. These courses are believed to be the first "computer sci­
ence" courses offered by any university. Watson Laboratory
personnel also held symposia, published reports, and made their
computing resources available to members of the engineering
and scientific communities. 28
To further build the customer base,John McPherson and Wal­
lace Eckert proposed in May 1948 that people with advanced
training in mathematics be hired to assist customers in doing
scientific computing.29 In response IBM's general sales manager
requested nominations from each sales district of at least three
individuals interested in computation. These individuals were to
be of "the highest type, from recognized universities or colleges,
with a Bachelor's or Master's Degree" in the fields of physics,
aerodynamics, chemistry, or metallurgy. The nominations were to
be sent for review to John McPherson who had just been pro­
moted to vice president for engineering.30
The request produced only four appropriate candidates during
the year. This was not adequate. There were already more than
fifty IBM installations where the primary activity was scientific and
engineering computing, and these were believed to "represent
only a fraction of the available business." Furthermore, even
people with advanced technical degrees required six or more
months of special training with IBM equipment and automatic
computation methods before they were ready for assignment to
a branch office.31

Concluding that success in this endeavor required far greater

emphasis, McPherson recommended in August 1949 that a new
marketing function be established with the promotion of techni­
cal computation as its primary responsibility. The recommenda­
tion was accepted, and Cuthbert C. Hurd, recently hired into
headquarters from the Oak Ridge National Laboratory, was as­
signed to ser ve as director of a new organization, called the
Applied Science Dep�Jm.�Jghted
Material
158 Building IBM

Cuthbert Hurd was exceptionally well qualified for the position.

He held a bachelor of science degree from Drake University, a
master of science degree from Iowa State College, and a doctor's
degree in mathematics from the University of Illinois. During
World War II he had served as professor of mathematics at the
U.S. Coast Guard Academy in New London, Connecticut, where
he inaugurated the use of IBM machines for statistical studies. He
served as dean of a small college for two years after the war before
accepting a position at the Oak Ridge National Laboratory. There
he chaired the committee that introduced IBM machines for
calculations of gaseous diffusions processes and organized a cen­
tralized computing facility to serve twelve divisions.32
At IBM Hurd rapidly created a department of highly qualified
people, many of whom had Ph.D.s in mathematics, engineering,
or science. Mter receiving training in sales and automatic com­
puting techniques, members of the department were assigned to
branch offices where they assisted salesmen in the sale and instal­
lation of CPCs and subsequent computers for solving scientific
and engineering problems. The unique training and experience
of Applied Science representatives enabled them to help custom­
ers in understanding and solving problems. They also advised the
company's product development groups of new requirements.33
Creation of the Applied Science Department to sell machines
for scientific and engineering calculations was an important step
in developing a market large enough to accommodate not only
the CPC but also the more powerful stored-program computers
that would soon be available. It also established a second distinct
marketing and product development culture in IBM. Up to this
point, the company had achieved its commercial success in auto­
matic computing through equipment installed primarily to solve
business problems. Now an increasing number of machines would
be installed primarily-or even entirely-for scientific and engi­
neering computations.

Supercomputer Strategies

Although IBM enjoyed market leadership in electronic comput­

ing, there was a significant missing ingredient: namely, a commer­
cial computer with the capability of a UNIVAC. This was a
genuine business need. It was the publicity attendant to UNIVAC,
however, that was most painful for Watson, Sr., who was proud of
IBM's leadership in a��&mmefMion.
 IBM's Initial RPsponse 159

Remington Rand achieved a major publicity coup during the

U.S. elections of 1952 by having a UNIVAC featured on national
television, programmed to predict the outcome of the election
when only a small fraction of the votes were counted.34 Its
spectacular success attracted worldwide attention and made
the name UNIVAC almost synonymous with large electronic
computers.
Watson, Sr., and James Bryce had established an informal strat­
egy for maintaining leadership in supercomputers. It was to use
IBM expertise and money to build and "contribute to science"
one-of-a-kind "supercalculators" that could be reproduced for a
few customers, if desired, and be used to test improved hardware
and computation methods for lower-cost products. This strategy
was thwarted in its first implementation by the breakdown of
cordial relations between Watson and Aiken during the dedica­
tion of the IBM-Harvard Mark I.
To recover from this setback, Watson hired Wallace Eckert to
establish the Watson Scientific Computing Laboratory at Colum­
bia University. A primary assignment was to design a supercom­
puter. The result was the SSEC, dedicated in January 1948. The
SSEC provided IBM with engineering experience, valuable exper­
tise in operating and programming computers, and the primary
patent on stored-program computers.
Unanticipated by Watson, however. was the vast government
program to develop electronic computers. This program created
a technical community outside IBM that benefited from many
independent engineering projects, simultaneously cooperating
and competing with each other. All of these projects were funded
by the government and stimulated by von Neumann's intellectual
leadership. These activities , rather than IBM's, would define the
next developments in large-scale electronic computers.
Watson's insistence on early completion of the SSEC, combined
with cost constraints. had forced its designers to limit the use of
vacuum tube circuits to a small number of critical functions.
Eckert understood, however, that any follow-on supercomputer
was to be designed and built on a schedule that would encourage
the use of new technologies. To initiate this activity. he hired
Byron L. Havens, who had worked at the MIT Radiation Labora­
tory, where many of the more significant wartime advances in
electronics had been made. Havens was recommended by I. I.
Rabi. who as associate director of the Radiation Laboratory had
himself hired HaventiJpl�Mb!M�ierial
160 Building IBM

Arriving at the IBM Watson Laboratory inJanuary 1946, shortly

before Palmer transferred from Endicott to Poughkeepsie, Ha­
vens began hiring engineers to conduct research studies on elec­
tronics. Like Palmer, Havens looked for engineers who had
gained practical experience with electronics through their war­
time assignments. Unlike Palmer's group which became involved
in product development within six months, Havens's group would
have several years to conduct research on improved electronic
circuits and devices.36 By mid-1950 Havens was ready to design
and build a large stored-program supercomputer. He projected it
would be at least two hundred times more powerful than the
SSEC and ten times more powerful than any computer then
under construction.
In a major change in policy, IBM sought government funding
for the project. A prime promoter of this policy shift was Louis
H. LaMotte, head of IBM's office in Washington, D.C. LaMotte
had become alarmed by a report written by a member of his staff
in August 1 948. Based on information gleaned at a symposium
held at the University of California in Los Angeles, the report
advised that at least fourteen organizations in the United States
were developing electronic computers. Most of these computers
were classed as "large," and all were funded at least in part by the
government. Not only was the Bureau of Standards funneling
money from various government agencies into computer devel­
opment, but government-funded organizations were being pres­
sured to purchase machines. According to the report, six
organizations were "definitely planning to order large computing
machines at the earliest possible moment."37
Forwarding this report to Tom Watson and the company's
executive vice president, J. George Phillips, LaMotte wrote, "It
appears there must be some very fundamental reason why so
many outside agencies are intensely interested and active in this
field." Then he posed a significant question: "Inasmuch as IBM
is the leader in the field of calculating equipment, does it not
seem reasonable that it too should be kept abreast of all develop­
ments by active participation in this field, either to an equal or
greater extent than any of the listed organizations?" LaMotte
urged that the company supplement its own funding of research
and development by offering to build computers for the govern­
ment "for any legitimate purpose on a cost-plus basis."38 The
company's strategy of "going it alone" was beginning to crumble
Copyrighted Material
 IBM's Initial Response 16 1

under the realities of government-funded competition. McPher­

son and others began cautiously courting government agencies.
Following receipt of a request for a "high-speed electronic
calculator" from the U.S. Naval Ordnance Laboratory in Silver
Spring, Maryland, in September 1950, John McPherson re­
sponded, 'We should be pleased to undertake the development
and construction of such a machine for you on a research and
development basis, under which we would be reimbursed for our
costs plus a fixed fee." The fixed fee was set at one dollar by IBM,
whose objectives were not immediate profit, but rather service to
a valuable customer, favorable publicity, experience in design,
construction, and maintenance of a supercomputer, and patents,
which the contract permitted it to retain. The monies ultimately
received were about twice the $1.3 million estimated in May 1951
when authorization for the project was received.39
Designed and built under Havens's leadership at the Watson
Laboratory, the resulting machine became known as the NORC
(Naval Ordnance Research Calculator ). It was characterized by
use of decimal (rather than binary) arithmetic, floating-point
operations, automatic address modification, and substantial error
detection capability. With a I-microsecond clock period and a
2,000- to 3,600-word capacity, 8-microsecond access, cathode-ray­
tube memory, the NORC reigned as the world's fastest computer
for several years after it was publicly demonstrated in December
1954.40
The acceptance of government money to pay for developing
NORC was evidence of the increasing influence of Tom Watson,
who was eager to move into large-scale electronic computing as
fast as possible. Through his initiative (see chapter 13) the com­
pany embarked on a program to manufacture and sell large-scale
electronic computers at the same time negotiations with the Naval
Ordnance Laboratory were entering their final phase. As a result,
Project NORC was soon relegated to a secondary role in the
company's plans. Tom Watson had found an even better way to
tap into the country's cold-war funding for electronic computers.

Copyrighted Material
13
Watson, Jr., Takes Charge

In September 1949, not quite four years after he returned from

military service, Tom Watson was promoted to executive vice
president. He was now next in line for the presidency, which
reported to the chief executive officer-his father. Still largely
disconnected from the EAM business, he had been given a rela­
tively free hand in electronics. The success of the IBM 603, 604,
and epe had convinced him that customers were ready for the
new technology and that IBM had the capability to develop and
manufacture it.
He was increasingly frustrated, however, by the relatively low
percentage of research and development resources allocated to
electronics. Only at his insistence had a plan been made at the
end of 1948 to hire twenty-five electronics engineers in 1949, half
of whom would be assigned to Ralph Palmer in Poughkeepsie.
This was approximately twice the number of engineers of all types
hired in the previous year, but it represented only about 5 percent
of the company's engineering staff.! In his new role as executive
vice president, Tom Watson planned to shift resources into elec­
tronics even more rapidly. He also planned to benefit somehow
from the huge sums of money the government was spend ing on
computers.

Choosing Team Members

To accomplish his objectives, he would first have to assert control

over the development of EAM products, which generated most
of the company's revenue and consumed most of its R&D funds.
A major impediment, as he saw it, was John McPherson, who
enjoyed his father's full confidence. McPherson was also well
respected by the engineers in Endicott, who had primar y respon­
sibility for EAM deveI�JJ}f��Jl3/what the secretive Endi-
164 Building IBM

cott inventors were doing would have been difficult in any case,
but the strong link McPherson provided between them and the
older Watson made Tom Watson's task virtually impossible.
Deeper frustration, and indeed anger, was caused by McPher­
son's support of the senior Watson's view that the company's
research was outstanding. Their joint stance thwarted Tom Wat­
son's efforts to increase dramatically the company's expenditures
in electronics. McPherson "made a big mistake," the younger
Watson later asserted. By remaining loyal to Watson, Sr., who
"wasn't going to live very long," McPherson lost any chance of a
good relationship with the son.2
The person who ultimately provided the rationale for hiring
more engineers was Albert L. Williams. With two years of college
education, Williams had joined IBM as a student sales repre­
sentative in 1936. He was elected controller in 1942 and treasurer
in 1947.3 On his own, Williams studied the research and develop­
ment expenditures of large companies in the electrical equip­
ment and electronics businesses. These companies were typically
spending more than 3 percent of their gross income on R&D,
whereas IBM was spending somewhat less.4 It was a simple, yet
convincing argument for the older Watson, who was pleased to
have a sound business reason for accommodating his son.
Given approval for a significant increase in R&D expenditures,
Tom Watson needed a technical leader to carry out the plan. It
would not be John McPherson. Although he could not remove
McPherson from the position of IBM vice president, he did have
an easy way to decouple him from management responsibilities.
The position of director of engineering had remained empty
following McPherson's promotion to vice president in 1948. Tom
Watson filled the position in May 1950 with W. Wallace McDowell,
who had served as manager of the Endicott laboratory since
1942.5 Ironically, John McPherson-the man most responsible for
the company's early postwar push into electronics-was to be
excluded from any further decision-making responsibilities.
Wally McDowell was just the type of technical administrator
Watson needed. Mter receiving his bachelor's degree in engineer­
ing management at MIT in 1930, he had joined IBM and worked
briefly as a draftsman and served in several engineering admin­
istrative positions in Endicott before becoming laboratory man­
ager. Unlike previous laborator y managers-who primarily
provided administrative support for engineers and carried out the
wishes of Watson, Sr.l: �yFl gtYi61J MMreilafed directly with govern-
 Watson,Jr., Takes Charge 165

Figure 13.1. W. Wallace McDowell

Wally McDowell presided over IBM's rapid growth in engineering, with em­

phasis on electronics, following his promotion in 1950 from manager of the

Endicott laboratory to director of engineering. A 1930 graduate of MIT, he
enjoyed the confidence and respect of both Tom Watson and Watson, Sr.
This photograph was taken one year before he retired from IBM in 1968.

ment agencies during World War II and made many i ndepend ent
decisions.6 He lacked the technical capabilities of Palmer or
McPherson, but he was a good listener and was well respected for
his decisions and f ollow through. Of particular note was his sup­
-

port of research on electronic circuits before the war.

McDowell had other important attributes for his new position.
He was well respected by both Watsons, politically wise, and able
to handle delicate relations between the two. Unknown to Wat­
son, Jr., at the time, for example, his father met privately with
McDowell at lunch time almost every day for several months after
McDowell was appointed director of engineering. This provided
the older Watson with an opportunity to keep informed of his
son's activities and to influence them without appearing to do
SO.7
Within five years, under McDowell s leadership the total num­
' ,

ber of employees working in R&D increased from fewer than

1,000 to more than 4,000. Approximately 60 percent of these were
classed as technical rather than administrative. During the same
period, the total number of IBM employees worldwide had nearly
doubled to 56,300. (See appendix A.)
Copyrighted Material
166 Building IBM

Tom Watson found another natural ally in Louis LaMotte, a

senior sales executive who had never been cowed by the elder
Wat'lon's imperious manner. In 1948 he had warned, from his
post in Washington, D.C., of the growing threat of government­
funded competition in computers. He had urged that IBM poli­
cies be altered so as to accept government development contracts.
In addition to McDowell, Williams, and LaMotte there were two
other members of Tom Watson's team in 1950. The first of these
was a protege of Charlie Kirk, James W. Birkenstock, who had
been promoted by Watson, Sr., from a minor headquarters sales
post to general sales manager in January 1946. Although he was
intelligent and capable, Birkenstock was unprepared for this level
of responsibility. Twelve months later he was removed as sales
manager and made manager of the Future Demands Department,
which had been dormant since McPherson vacated its leaders hip
four years earlier.
Despondent over what he considered a demotion, Birkenstock
became an adviser to Tom Watson, following a session in which
he proposed quitting the company and the young Watson urged
him to stay. Soon thereafter Birkenstock became one of the first
in IBM to recognize that electronic computers could use mag­
netic tapes to replace many functions of punched cards. By 1948
he was predicting that electronic computers with ma gnetic tape
storage would take over IBM's traditional markets and that
punched cards were doomed. He urged Tom Watson to act
quickly.s
The final member of Watson's early team was Cuthbert Hurd .

With a doctorate in mathematics and extensive experience in

digital computation methods, he had already demonstrated his
value as head of the Applied Science Department. Now as a
member of Watson's inner circle, he provided intellectual leader­
ship that was crucial to the company's entry into the market for
large-scale electronic computers.
Even before he selected Wally McDowell to head engineering
in 1950, Tom Watson had urged John McPherson to leave IBM,
telling him he "would do better elsewhere."9 But McPherson saw
insufficient reason to do so. He still had the support of the older
Wat'lon, and he was deeply involved in getting the company into
electronics.
Deprived of influence at the top clearly not a member of Tom
-

Watson's team-McPherson nevertheless continued to play much

the same role as he h�� ��� MR �ept abreast of engineer-
Pn l:i a
 Watson, Jr., Takes Charge 167

ing projects. He aided their progress, primarily by making engi­

neers aware of relevant advances of others inside and outside the
company. It was an important role in a company in which few
technical people had outside contacts and in which the older
engineers still kept their work secret from one another.
Ralph Palmer was also left out of the inner circle. Tom Watson
recalled his reasons for not appointing him director of engineer­
ing as follows: "Putting the leader of the Poughkeepsie mavericks
in charge would be an unnecessary insult to Dad's inventors in
Endicott. Besides, Palmer was simply too important in the lab." 10
Indeed Palmer's interests were too strongly oriented toward en­
gineering and research to satisfy the needs of Tom Watson, who
kept more aloof of detailed engineering decisions than had his
father.

The Defense Calculator

An opportunity for action came with the outbreak of the Korean

conflict in June 1950 just one month after McDowell's promotion
,

to director of engineering. Recalling his World War II creation of

a separate munitions manufacturing organization, Watson, Sr.,
asked Birkenstock to establish a military products division to
provide specialized products for national defense. Watson's mo­
tivation was primarily patriotic, but his son seized on the idea as
an opportunity to speed the company's development of large­
scale electronic computers.ll
To help Watson,Jr., assess the opportunities, Birkenstock asked
Cuthbert Hurd to join him in visiting prospective customers. As
director of the Applied Science Department, Hurd already had a
list of nearly sixty installations where IBM equipment was being
used in defense-related engineering and research applications.
Jointly and then separately the two of them visited more than
twenty of these. They found people eager to discuss their compu­
tation needs, which were growing rapidly because of the military's
pressure for significantly improved armaments.12
By November Hurd had created his own assessment of the
military's needs for computing capability. As submitted to Tom
Watson, they were in priority order: (1) atomic energy; (2) guided
missiles; (3) strategic planning (cryptanalysis, weather forecast­
ing, and game theor y); and (4) jet engines. For a variety of
reasons Hurd believed IBM could have the greatest impact by
directing its efforts "avp>�61IBt.he basic guided missile
168 Building IBM

control problems."13 Two weeks later, however, Hurd had substan­

tially revised his thinking. He now concluded that the "special
needs" of different users could best be served by a single, properly
designed machine.
The plan was thus evolving from one in which three or four
different computers would be built to the specifications of se­
lected customers to a plan in which only one computer would be
designed and manufactured to satisfy the needs of as many cus­
tomers as possible. As described by Birkenstock, "The machine
should be general purpose in concept to permit our duplicating
the prototype in the order of 25 to 30 times. "14 It was a riskier
strategy because the cost of development would be paid entirely
by IBM. If successful, however, the rewards would be greater.
Engineering responsibility would rest with Ralph Palmer's group
in Poughkeepsie.
Even before completing work on the 604 electronic calculator
in mid-1948, Palmer had begun to broaden his group's research
activities to encompass all technologies needed for large-scale
electronic computers. At the top of his priority list was high-speed
memory. His preferred solution (called the Williams-tube mem­
ory after its British inventor) made it possible to write and read
digitally encoded information electronically on the face of an
ordinary cathode ray tube. Not only did the Williams tube provide
faster access to information than the delay-line memory planned
for UNIVAC, but it had the added advantage that the storage
tubes were commercially available. Achieving the desired reliabil­
ity was, however, a serious problem.15
Also under study was diode logic in which vacuum tubes pro­
vided power amplification and the lower-power germanium
diodes performed the switching function. By mid-1949 informa­
tion storage by magnetic material on the surface of a rotating
drum was under study, using a drum supplied by the Endicott
laboratory; and the storage of information on magnetic tape,
initially assigned to Endicott, had also become the responsibility
of Palmer's group.16
That fall Steve Dunwell drew up specifications for an electronic
computer using components being developed by Palmer. In­
tended for business applications, it was called the Tape Processing
Machine (TPM) because a key question to be addressed was the
appropriateness of magnetic tape for storing business records. By
the spring of 1950 the Tape Processing Machine had become the
Copyrighted Material
 Watson,Jr., Takes Charge 169

primary focus of Palmer's group.!' It was to be IBM's most direct

response to UNIVAC. To speed the work, Palmer had been
authorized to hire more than a dozen additional electronics
engineers. 18
By late in the fall, the possibility of redirecting Palmer's effort
to build a computer to meet the needs of military organizations
and defense industries was under consideration. To facilitate a
decision, Tom Watson established a committee to draw up spe­
cifications for the proposed computer. Headed by a member of
the sales organization and with representatives from engineering,
Future Demands, Applied Science, and the Watson Laboratory,
the committee carried out its task between Christmas and New
Year's Day.19
Hardware component specifications were based on compo­
nents under development in Palmer's group, but a substantial
deviation from the TPM design was required to satisfy the needs
of engineering and scientific computing. The most significant
change was from a bit-at-a-time serial operation to a data path
wide enough to handle an entire word at a time. This increased
both the speed and the cost of the computer.
Largely on the basis of Hurd's assertion that he could sell at
least six of the machines, Watson made a tentative decision to
proceed.2O Palmer immediately began reassigning engineers to
the new project and made preliminary schedule and cost esti­
mates. In February 1951 Hurd was authorized to approach po­
tential customers using an estimated rental of $8,000 per month
and delivery in little more than one year. 21
When more than twenty letters of intent were quickly obtained,
Watson gave final approval. By the spring of 1952 firm engineer­
ing designs were in place. The projected rental price had dou­
bled, with a range of $II,900 to $17,600 per month, and deliveries
were estimated at just under a year. In spite of the higher price,
orders for thirteen were confirmed by August and LaMotte (now
vice president for sales) had authorized the purchase of materials
for a production lot of eighteen machines.22
The shift in focus from commercial to defense-related applica­
tions for IBM's first large-scale electronic computer was justified
by the uncertain market for the former and the well-defined
demand for the latter. The shift had also made it easier to sell
the project internally. Issues of marketability and profitability
could be set aside by arguing its importance to national security.
Copyrighted Material
170 Building IBM

The name, Defense Calculator, was specifically selected to appeal

to the patriotism of the older Watson and to avoid use of the
unacceptable word, computer.
The internal arithmetic and control features of the Defense
Calculator were based, with a few significant departures, on von
Neumann's lAS computer, functional plans for which had been
circulated in 1946 and again in revised form in September 1947.23
A principal result of this decision was use of the binary number
system rather than the decimal system as in all previous IBM
machines. In addition to reducing the design time, the decision
was expected to facilitate sales. "I was aware that practically eve­
r ybody in the country had already been sold on John von Neu­
mann's idea of how to build a machine," Ralph Palmer recalls.
Through use of his design, von Neumann inadvertently became
"IBM's best salesman."24
In Januar y 1952 von Neumann also became a consultant to
IBM. Most of the Defense Calculator design decisions had already
been made, however, and the engineering work was well ad­
vanced. The hiring of von Neumann is best likened to an insur­
ance policy for Tom Watson. No specific contributions can be
attributed to him, but the ready availability of his advice bolstered
Watson's confidence.25
Several significant deviations from von Neumann's lAS design
were incorporated in the Defense Calculator. These included the
use of 3frbit rather than 40-bit words, the capability to address
and process half-length as well as full-length words, and the
representation of numbers by their absolute value and their sign
rather than by using complements for negative numbers. Prob­
ably more significant to users was the emphasis on fast and
reliable input-output equipment.
Palmer also determined that each functional unit would be
packaged in its own box (frame and covers) made small enough
to pass through standard doorways and to fit on commercial
elevators. After preshipment testing, a complete system could be
disassembled, shipped, and reassembled in the customer's facility
simply by disconnecting and reconnecting the cables between the
boxes. Repairs and future upgrades would also be easier.26
The functional units comprising the Defense Calculator were
the electronic analytical control unit, electrostatic storage unit,
punched-card reader, alphabetical printer, punched-card re­
corder, magnetic-tape reader and recorder, and magnetic-drum
reader and recorder. fJeYJ9fi�f1lirt�Nlfll8 also housed in separate
 Watson, Jr., Takes Charge 171

Figure 13.2. mM 701

Known as the Defense Calculator during development . the IBM 701 com­
puter i� pictured (top) in the Poughkee psi e laboratory in March 1952. Clock­
wise from the left are two tape units, a magnetic drum storage device, the
cathode-ray-tube memory unit, arithmetic and control unit with operator ' s
panel, card reader, prin ter, and card punch . A panel of the 701 with eight­
tube pluggable units is shown (bottom).

Copyrighted Material
172 Building IBM

boxes. The product type number of the first unit gave the system
its widely used informal name, the IBM 70J.27

Measures of Success

At the April 1952 annual meeting, Watson, Sr., advised stockhold­

ers of a doubling of the size of the Poughkeepsie plant to handle
defense production, continued growth in demand for the com­
pany's "regular products," and rapid progress in electronics.
About 35 percent of the laboratory staff was now working on
electronics. He also advised stockholders of the effort to develop
a large-scale electronic computer, which was twenty-five times
faster than the SSEC, yet would occupy less than one fourth the
space. He reported that ten of these computers were ordered only
two days after it was announced on a limited confidential basis.28
Watson's pleasure in this announcement was heightened by the
evidence it gave of the success of his older son, who had been
promoted to president in January.
For Ralph Palmer and the more than 150 people assigned to
the 701 development effort, the December 1952 shipment of the
first production machine to the company's headquarters in New
York City was a major achievement. The shipment occurred just
two years after tentative specifications had been proposed for the
Defense Calculator and less than eight months after orders for
the first ten systems had been confirmed.
An even better measure of success occurred two months later
when the second production machine was shipped to the Los
Alamos Scientific Laboratory and was installed and operational
only three days after it arrived. Debugging to assure satisfactory
performance on all customer tasks was continued for another
fourteen days before rental payments began.29 Operated twenty­
four hours per day, five days per week, during the first six full
months of operation, the system was reported to have been down
for errors or maintenance less than 24 percent of the time.30
Palmer's plan for simplifying installation by shipping functional
units in individual boxes was a success. Contributing to that
success were engineering designs intended to facilitate manufac­
turability, reliability, and ease of service of all components.
But IBM did not just design, manufacture, and sell products.
It provided a full service that required well-trained employees.
More than a year before the first production machine was in­
stalled at headquart� 9nbflHfUM1rMri9fgun for twenty-five re-
p
 Watson, Jr., Takes Charge 173

cently hired graduate engineers to train them in all aspects of the

design, construction, and operation of stored-program comput­
ers. These engineers were to assist in constructing, installing, and
servicing the machines. Classes were conducted part of each day
for a year. T he rest of the day was spent on assigned tasks in the
701 development project. Homework was done at night. A second
class of fifteen students was begun in May 1952 to train additional
servicemen for the 701s. This class consisted primarily of experi­
enced customer engineers who had previously installed or serv­
iced 604 electronic calculators.31
The company's primary competition during the 1950s was
Remington Rand, which had acquired the Eckert-Mauchly Com­
puter Company (EMCC) in 1950 and Engineering Research As­
sociates (ERA) in 1952. Because ERA had begun its computer
developments with a larger cadre of experienced electronics en­
gineers than either IBM or EMCC, it is revealing to compare the
IBM 701 experience with that of the ERA 1103.
Several months before ERA delivered its first electronic com­
puter (Atlas) to the government in 1950, its engineers began work
on a far more sophisticated computer, code-named Atlas II.
Learning in late 1951 that IBM was building the Defense Calcu­
lator, the ERA engineers sought permission to announce an un­
classified version of their Atlas II computer to compete with IBM.
Work on Atlas II continued in mandator y secrecy until their
request was granted by the government in the fall of 1952. Then,
for the first time, the ERA engineers were able to advise the top
management of Remington Rand of the existence of the pre­
viously classified computer. None of the top executives had secu­
rity clearance.
'The astounded executives approved a program to produce two
machines and to buy sufficient parts for two more." The declas­
sified Atlas II computer was announced as the ERA 1 103 in
February 1953. Like the IBM 701, the ERA 1103 used a 36-bit
word length and was equipped with a Williams-tube main mem­
ory, magnetic drum and magnetic tape storage, and punched­
card units for input-output. When all four ERA 1103s were sold
within a few months, a decision was made to produce them in
quantity. Full-scale manufacturing was underway by 1954, and
about twenty were ultimately built. Thus ERA suddenly found
itself in the computer business without a business plan. According
to two participants, "Questions arose regarding pricing, rental,
field service, installa��t�ping, and support. The
174 Building IBM

absence of even a modicum of applications and field support

became a glaring deficiency. "S2
Engineers at ERA hoped that superior engineering design of
the 1103 would offset the company's weakness in applications and
field support, but the predicted superiority of the computer over
the IBM 701 failed to materialize. A study team commissioned by
the Joint Chiefs of Staff to compare the two computers for
weather prediction applications in 1954, for example, concluded
that the "speeds of the two machines are comparable with a slight
advantage in favor of the 70 1." The input-output equipment of
the 701 was significantly faster, causing the study team to recom­
mend it unanimously over the 1103. ss The marketing problems
for the ERA 1103 were exacerbated in May 1954 when IBM
announced a substantially improved version of its 701 computer,
the IBM 704.

The 700 Series

Heavy emphasis on the 70 1 project during 1952 brought engi­

neering work on the Tape Processing Machine (TPM) to a near
standstill, but the TPM product planners benefited by having
more time to improve its specifications. Greater flexibility in
attaching input-output equipment was provided by a "common
input-output trunk," and a decision to process all bits of a char­
acter in parallel rather than one at a time produced a fivefold
increase in processing speed that more than offset the cost
increase.
These and many other proposed improvements appeared to
make the TPM superior to Remington Rand's UNIVAC. Ironically
in early 1953, however, careful analyses indicated the company's
conventional punched-card equipment would be more cost-effec­
tive for most customer tasks. Such analyses failed to take into
-account new applications that might be made possible with the
TPM, and they ignored the huge customer interest already stimu­
lated by four installed UNIVACs and the soon-to-be-delivered IBM
701. M any IBM salesmen were distressed by the company's deci­
sion to develop its first electronic stored-program computer for
engi neering and scientific computations rather than for business
data processing. The possibility of losing large data processing
customers to Re m i n g ton Rand threatened affected salesmen with
a major loss of sales commissions.
Copyrighted Material
 Watson, Jr., Takes Charge 175

Thus it was that concerns of the sales organization and antici­

pated future developments-rather than cost-benefit analyses­
drove the decision to announce the improved TPM. In
September 1953 it was announced as the IBM 702 Electronic Data
Processing Machine (EDPM) . The name was selected once again
to avoid the politically unacceptable word, computer: Deliveries
were not to begin until early 1955, four years after the first
UNIVAC had been <l.ccepted by the Census Bureau. In the mean­
time IBM salesmen had nothing equivalent to offer. Only the
inability of Remington Rand to manufacture more UNIVACS­
and to provide adequate support services-saved many IBM sales­
men from loss of customers.34
Customer enthusiasm for UNIVAC and the IBM 702 was also
constrained by important business considerations. One consid­
eration was the high cost and long time required to modify
business practices and retrain personnel. Another was concern
over loss of visual access to data. " Perhaps the most radical idea
which business is being asked to accept is the idea that a reel of
tape can safely be used to carry information now being entrusted
to visual card files," the vice president and chief actuar y of the
Metropolitan Life Insurance Company observed in late 1953.35
Recognizing the various concerns of IBM salesmen and poten­
tial customers, and the need to coordinate company activities in
electronic computers, Tom Watson turned to T. Vincent Learson
for help. A towering personage, mentally and physically, Vin
Learson had been a fast-rising member of the sales organization
since joining the company in 1935 with a bachelor's degree in
mathematics from Harvard University. He had served as sales
manager for the EAM Division from 1949 to 1953, when he was
promoted to general sales manager. Now in April 1954, six
months after the IBM 702 was announced, Tom Watson ap­
pointed him director of Electronic Data Processing Machines
(ED PM) with a mandate "to coordinate the various aspects of this
complicated work."36
Moving from the position of general sales manager to respon­
sibility for a tiny new product segment might have been regarded
as a demotion, but Learson understood the younger Watson's
nearly single-minded dedication to this segment. Working closely
with the engineering, programming, manufacturing, sales, and
service organizations, Learson moved vigorously to better meet
customer needs. That April he created a new Programming Re-
Copyrighted Material
176 Building IBM

Figure 13.3. Learson, Watson, Jr., and Palmer

T. Vincent Learson (left) was photographed with Tom Watson (center) and
Ralph L. Palmer (right) when he was promoted to the new position of direc­
tor of electronic data processing machines (EDPM). The term EDPM was
chosen to distinguish the new large-scale electronic machines from the ear­
lier electromechanical accounting machines (EAM). Learson's task was to
coordinate all EDPM activities from development through sales and service.

Copyrighted Material
 WaiSon, Jr., Takes Charge 177

search Department to "develop a library of advanced EDPM

programs for the use of customers and [to] assist the sales organi­
zation in evaluating customers' potential uses.» 37
With fifty 702s on order in June, he limited shipments to those
scheduled for delivery in 1955 and early 1956 so that a number
of significant improvements could be made. As a result, only
fourteen 702s were built and installed.38 The remaining orders
were to be satisfied (at a higher rental price) by an improved
version, which was announced in October 1954 as the IBM Type
705 EDPM. An improved version of the 701 had already been
announced in May as the IBM Type 704 EDPM.
That August Learson had expanded an advanced technology
effort into a product development program and given it respon­
sibility to provide magnetic-core memories for the IBM 704 and
705 computers. Not wanting to draw attention to reliability prob­
lems of the 701 and 702 computers, IBM chose to highlight only
the slightly improved processing speed provided by magnetic-core
memory. Superior reliability was, however, the primar y reason for
changing from cathode-ray-tube (Williams tube) memories. For
the same reason, magnetic-core memories were retrofitted on
installed 701 and 702 computers as soon as shipments of the 704
and 705 computers commenced in January 1956.39
The 700 series of computers was capped by the 709 and 705-III,
announced in 1957. For scientific and engineering applications,
the 700 series offered computers that used binary notation and
emphasized fast computation speeds. For business data process­
ing, it offered computers that emphasized alphanumeric nota­
tion, magnetic-tape storage, nonstop processing of huge
accounting files, and high-quality printing of reports. So success­
ful were these computers that the 700 series evolved into the 7000
series when transistors replaced vacuum tubes. In particular, the
IBM 7090 computer, first shipped in November 1959, was made
to operate like the earlier IBM 709 in order to facilitate customer
migration to the newer machine.4O
Learson's effectiveness in strengthening the thrust into elec­
tronic computers won him a promotion, in only eight months, to
vice president in charge of sales. Once again he had responsibility
for all of the company's sales activities, but now he was armed
with in-depth knowledge of the rapidly growing electronic com­
puter segment-and he was an IBM vice president.41
Success of the 700 series was crucial for Tom Watson, who has
described himself as �J'i�1Ma� of failure" and "abso-
178 Building IBM

lutely panicked" when he first learned that a second UNIVAC had

been installed at the Census Bureau.42 Driving himself and his
associates, he was determined to achieve leadership in this part
of the industry. By his own count, it was in 1955 that the number
of customer installations with 700 series computers first exceeded
the number with computers manufactured by Remington Rand.
It was then, in Watson's view, that IBM had once again "con­
clusively asserted its leadership" in the information processing
industry.43

The Most Populnr Computer

At a time when Tom Watson, his close associates, and most outside
observers had their attention riveted on large-scale electronic
computers, an important story was unfolding in the IBM Endicott
laboratory. It was here that Frank Hamilton had set about early
in 1948 to carry out the senior Watson's instructions. Using what
he had learned from the SSEC supercomputer project, he was to
devise a small machine "to meet the requirements of the ordinary
businesses we serve. ,,«
Advised by the Future Demands Department that the soon-to­
be delivered Type 604 electronic calculator would satisfy the
foreseeable needs of business applications, Hamilton (with
McPherson's support) decided to develop a small computer for
scientific applications. His initial approach was to supplement the
604 with a reduced complement of SSEClike paper-tape devices
for storing tabular data and sequence control information, an
electromechanical table-lookup unit, and perhaps 200 digits of
vacuum tube storage for intermediate results.
Before the end of the summer, however, he had dropped the
bulky vacuum tube and electromechanical storage in favor of
'
digital magnetic recording on the cylindrical surface of a rotating
drum. Paper-tape storage was also dropped when he recognized
that instructions could be read from a rotating drum and exe­
cuted in the same manner as from paper tape. With a single
magnetic drum storage device holding both data and instruc­
tions, the architecture of the system had evolved by April 1949 to
one having less similarity to its SSEC predecessor than to von
Neumann's lAS computer. Unlike the lAS computer, however, it
used decimal rather than binary numbers, and it processed infor­
mation one decimal digit (several bits) at a time, whereas the lAS
computer had an inteaJ M-� 8�fM1fcJ/arithmetic circuits capa-
 Watson, Jr., Takes Charge 179

ble of processing an entire forty-bit binary number at a time.

Hamilton's proposed machine was now called the Magnetic
Drum Calculator.45
One month later, Tom Watson redirected the project in the
first of a series of corporate-level decisions destined to deprive
Hamilton and his Endicott engineers of any opportunity to de­
velop IBM's first electronic, stored-program computer product.
Watson's sales manager had found an urgent need for a table­
lookup device to support rate table applications on the 604. The
Endicott drum was seen as the answer. Mter three months of
work, however, Hamilton's cost and schedule projections for a
drum attachment to the 604 were judged to be unacceptable.
Negotiations were then undertaken to purchase a magnetic drum
storage device from Engineering Research Associates (ERA),
which was the recognized leader in magnetic drum technology.46
Although it normally built devices for classified government pro­
jects, its engineers had given unclassified technical presentations
on their magnetic storage drums at engineering conferences. The
information storage medium consisted of magnetic tape, bonded
to the surface of a cylindrical aluminum drum. In operation the
drum rotated continuously.47
To improve on ERA's crude recording surface, the engineers
in Endicott had wrapped magnetic wire tightly around a drum
and then ground the outer surface to achieve a smooth cylindri­
cal surface. Despite good results with the Endicott drum, Watson
approved the ERA initiative over the objections of Palmer and
McPherson. He believed engineers at ERA were superior to those
in IBM's Endicott laboratory, whom he described as "a bunch of
monkey-wrench engineers" working under the tutelage of his
father and John McPherson.48
Representatives of IBM were soon convinced by ERA that a
drum-augmented 604 would be less desirable than a magnetic
drum computer-the very thing Hamilton was working on before
Watson redirected the project. The scope of the contact with ERA
was accordingly expanded to encompass the design of a magnetic
drum computer. Simultaneously Hamilton was allowed to return
to his original objective, but now he was in competition with
ERA.49
An audit of the competing projects was conducted in March
1950. Both machines were judged to have met the functional
specifications, and there was no clear winner. A major concern
expressed in the aud�t§t}�/llWEfSJIMF �1j t cost. Neither machine
180 Building IBM

could be built at the desired entry-level rental of $400 per month.

Even the upper limit rental of $1,000 per month for a full-func­
tion system was in doubt.50
Frustration with unachievable cost-performance requirements
was compounded for Hamilton by a reduction in engineering
resources early in 1951 when major emphasis was given to the
Defense Calculator project. The number of people assigned to
the Magnetic Drum Calculator dropped steadily, reaching a low
of four employees by the fall of 1952. Nevertheless, Hamilton
pushed relentlessly forward, trying to complete an engineering
model to satisfy specifications established some months after the
audit.5'
Meanwhile IBM's pioneering 604 and CPC were facing increas­
ing competition in the market. By mid-1952, for example, the
Remington Rand Type 409-2 calculator, which was described as
having "several times the capacity" of the four-year-old IBM 604,
was selling well despite its higher monthly rental of $850 versus
$550. In addition there were now seven small stored-program
computers in the market competing with IBM's three-year-old
CPC.52
This competition was particularly evident to Cuthbert Hurd
who, as head of the Applied Science Department, was responsible
for the marketing of computers for scientific and engineering
computations. Unlike the better established market for business
applications, this market was more sensitive to new function and
technologies than it was to price. Accordingly Hurd agreed to
assign some members of his department to help tailor Hamilton's
computer to satisfy the needs of these customers.
In November 1952 Hurd convinced Tom Watson to give the
Magnetic Drum Calculator his full support. Almost immediately
Hamilton was authorized to increase his staff from four to fifty.53
Eight months later Hamilton's group had shown its design to be
superior to alternatives proposed by the Poughkeepsie laboratory,
had found ways to engineer around certain patents held by
Remington Rand, and had readied the machine for an­
nouncement. The main memor y used a four-inch-diameter drum
designed and built in Endicott. Ferromagnetic metal was electro­
plated on the surface to provide the storage medium. Spinning
at 12,500 revolutions per minute, the drum afforded the remark­
ably short average access time to data of 2.4 milliseconds.
The IBM 650 Magnetic Drum Calculator was announced in July
1953. Rental was $3
P
c8 y,pgnteW�tfJrit?r a machine with one
 Watson, Jr., Takes Charge 181

Figure 13.4. mM 650 Computer

Top: Announced as the IBM 650 Magnetic Drum Calculator in 1953, the 650
became the most popular computer of the 1950s. The woman is sitting in
front of the console, which is on the box that housed the processing unit
and drum memory. The box behind it housed the controllers and power sup­
ply. A Type 533 card read-punch is at left and a higher-performance Type
537 card-read punch is at right. Bottom: The magnetic drum memory is
shown with its read-write devices removed at the left to reveal the cylindrical
drum on which information was stored.

Copyrighted Material
182 Building IBM

thousand 10-digit words of memory, and $3,750 per month for

a two-thousand-word machine. T hese rentals were four times
smaller than the $15,000 monthly rental for an IBM 701 com­
puter, but they were also four times higher than the original
objective for' the 650.
Designed and initially marketed primarily for engineering and
scientific computations, the IBM 650 was unexpectedly successful
for business applications. The first 650 to be shipped was in fact
installed at the John Hancock Mutual Life Insurance Company
at Boston in December 1954. Part of the product's success can be
attributed to punched-card input-output that permitted the 650
to fit easily into established EAM facilities. Of the eight other
small digital computers then available, only the machines of the
ElectroData Corporation, National Cash Register Company, and
the Underwood Corporation offered punched-card input-output
comparable to that of the 650, and none of these came within a
factor of three of its 2.4-millisecond drum access time.54
To satisfy business needs even better, a feature allowing alpha­
betic information to be read from, and punched into, cards was
announced in mid-1954. Another feature added in 1955 permit­
ted attachment of the Type 407 accounting machine to produce
reports on-line with excellent print quality. 55
By the time the last IBM 650 came off the manufacturing line
in 1962, almost two thousand had been produced and installed.
No other electronic computer had been produced in such quan­
tity. Frank Hamilton-who had learned to build computers by
working on the one-of-a-kind Mark I and SSEC supercomputers­
had met the senior Watson's challenge. He had developed a
machine for "ordinary businesses" that became the most popular
computer of the 1950s.56

Copyrighted Material
 14
Programming Computers

Viewed broadly as the adaptation of general-purpose devices to

specific tasks, "programming" was an old data processing concept.
So was programming support. From the beginning Hollerith had
wired and rewired (programmed) his equipment to handle spe­
cific jobs. The introduction of interchangeable, prewired plug­
boards made it relatively easy for EAM equipment to be shifted
from one application to another. An IBM publication called Point­
ers was distributing information about novel plugboard wiring
diagrams to customers by the early 1930s. Some of these diagrams
were crafted by IBM personnel, others by customers who were
willing to share their solutions. (See chap. 5, "Methods Re­
search.")
The first commercially available product to permit more than
a few sequential computation steps to be wired on a plugboard
was the IBM 604. The forty-step capability initially offered in 1948
was later expanded to sixty. The CPC introduced yet another level
of programming methodology to the marketplace in 1949. Pro­
gram instructions stored in cards could be read sequentially by
the CPC and executed. As with conventional EAM equipment,
IBM's sales and marketing representatives prepared, solicited,
and distributed programs (called solutions) to customers.l
Thus a culture of programming support and sharing was well
in place among IBM and its customers long before the first 701
computer was installed. The adaptation of this culture to meet
the requirements of electronic stored-program computer users
would be crucial to the company's continuing success.

Some Early Effmts

The concept of a series of rudimentary processing steps to pro­

vide sequence contrc¢6'p,r��tnputing machines dates
184 Building IBM

back at least to 1843 when Ada Lovelace translated into English

a description, written by an Italian mathematician, of Babbage's
Analytical Engine. She published the translation together with
explanatory notes that contained a list of instructions for direct­
ing such an engine to solve assumed problems. Written under
Babbage's guidance, these notes may contain the first computer
program to be published. Although Babbage worked on the Ana­
lytical Engine for many years, his work never reached fruition in
an operational device.2
The largest electromechanical computer ever built was the
Mark I, presented to Harvard University by IBM in August 1944.
It was programmed by the staff of the Harvard Computation
Laboratory. The electronic E�IAC, which was dedicated in Feb­
ruary 1946, was initially programmed by a laborious rewiring
process more closely related to the programming of Hollerith
equipment than to programming the architecturally advanced
Mark I or later computers.3
The IBM SSEC, dedicated in January 1948, placed great de­
mands on programmers by offering them more flexibility than
did the Mark I. For example, the benefits of large storage capacity,
achieved through a multilevel storage hierarchy, were accompa­
nied by increased programming complexity needed to control the
flow of data and instructions between storage devices and the
arithmetic unit. Other programming challenges arose out of in­
novative features such as the ability to alter instructions based on
intermediate results. Because the SSEC's electronic circuits exe­
cuted instructions hundreds of times more rapidly than the Mark
I, the programming effort to keep it busy effectively limited its
use to very large computations.
The benefit of crafting frequently used subroutines (Le., pro­
gram parts) only once rather than each time a new program
needed them gradually became obvious. The kinship to inter­
changeable plugboards was apparent. Furthermore, as proposed
by Eckert and Mauchly in 1945, subroutines used more than once
by an application program should be "inserted in the memory
only once, and the basic order code must provide for finding
these instructions each time they are needed. "4 At about the same
time, A M. Turing, a leading figure in British code-breaking
activities during World War II, was planning a computer for the
National Physical Laboratory in England. He proposed that fre­
quently used subroutines be stored on punched cards in a man-
Copyrighted Material
 Programming Computers 185

ner to facilitate their selection and assembly into a program using

standard EAM equipment.5
By 1951 Maurice V. Wilkes and two of his colleagues had
published a book describing the programming of their stored­
program computer (EDSAC) and its library of subroutines. Ex­
hibiting remarkable vision two years earlier, Wilkes had predicted
that when a machine was operational, and a number of subrou­
tines were in use, "There would be almost as much capital sunk
in the library of subroutines as the machine itself, and builders
of new machines in the future might wish to make use of the
same order code [i.e., machine language] as an existing machine
in order that the sub-routines could be taken over without
modification. "6
The reuse of complete, as opposed to partial, programs also
received early attention. The basic concept of a stored-program
computer is that an entire program (or its initial portion) is
loaded into the computer's main memory before program exe­
cution begins. Initiation and control of the program-loading
process required a special set of instructions, which could be the
same for loading many different programs. Short programs called
"program loaders" were therefore among the earliest reusable
programs written for the Defense Calculator. One of these was
short enough to be held on a single punched card, yet detailed
enough to load application programs occupying hundreds of
punched cards. Many other short programs were needed to per­
form a variety of common utilitarian functions.7
Such programs, known as "utility programs" or simply "utili­
ties," thrived by 1951 in the UNIVAC environment, where all
programs and data files reached the computer in the form of
magnetic tapes. The UNIVAC operator was helpless without utili­
ties of various kinds to manage job setups and instructions, de­
bugging sessions, simple copying tasks, and so on, and UNIVAC
had no provision for entering data or instructions directly into
main memory from a card reader; information transfer between
cards and tapes was performed off-line. Unlike UNIVAC, the IBM
Defense Calculator could receive data and instructions directly
from a card reader as well as from magnetic tape units. This
proved to be a significant convenience.
In August 1952, when some thirty representatives of potential
customers for the Defense Calculator (already better known as
the IBM 701) came to Poughkeepsie to attend a week-long cus­
Copyrighted Material
186 Building IBM

tomer training class, IBM programmers described to them three

dozen or so utilities written for the 701. These included ones for
loading programs into memory, for converting numbers from
decimal to binary form, for moving information from magnetic
tape to memory, for copying tapes, and for printing results.
The IBM programmers also described two different "assembly
programs" that automatically translated instructions from deci­
mal to binary form and combined subroutines and other sepa­
rately written program pieces into a single program. The latter
function relieved programmers of tedious clerical activity such as
the renumbering of program instructions. For example, each
time a subroutine was incorporated into a program, its instruc­
tions had to be given addresses corresponding to the memory
locations it would occupy. Subroutines were therefore normally
written with symbolic m�mory addresses such as S1, S2, and S3,
which were converted to real memory addresses for each use.
Assembly programs automatically handled this conversion.8

Sharing Information and Programs

The IBM-hosted training class also provided customers with their

first opportunity to test programs they had written, in this case
by running them on an engineering model of the 701. For cus­
tomers, an exciting aspect of acquiring a computer was the op­
portunity to program it to handle their own jobs in their own way.
Programs and subroutines written at IBM initially were viewed as
examples of the types of programs that installations might require
rather than as programs any installation would actually use. As
the magnitude of the programming effort came to be better
understood, however, customers were increasingly willing to use
IBM-developed programs, particularly in the cases of utilities and
standard mathematical functions. Customers also began sharing
programs among themselves. Installations took pride in develop­
ing programs and subroutines that were particularly effective and
used by others.
Activities leading to organized customer sharing of computer
programs and other information began when several participants
at the August 1952 customer training class decided to continue
discussing mutual problems on an informal basis. Their first
meeting was held in February 1953 during an AlEE-IRE Com­
puter Conference in Los Angeles. It was then agreed to hold a
second meeting after �6p CWtJ h1f�grMH�1iJpeir own 701s installed.9
 Programming Computers 187

The second meeting was hosted by the Douglas Aircraft Company,

in Santa Monica, California, in August 1953. It was attended by
representatives from five companies with installed 701s and six
customers that had ordered but not yet received theirs.
Representatives of IBM also attended. Among these was
Nathaniel Rochester, chief architect of the 701. Upon returning
from the conference, he reported that the most "striking feature"
of customer activities was the extent to which they were adopting
or improving on techniques "to let the 701 aid in problem prepa­
ration. " Commenting on this "use and extension of the principal
techniques in which IBM did pioneering work in New York and
Poughkeepsie," he observed, "interpretive programs and assem­
bly programs seem likely to play an important part in the use of
any such general purpose machines that we make. "10
Informal meetings of customers continued. Then following an
IBM symposium in Los Angeles in August 1955, the RAND Cor­
poration hosted a meeting of representatives of seventeen organi­
zations that had ordered IBM 704s. During this meeting, the
computer industry'S first "user organization" was formally cre­
ated. Given the name SHARE, its purpose was to promote the
sharing of information and programs among users of IBM's scien­
tific- and engineering-oriented 704 computer and to influence
the company's future developments in hardware and program­
ming support. Nine months later SHARE could boast a member­
ship of forty-seven installations, including all "announced
prospective users" of the 704. Among the benefits of membership
was a collection of three hundred computer programs, available
from an IBM-maintained library.ll
The benefits of a user organization were soon evident to others.
In December 1955 early purchasers of Remington Rand's ERA
1103A formed a user organization called USE, an acronym for
Univac Scientific Exchange.12 User organizations for Remington
Rand's UNIVAC, and for Burroughs and Bendix computers, were
subsequently formed in 1956, as was GUIDE, the organization for
users of IBM's business-oriented computers.13
The cooperative nature of a user organization is suggested by
excerpts from "Obligations of a SHARE Member" printed in its
manual: "It is expected that each member approach each discus­
sion with an open mind, and, having respect for the competence
of other members, be willing to accept the opinions of others
more frequently than he insists on his own .. . . When it comes
to stan dards, SHARE��HgR1ea.�1Me to them for communi-
188 Building IBM

cation purposes through SHARE channels to the extent that it

refuses to distribute material not in SHARE language. Of course,
decisions of SHARE can in no way be binding on any member
installation so far as its internal operation is concerned." The
manual also urged that each member be represented at each
meeting by at least two people, "one empowered to make basic
policy decisions and another thoroughly familiar with techniques,
programming and detailed operating matters. "14
The SHARE organization made numerous contributions to the
development of computer programming. Of particular sig­
nificance was the pioneering work of several West Coast members
on operating system.r---programs that manage and allocate the hard­
ware and software resources of a computer to facilitate the writing
and running of application programs. The company's first oper­
ating system was a direct outgrowth of the pioneering activities of
these members of SHARE. Developed for the IBM 709 computer
and introduced in 1959, its origin is suggested by its name, SOS,
an acronym for SHARE Operating System.15

Customer Support

Support for customers of the IBM 701 had begun more than six
months before deliveries, with the August 1952 training class in
Poughkeepsie. More hands-on experience was available to cus­
tomers following delivery in December 1952 of the first produc­
tion model to the Technical Computing Bureau located at IBM
World Headquarters in New York City. Here the day's activities
were divided into three eight-hour shifts. One shift was available
for program debugging by organizations that had ordered 701s.
Another shift was devoted to conventional service bureau work,
with personnel and machine time charged to customers. A third
shift was reserved for maintenance. Classes on programming the
701 were also offered more or less continually.16
A staff of approximately thirty people was responsible for pro­
gramming and operating the system. This was somewhat fewer
than were employed by most 701 installations, but the average
experience and education was higher. All members had at least
a bachelor's degree in mathematics, about half had master's
degrees, and a few had doctor's degrees. Many had experience
programming and operating the CPC and SSEC computers.17
"Although we ran the 701 seven days a week, 52 weeks a year,
it was during the day �r�e*,�gh Friday, that "spit and
 Programming Computers 189

polish" prevailed," the manager of the operation recalls. This of

course meant the unwritten IBM dress code was in force: white
shirts, neckties, and dark suits for the men and office attire for
the women. The 701 had replaced the SSEC as IBM's showcase
computer.
So proud was Watson, Sr., of this evidence of the company's
entry into large-scale electronic computing that he told one of
his assistants to get rid of the electromechanical card reader in
the same room. As recalled by the manager of the bureau, "When
I explained to the assistant that the card reader was the only way
we could get information into the 701, he said that made no
difference; it had to go. He said Watson felt that the card reader
detracted from the rest of the totally electronic aspects of the
701." In a significant departure from the norm, the older Wat­
son's desires in this case had to yield to necessity. The card reader
stayed. IS
Programming and customer support for the 702 developed
somewhat differently than for the 701, but the impact on custom­
ers was similar. In April 1954, six months after the IBM 702 was
announced, Yin Learson ordered the creation of a new depart­
ment to write application programs for the 702 for demonstra­
tion purposes and for use in courses offered to customers. In
addition to showing customers how the 702 could handle tasks
done with EAM equipment, it was hoped that these model pro­
grams could be adapted by customers to satisfy many of their own
requirements.
Customers were also given the opportunity to run their own
programs on a 702 before their computers were installed. Initially
provided on the engineering model, this service was shifted to
the first production machine after it was installed in Poughkeep­
sie in February 1955. IBM 701 and 702 installations were typically
doing useful work within a few days or weeks, in part because
many customer programs had already been debugged on a system
at IBM to assure they could perform satisfactorily.19 Customer
support, of course, went well beyond activities cited here and was
a major responsibility of the branch offices.
The importance of customer support can be gauged in part by
the experience of General Electric, the first UNIVAC customer
with conventional accounting applications. The computer was
shipped to its Major Appliance Division in Louisville, Kentucky,
in January 1954. Installation and checkout were completed in
April, but the first us6tuJ>�t achieved until October,
190 Building IBM

when the payroll program was finally operationa1.20 Even then, its
performance was not satisfactory. According to one account:
"Even when the computer was run 24 hours a day, there was not
enough time to get out the payroll. The first group who worked
on the project at GE was fired, and a new group was brought in.
By the time the installation was working, late in 1956, the whole
area of automated business computing, and Univac especially,
had become very questionable in the eyes of many businessmen."
More than six months before the UNIVAC at GE was able to
handle its work load, the last-installed IBM 702 was successfully
processing a far larger payroll for the Ford Motor Company.21

Toward High Level Languages

Among those assigned to the 701 computer operation in New

York City was John Backus. He had joined IBM as a programmer
for the SSEC in September 1950, soon after earning a master's
degree in mathematics at Columbia University. In an early IBM
assignment, Backus teamed up with Ted Codd to devise an auto­
mated means for locating the source of machine errors during
the running of programs on the SSEC. Another early assignment
was to study automatic means for converting programs written
and debugged on the CPC to operate on the SSEC. The knowl­
edge Backus gained about CPC programming, and his experience
in automating SSEC programming and operations, was influential
in his later work.22
Beginning in January 1953, Backus and five principal col­
leagues developed a programming system for the 701, called
Speedcoding. Its primary purpose was to facilitate the writing of
programs using floating-point arithmetic. Floating-point (also
called scientific) notation was widely used by engineers and sci­
entists because it expressed very large and very small numbers
just as simply as numbers in more common use. To programmers,
the benefits of floating-point arithmetic were more mundane, but
just as important. A machine performing floating-point arithme­
tic kept track of decimal-point locations, thus relieving the pro­
grammer of this burdensome, error-prone activity.
The inspiration for Speedcoding was a CPC mode of operation
devised at the Los Alamos Scientific Laboratory in 1950. Rather
than using a differently wired 604 plugboard for each problem,
the Los Alamos group had devised a general-purpose, floating­
point plugboard wiri�te�mmJtl perform the four basic
 Programming Computers 191

Figure 14.1. John W. Backus

John Backus led the development of the FORTRAN language and its first
compiler, which was delivered to customers beginning in April 1957.
Awarded the National Medal of Science in 1976 and the Draper Prize for en­
gineering in 1993 for his work on FORTRAN, he was the first IBM employee
to receive either of these top-level awards.

Copyrighted Material
192 Building IBM

arithmetic operations (addition, subtraction, multiplication, and

division) and also square roots. When numbers were represented
on punched cards in the prescribed floating-point format, and a
deck of cards was used to sequence through the desired arithme­
tic computations, the plugboard carried out the specified opera­
tions, automatically maintaining the decimal-point location and
the correct power of ten.23
To provide this capability on the 701, Backus and his colleagues.
devised a Speedcoding instruction format that provided simulta­
neously for one floating-point and one fixed-point operation
code. The floating-point operation code was associated with three
memor y addresses, as had been the floating-point implementa­
tion for the CPC. This format allowed the programmer to specify
the memory addresses of two floating-point numbers (operands) ,
the operation to be performed, and the memory address to which
the floating-point result was to be sent. The fixed-point operation
code referenced a single memory address and was used to imple­
ment control functions such as "indexing" and "branching."
The Speedcoding system gave the user the illusion of perform­
ing all functions of each four-address instruction in one step. Had
the instructions been executed on the SSEC, the illusion would
have been closer to reality because the SSEC was equipped with
two high-speed electronic registers for instructions and six for
numerical data.24 The 701, however, had only one high-speed data
register, so a program was needed to break each Speedcoding
instruction into a series of steps and also to call on subroutines
for performing floating-point arithmetic and other operations
not built into the 701. Because this interpretive function had to
be performed each time the application program was executed,
the time for execution was typically fifteen times longer than for
an equivalent program written in machine language. Neverthe­
less, by the fall of 1953, the ease of programming offered by
Speedcoding had attracted several users. 25
As planning for an improved 701 (the IBM 704) began in 1953,
Backus was instrumental in the decision to implement floating­
point and indexing capabilities directly in electronic circuits.
Speedcoding would thus have little value on the 704. The addi­
tion of these and other instructions to the 704 repertoire made
assembly language programming easier and more effective, but
it did not eliminate the need for more sophisticated tools to
reduce the cost of programming.

Copyrighted Material
 Programming Computers 193

His work at the Scientific Computing Service Center provided

Backus an opportunity to observe the economics and practical
problems of computing. At the time, the Service Center had the
only stored-program computer installation in the country with a
primary function of selling computer services. Thus it was the
only installation where cost accounting for equipment, personnel,
and supplies was intrinsic to its business. Its second primary
function-of helping customers test programs before they re­
ceived their own computers-provided further insight into cus­
tomer requirements.
Information from customer facilities, and IBM's own, revealed
that the cost of programmers associated with a typical 701 com­
puter installation was at least as great as the cost of the machine
itself. Also nearly half of the useful computer time was spent
debugging programs. "Thus programming and debugging ac­
counted for as much as three-quarters of the cost of operating a
computer," Backus observed, "and obviously, as computers got
cheaper, this situation would get worse. "26 The economics of
computing provided a compelling rationale for developing easier
and more cost-effective means for programming computers.
From his own work on Speedcoding and his observation of
other early steps toward "automatic programming," Backus saw
the possibility of creating what came to be known as a high level
language-that is, a computer language whose operational codes
and structures are designed to conform to the needs of the user
rather than to the machine architecture. A cost-effective means
for automatically translating programs written in the high level
language to machine code was, of course, mandatory.
For application programs that might be executed repeatedly,
Backus believed a fifteenfold penalty in execution time (as expe­
rienced with Speedcoding) would be unacceptable.27 The solu­
tion he proposed was to translate the entire program from high
level language to machine language before execution rather than
interpreting each high level instruction individually just before
execution. This process, known as compiling, caused the time-con­
suming translation of instructions to be done only once-rather
than each time the program was run.
The advantages of a compiler had been advanced a year earlier
by Grace Murray Hopper, who had earned a Ph.D. in mathemat­
ics at Yale in 1934 and had done her first computer programming
on the IBM-Harvard Mark I while serving in the navy. In 1949

Copyrighted Material
194 Building IBM

she joined the Eckert-Mauchly Computer Corporation as a senior

mathematician and did pioneering work on computer program­
ming.28 Describing her activities at a May 1952 meeting of the
Association for Computing Machinery, Hopper provocatively as­
serted, "It is the current aim to replace, as far as possible, the
human brain by an electronic digital computer." A key element
of her presentation related to the benefits of a compiler.29 By late
1953 some UNIVAC customers were making use of her "A-2"
language and compiler. The statements in the A-2 language were
similar in format to those in Speedcoding, but the language
lacked the control and indexing features of Speedcoding.3o
Backus's objective, by contrast, was to create a high level lan­
guage far richer and more user-friendly than Speedcoding and
still be able to compile its programs into efficient machine
language.

The Birth of FORTRAN

By the end of 1953, Backus had won approval for a project to

create an "algebraic programming language" that would make
use of common mathematical notation. It was to be structured
for easy learning and use by mathematicians, scientists, and engi­
neers familiar with standard algebraic notation. A compiler was
to be created to translate programs written in the high level
language to machine language instructions.
Backus was joined by Irving Ziller, who had been hired into
IBM's computing service bureau in 1952 after graduating from
Brooklyn College. As part of his initial assignment, Ziller had
made use of the similarity between CPC programming and Speed­
coding to transfer many customer applications from the CPC to
the more powerful and cost-effective 701 computer. The third
member of the team was Harlan L. Herrick, who had joined the
SSEC programming staff in November 1948. Herrick had a mas­
ter's degree in mathematics from the University of Iowa and two
additional years of teaching and graduate studies at Yale. Crucial
support was provided by Roy Nutt, an experienced programmer
on loan part time from the United Aircraft Corporation.
By November 1954 these individuals had established the gen­
eral features of the proposed language. They had also written a
descriptive document titled "Preliminary Report, Specifica­
tions for the IBM Mathematical FORmula TRANslating System,
FORTRAN. "
Copyrighted Material
 Programming Computers 195

That winter Backus, Ziller, and Herrick visited several custom­

ers who had ordered 704 computers, seeking their reaction to the
FORTRAN proposaJ.31 Most of the customers were skeptical. They
shared a concern Backus and Herrick had themselves expressed
during a symposium on "Automatic Programming for Digital
Computers" in May 1954: "Can a machine translate a sufficiently
rich mathematical language into a sufficiently economical ma­
chine program at a sufficiently low cost to make the whole affair
feasible?"32 An unexpectedly strong endorsement for the project
was provided by United Aircraft, which made Nutt available on a
regular basis to help develop the compiler.33
The programming and testing of the FORTRAN compiler be­
gan early in 1955 and required more than two years. A key
problem was to achieve efficient use of index registers as aids to
accessing arrays of data during iterative computations. More than
a dozen programmers were assigned to the compiler project
before it was completed. A programmer's manual was made avail­
able in October 1956, and a paper (authored by ten IBM employ­
ees and three people on loan to the project from other
organizations) was presented at a computer conference two
months before shipment of the FORTRAN compiler commenced
in April 1957.34
So enthusiastic were users of the new language that IBM devel­
oped FORTRAN compilers for all its stored-program computers.
Most competitors did the same within only a few years. Applica­
tions programmed in FORTRAN for one machine could be con­
verted with relative ease for use on another. Thus FORTRAN
became the common computer language by which scientists and
engineers shared programs and computational processes. So
pervasive was the use of the language that most college courses
on computing during the 1960s caused students to believe that
learning FORTRAN was synonymous with learning to program
computers.35
''We did expect our system to have a big impact, in the sense
that it would make programming for the 704 very much faster,
cheaper, more reliable," Backus recalls. ''We also expected that ,
if we were successful in meeting our goals, other groups and
manufacturers would follow our example in reducing the cost of
programming by providing similar systems with different but
similar languages." Beyond that, however, the objectives of the
creators of FORTRAN were relatively modest. According to
�
Backus, ''We certainly �a �naNlaM�rguages almost identical
196 Building IBM

to the one we were working on would be used for more than one
IBM computer, not to mention those of other manufacturers. "36
Consistent with the practice of the time, IBM made no proprie­
tary claim to the language, and a reel of magnetic tape containing
the compiler was provided to customers at no additional charge.

COBOL Arrives

More than a hundred high level languages were developed dur­

ing the next few years. Many were inspired by the remarkable
success of FORTRAN. Most were little used and have been long
forgotten. An important exception is COBOL, a language de­
signed for business applications. Its origins can be traced to an
April 1959 gathering of six representatives from computer users
and manufacturers. Their purpose was to prepare an agenda for
a meeting to consider the development of a "common business
language." A long-range goal implicit in the word "common" was
to forge an industrywide standardization that was not likely
to result from the competitive language developments then
underway.
The Department of Defense believed such standardization was
in the national interest and agreed to host the meeting in the
Pentagon in May. In attendance were representatives of Bur­
roughs, CE, Honeywell, IBM, NCR, Philco, RCA, Remington
Rand, Sylvania, and ICT, as well as seven government organiza­
tions and eleven users. The key decision of the meeting was to
proceed with a program having both short-range and long-range
goals. A hierarchy of committees, with broad industry repre­
sentation, was established to carry out the program.
The short-range goal was rapid development of an initial busi­
ness language, using simple English terms as much as possible.
This language was to be used until more sophisticated long-range
language goals could be achieved. Two already-defined business
languages, IBM's COMTRAN (later renamed Commercial Trans­
lator) and Sperry-Rand's Flow-Matic were influential in commit­
tee deliberations. In September the short-range language was
named COBOL, an acronym for COmmon Business Oriented
Language. The initial language definition process was completed
by December, although substantial revisions were subsequently
made, especially during the first year after its initial availability.
By the end of 1960, RCA and Remington Rand had produced
operational compilere l.f}J-
o 1Qj'ite'i'FMat�
� in a public display to
 Programming ComputiTS 197

demonstrate that the same COBOL program could be run on

different computers.37 Busy with implementing its Commercial
Translator, IBM deferred COBOL until the revised specifications
(COBOL 61) were developed. Its first COBOL compiler was
delivered in January 1962.38
Within a few years, COBOL became as dominant for program­
ming business applications as FORTRAN was for engineering and
science. The attempt by IBM to achieve broad market acceptance
of its own Commercial Translator for business applications had
failed. Important to COBOL's success was ongoing support by the
industry committee. Originally intended only as a stop-gap meas­
ure, COBOL was so successful that it obviated the need for the
planned longer-range efforts.
Within ten years after FORTRAN was introduced, it and the
subsequently introduced COBOL were used for writing nearly
half of all application programs, and their usage was continuing
to increase.39 T he cost-effectiveness of high level languages pio­
neered by FORTRAN had been confirmed. Even a quarter of a
century after FORTRAN was first introduced, a survey revealed
that "90 percent of all computers-both IBM and non-IBM-do­
ing engineering and scientific work use this programming lan­
guage more than 70 percent of the time."4O

Copyrighted Material
 15
An Air Defense System

More than a year before the first UNIVAC was accepted by the
Census Bureau-at a time when the IBM CPC was the most
advanced electronic computer in the market-a government­
sponsored study recommended development of an on-line, real­
time air defense system. Dozens of large-scale stored-program
computers were to be interconnected with each other and with
radar units and other specialized input-output equipment dis­
persed across the country. It was January 1950, five months after
the Soviet Union tested its first atomic bomb. Many of the tech­
nologies needed for the proposed system did not exist, but no
other solution seemed more likely of success in defending the
United States against possible attack by Soviet warplanes carrying
atomic b o mbs l.

Recognizing the significance of the project when he learned

about it in 1952, Tom Watson decided participation was essential
to the company's future. "Pulling out the stops" to win an initial
contract to work with the Massachusetts Institute of Technology
(MIT) in building prototype computers, he was still not assured
of getting the really big contract to manufacture, install, and
service the dozens of large computers needed for the system. "I
worked harder to win that contract tha n I worked for any other
sale in my life," Watson recalled.2 Ironically the person who
initiated IBM's involvement was John McPherson-the man Tom
Watson had tried to force out of the company only a few years
earlier.

The Whirlwind Computer

The technological basis for the proposed air defense system was
Whirlwind, an electronic digital computer under development at
MIT since 1946. Th�tlih �h j � nd, however, dates hack
200 Building IBM

three years earlier when Capt. Luis de Florez, director of the

Special Devices Division of the U.S. Navy Bureau of Aeronautics,
approached the Bell Telephone Laboratories and MIT (his alma
mater) about the possibility of developing a flight trainer capable
of simulating many different aircraft. Out of these discussions
evolved a navy-funded project to develop an "airplane stability
and control analyzer." The project had two objectives: first, a
universal trainer, and second, a means for simulating the perfor­
mance of new aircraft.
The project was officially launched in December 1944 when a
$75,000 feasibility study contract was awarded to the MIT Servo­
mechanisms Laboratory. In charge of the project was Jay W.
Forrester, a self-assured twenty-six-year-old assistant director of the
Servomechanisms Laboratory. Forrester's credentials included a
bachelor's degree in electrical engineering from the University of
Nebraska and five years' experience as a graduate student and
research assistant in electrical engineering. Less than half a year
later, Forrester endorsed the replacement of the small feasibility
study contract with one calling for completion of the project in
eighteen months for nearly twelve times the money, approxi­
mately $875,000. As the summer of 1945 wore on, however, he
became increasingly concerned that the equipment his group was
developing lacked the necessary flexibility and performance.3
Seeking alternative solutions, Forrester began considering a
shift from traditional analog methods to newer digital methods
recommended by Perry O . Crawford, Jr., a fellow graduate stu­
dent in electrical engineering. Crawford's master's degree thesis
(submitted three years earlier) had made the innovative proposal
that "automatic control" of "real-time" activities such as directing
antiaircraft fire could be carried out by means of numerical
co mputations rather than the analog methods then in use.4
When submitted in 1942, Crawford's proposal seemed aca­
demic. Computing devices with the speed of electronic circuits
were needed, but ver y little had been done to develop them.s By
late 1945 the situation had changed dramatically. The ENIAC,
whose electronic circuits contained 18,000 vacuum tubes, was
being readied for its first full-scale test, and von Neumann's "First
Draft of a Report on EDVAC" (distributed that summer) had
stimulated additional interest in electronic digital computers.
Eager to participate in the exciting new field of electronic
digital computers, Forrester also saw it as the best solution to the
problems his projec tc;ttJj1i-t!JhMtt �5la/1946, following several
 An Air Defeme System 201

months of discussions with the navy, he formally proposed that

his project's objectives be revised. Its new tasks would be, first, the
design of an electronic digital computer capable of handling the
aircraft stability and control analyzer problem, and second, con­
struction and incorporation of the computer in the flight trainer
and analyzer equipment.
The cost of the project was now estimated at $2.4 million-ap­
proximately three times more than had been estimated for an
analog solution. Nevertheless the navy liked what it saw. If the
project was successful, it would be the first time digital techniques
were used to implement automatic control of a large system.
While substantial risks were inherent in such a pioneering effort,
there was also the prospect of achieving substantially improved
results.
The navy authorized Forrester to undertake research and de­
velopment of digital technologies for incorporation in a small
computer to demonstrate feasibility and to design (but not build)
the proposed electronic computer and aircraft analyzer. These
two phases of the project were to be completed by June 1948 at
a cost of $1.2 million. The newly defined project was named
''Whirlwind,'' consistent with the names Hurricane and Tornado
given to other newly funded navy computer projects.6
Among those who encouraged the navy to accept the shift from
analog to digital methods was Perry Crawford. Having left MIT
in October 1945 to join de Florez in the Special Devices Division
of the navy, Crawford now included among his responsibilities the
review of Forrester's project for the navy. If the use of digital
methods for automatic control could be validated by this project,
the significance of his thesis and his own reputation would be
greatly enhanced. This cozy relationship between project leader
and project evaluator helped support Forrester's normal re­
sponse to difficult problems, which was to seek more resources.
By 1947 Project Whirlwind had grown to fifty people, half of
whom were highly trained engineers.7
Pressure on the navy to reduce expenditures was growing in
the postwar era, causing Crawford and Forrester to consider
terminating work on the cockpit and ancillar y gear. This would
free up resources for development of the all-important computer,
but it would also cause political problems because the cockpit and
ancillar y gear represented to many the primary purpose of the
project. Nevertheless Forrester, with full support of Crawford, had
terminated all such � ��Ll��Jfhere was now only one
202 Building IBM

objective for Project Whirlwind: the design of a general-purpose

computer capable of handling real-time problems.s
A critical evaluation of Project Whirlwind, by an ad hoc panel
commissioned in 1949 by the Department of Defense, stiffened
the navy's resistance to further cost increases. The cost of com­
pleting Whirlwind was projected to be two to three times higher
than the combined costs of completing von Neumann's lAS com­
puter, Eckert and Mauchly's UNIVAC, and the Moore School's
E DVAC. The ad hoc panel commended Project Whirlwind for the
excellence of its staff, for the quality of its reports, and for training
graduate students. It concluded, however, that the development
costs were not justifiable for any known application. It recom­
mended the project be terminated if no suitable application
could be found.9
Seeking a use for Whirlwind that was sufficiently important to
justify its development costs, Forrester undertook a one-year study
of its potential use in air traffic control. This study was begun in
March 1949 with air force funding.

The New Mission

Five months later the Soviet Union detonated its first atomic
bomb. In response to this threat, U.S. militar y organizations
initiated a series of actions that resulted in the selection of Whirl­
wind as the central element of a nationwide air defense system.
Among these early actions was creation by the air force of the Air
Defense System Engineering Committee to find "the best solution
of the problem of Air Defense." The committee was chaired by
George E. Valley, an MIT professor who was also a member of the
Air Force Scientific Advisor y Board.
The committee's preliminar y findings, submitted in Januar y
1950, were consistent with those of a separate study carried out
by the Weapon System Evaluation Group (WSEG) of the Depart­
ment of Defense. Both studies found the existing air defense
system to·be woefully inadequate. Improvements were needed in
radar, antiaircraft artiller y, ground-to-air missiles, and fighter air­
craft. The most important need, however, was for command and
control centers capable of receiving and evaluating information
from many sources and of working together to create a coordi­
nated nationwide response to any airborne threat. 10
Electronic speed would be required for all parts of a modern­
ized system, includina:. obser �ation. eVal }lation, and communica­
l;;;opyngfjrea Marena
 An Air Defense System 203

tion. Such a system had much in common with the air traffic
control system Forrester's group was already studying for the air
force. It was also reminiscent of Crawford's 1942 master's degree
thesis in which he analyzed the possible use of electronic digital
computation methods for "the prediction of the future position
of the target" for the "control of anti-aircraft gunfire. nIl
These facts would have made Whirlwind a strong candidate in
the thinking of any group asked to consider solutions to the air
defense problem. But with MIT professor George Valley as chair­
man, the committee quickly and decisively endorsed Whirlwind
as the "best solution." Informal approval from the air force came
almost as rapidly. Fear that Soviet aircraft carrying nuclear bombs
were a primary threat to the United States gave the project the
highest priority of the Department of Defense.I2 At last an appli­
cation had been found for Whirlwind with requirements for high
performance and reliability-and sufficient urgency-to justify
the level of expenditure Forrester had always deemed necessary.
In response to a December 1950 request by the air force, MIT
established Project Lincoln and began to construct a laboratory
dedicated to the air-defense problem. Located at at the Hanscom
Air Force Base a few miles west of Boston, the facilities were called
the Lincoln Laboratory. IS Project Whirlwind continued to be
located in Cambridge at the MIT Digital Computer Laborator y,
which was directed by Forrester. Activities and persons concerned
with Whirlwind and its use in the air defense system were organi­
zationally part of the Lincoln Laboratory Division VI, which was
also headed by Forrester.
Forrester did not wait for these complex organizational matters
to be resolved before shifting the direction of his project. In April
1950, with full agreement of the air force, he discontinued work
on the air traffic control problem and began work on the air
defense system. More than a year later Oune 1951) a contract
went into effect that covered the new project objectives and
agreements between MIT and the air force.I4
Final assembly and test of the Whirlwind computer was under­
way during the last half of 1950. The most critical component was
memory. It consisted of specially constructed cathode-ray tubes,
each costing approximately $1,500 and designed to store 256 bits
of information in a 16 x 16 array of electrostatically charged
regions on its front surface. The first bank of 16 tubes was
installed in Whirlwind in June 1950. All tubes functioned as
designed, except tha rcBpy}t�rnt9t�1 reliability of the storage
204 Building IBM

surfaces were not satisfactor y. Mter three months of intensive

effort to solve the problem, Forrester could only report that "final
testing and alignment of the first storage bank is moving along
steadily. "15
By early 1951 the memory was functioning well enough that
Whirlwind was connected to microwave early warning (MEW)
radar for experiments in tracking the course of two airplanes,
predicting their future locations, and calculating data for voice
transmission to intercepting aircraft. Soon after this capability
had been demonstrated in April, a decision was made to construct
a feasibility test model of the air-defense system in the region of
Cape Cod. Named the Cape Cod System, it was to incorporate at
least one of each essential element of the final system so that
full-system operation could be simulated.
The Cape Cod System was fully operational by September 1953,
less than two and one half years after work on the system began.
Its success was made possible by improvements in electronic cir­
cuitry, display technology, data handling procedures, methods of
analysis, computer programming, and-perhaps most important
of all-the use of a new memory technology pioneered by Jay
Forrester. Known as magnetic core, ferrite core, or simply core,
the new memory technology provided a dramatic improvement
in computer reliability. 16

A Better Memory

Main memory for Whirlwind had been a critical concern for

Forrester from the beginning. The high-speed random access
capability needed to satisfy real-time response requirements could
not be achieved by the magnetic drums pioneered by ERA or by
the mercury delay-line memor y developed for EDVAC. Of all
memory technologies then being developed, only electrostatic
storage of information within a cathode ray tube provided ade­
quate speed. But this type of memory occupied a large space, was
expensive to build, and was difficult to make reliable .
"Storage tubes do not represent the ultimate in data storage
devices," Forrester obser ved in April 1947. He then put forward
his own ideas for "efficient storage," which consisted of individual
elements "closely spaced in a 3-dimensional volume." In his first
attempt he used gas glow discharge tubes. These possessed a
"suitable form of nonlinear impedance," but seemingly uncon-
Copyrighted Material
 An Air Defmse S.vslem 205

,,8 Jrl Jla R!Lx.HLlL

.. ��

Figure 15.1. Jay W. Forrester and Memory Devices

Jay Forrester is shown in 1950( top) with an electrostatic storage tube of the
type initially used in Whirlwind. A small corner of a 64 x 64 core plane pio­
neered by Forrester and used in the XD-J computer (bottom, left) shows
how four wires were threaded through each core: three for reading and writ­
ing information and the fourth to detect the output signal . Core size is indi­
cated b y U.S. dime; outside diameter was 80 mil (0.2 centimeter). The SAGE
computer memor y array (bottom, right) contains thirty-six core planes,
three of which are sp ares . (Top photograph courtesy of the MITRE Corpora­
tion Archives.)

Copyrighted Material
206 Building IBM

trollable electrical characteristics forced him to abandon the

effort. 17
Two years later in April 1949, Forrester saw an advertisement
for Deltamax, a ferromagnetic metal produced by the Arnold
Engineering Company of Chicago. Its remarkably sharp magnetic
switching threshold caused him to consider i ts use for storing
information in a compact 3-dimensional array of tiny magnetic
rings (cores ). The Is and Os of binary information would be
represented by the clockwise or counterclockwise direction of
magnetization of doughnut-shaped cores made of Deltamax.I8
The idea of storing information in magnetic cores was not new.
Several other researchers had considered this possibility.I9 But
Forrester's method for wiring magnetic cores into a three-dimen­
sional array and selecting them electrically was new. Fundamental
to his proposal was a magnetic material with a well-defined switch­
ing threshold. This threshold needed to be sharp enough that
coincidence of electric current pulses in two wires through a core
would reverse its magnetization direction, whereas one of these
pulses alone would be too small to alter the magnetization state
of any core through which it passed. Using this concept, Forrester
devised a way to wire cores into a three-dimensional array so as
to minimize the number of electronic circuits needed to read and
write information anywhere in the memory.
Forrester carefully documented his ideas in his engineering
notebook and carried out a number of experiments that sum­
mer.20 Simple theory and experiment did not agree as well as he
hoped, however, so he selected a graduate student to carry out
more detailed experiments as part of the master's degree pro­
gram. The task was to relate theory to experiment and to find a
suitable core material for the proposed memory.21
An early suspicion, subsequently verified by experiment, was
that metal cores would not switch fas t enough because of electric
eddy currents induced within the core itself. As a result, Forrester
initia ted a program at MIT to develop materials with suitable
magnetic characteristics that did not conduct electricity. Some of
the work on ferrite materials was subcontracted to the General
Ceramics and Steatite Corporation of Keasbey, New Jersey. With
Whirlwind's huge government funding, Forrester had little
trouble paying for activities he regarded as importan t.
In May 1952 successful operation of a 16 x 16 array of ferrite
cores had been demonstrated. One year later the first ferrite core

Copyrighted Material
 An Air Defense System 207

memory with enough storage capacity to ser ve as a main memory

of a computer was assembled and installed in a small computer,
specially designed to test memory functions. The memory had a
capacity of 17,408 bits of information in a 32 x 32 x 17 array of
ferrite cores. By September 1953 two ferrite core arrays of this
size had been tested and installed in Whirlwind .
Somewhat faster than the cathode-ray tube memories they re­
placed, the ferrite core memories were most notable for their
superior reliability. The electronic drive and sense circuits some­
times failed, but the tiny doughnut-shaped cores almost never
failed. A new memory technology had been developed that could
satisfy the speed and reliability needs of Whirlwind.22

Selecting IBM

In June 1952, fifteen months before ferrite core memories were

installed in Whirlwind, Forrester was already considering prob­
lems of manufacturing, installing, and servicing the computers
for the proposed air defense system. During a professional society
committee meeting in New York City, one of his group leaders
had informally discussed the subj ect with John McPherson.
Learning that "McPherson was very interested ," Forrester said ,
"the trouble with IBM would be its traditional secretiveness."
Then he cautioned , "A number of companies need to be consid­
ered. . . . However, we must avoid indiscriminate discussion of
this problem with different manufacturers who might create con­
siderable disturbance in competition and pressure to receive the
job."23
The "competition and pressure to receive the job" were as
intense as Forrester anticipated, but IBM's "traditional secre tive­
ness" was not a problem. Tom Watson wanted the con tract in spite
of what he considered to be large inherent risks. One of these
risks had been identified in a particularly pointed manner by the
company's treasurer, AI Williams. Could a mistake in computa­
tion, he asked, "result in the accidental destruction of one of our
country's own airplanes, with the resultant financial exposure and
publicity such an accident might entail?"24
Watson believed the contract was so significant that whoever
won it would become the kingpin of the emerging electronic
computer industry. Knowing he could not obtain the contract
without revealing IBM's engineering and manufacturing

Copyrighted Material
208 Building IBM

strengths , he treated Forrester and his colleagues to one of

the more open tours of laboratories and plants ever given to
outsiders.25
During the summer of 1 952, Forrester and three of his top
technical leaders visited the research, development, and manu­
facturing facilities of IBM and of other leading contenders.
Raytheon was a strong contender because of its wartime activities
in electronics and its development of the Hurricane computer
under contract to the navy. Remington Rand was an even stronger
contender. Under the leadership of James H. Rand, Jr., it had just
acquired ERA, and it had acquired the Eckert-Mauchly Computer
Company two years earlier.
Jim Rand's methods were different from Tom Watson's, but he
was just as eager to win the contract. He personally hosted the
MIT evaluation team on the corporate yacht. Also present on the
yacht was the new chairman of the Remington Rand board, Gen.
Douglas MacArthur, one of the more popular military leaders of
World War II. Knowing little about computers, MacArthur enter­
tained the guests with his obser vations on political and military
matters. Leslie R. Groves, who had managed the Manhattan Proj­
ect in which the first atomic bomb was devel op ed, hosted the
group during the days. Recently hired as a Remington Rand vice
president and director of advanced research, Groves was an im­
pressive salesman for the company's technical capabilities.26
Before the visits were undertaken , Forrester and his team had
decided what capabilities a contractor should have, and they had
devised a method for quantifying the relative importance of these.
Following in-depth reviews, the four of them "were unanimous in
their relative placing of the companies and were unanimous in
feeling that a wide margin existed between each." First was IBM,
followed by Remington Rand and then Raytheon.
Documenting the reasons for selecting IBM over Remington
Rand, Forrester wrote: "In the IBM organization we obser ved a
much higher degree of purposefulness, integration, and esprit de
corps than we found in the Remington Rand organization. Also,
of considerable interest to us, was the evidence of much closer
ties between research, factory, and field maintenance in IBM." By
contrast. he noted that the Eckert-Mauchly and ERA organiza­
tions "had not yet worked with each other or with a Remington
Rand factor y."
Also cited were the superior technical abilities of key staff
members at IBM and the comp'any's leadership in developing,
CopyrightetJ Material
 An Air Defense System 209

manufacturing, installing, and serVIcmg electronic computing

equipment. "At the time of our visit IBM had built much more
electronic digital computing equipment. They had some
3,000,000 vacuum tubes operating in their own equipment in
commercial service." By contrast Forrester noted, "In the
Remington Rand organization suitable factory personnel have
not received training, since most of their electronic computer
construction has been on an expanded model shop basis con­
nected with the research groupS."27
Forrester's report did not reveal that IBM had already made
considerable progress in developing its own magnetic core memo­
ries. However, prior to giving IBM the contract to work with
Lincoln Laborator y in defining the computer for the Air Defense
System, Forrester sent two of his colleagues to review its research
on magnetic core memories.28
The lead engineer for this project was Munro K. (Mike)
Haynes, who had joined IBM in 1950 after completing his doc­
toral research in electrical engineering at the University of Illi­
nois. His doctoral thesis was titled, "Magnetic Cores as Elements
of Digital Computing Systems." It was primarily devoted to the
use of magnetic cores for computer logic, but the last section
described a two-dimensional memory array. Without any appar­
ent knowledge of Forrester's prior work, Haynes had inde­
pendently invented the same type of magnetic core memory
using coincident-current selection, albeit for a two-dimensional
rather than a three-dimensional array.29
Soon after joining IBM, Mike Haynes began seeing General
Ceramics advertisements for ferrite cores with sufficiently sharp
switching thresholds to be used in his proposed memory. He
ordered a variety of these cores, which were typically about an
inch in diameter. Teaming up with a more experienced engineer,
he built a core memory with just enough capacity to store all the
information in an IBM punched card: 960 bits in an 80 x 12 array.
In May 1952 it was successfully tested as a data buffer between a
Type 405 alphabetical accounting machine and a Type 517 sum­
mary punch. This first functional test of a ferrite core memory
was made in the same month that a four-times smaller 16 x 16-bit
ferrite core array was successfully tested at MIT.
Although Mike Haynes could claim some areas of leadership
in magnetic core memor y research, the MIT effort was more
advanced overall. It had started sooner, had a more clearly
defined mission, an dd3�Wi&»ft#t fw�UHr;gpvernment funding, the
210 Building IBM

bread th of MIT's technical capabilities, and Forrester's inspired

leadership .
Particularly significant was Forrester's decision to establish a
ferrite materials research effort in 1 950 when he learned about
ferrite materials with sharp thresholds, produced by General Ce­
ramics. This led to an unwitting cooperation between IBM and
MIT that began in 1 952 when Mike Haynes ordered smaller
ferrite cores than had ever been made by General Ceramics. The
cores had 90 mil (0.23 cm ) outside diameters and 60 mil (0.15
cm) inside diameters. IBM paid for the special dies to fabricate
these cores because there appeared to be no other customer for
them. Learning of their existence, the MIT Lincoln Laboratory
contracted with General Ceramics to develop improved materials
for these smaller cores. Forrester and others at MIT were appar­
ently unaware that I BM had paid for the dies that would soon be
used to fabricate ferrite cores for the first magnetic core memory
installed in Whirlwind.3o

Working Together

In October 1952, IBM received a subcontract from the Lincoln

Laboratory to assist in the final design of the computer and
related equipment for the air-defense system. Off to a fast start,
the company had already assigned a few people in September,
and by November there were fourteen staff-level employees. An
old necktie factory on High Street in Poughkeepsie ser ved as
headquarters. It also provided the name, Project High, for I BM's
part of the air-defense effort. By July there were 203 technical and
26 administrative people assigned. By November 1953 (only thir­
teen months after the initial contract was signed ) the total num­
ber of employees had reached the planned level of about 300.31
Assigning qualified people was only the beginning. The newly
assigned engineers with their industrial culture had to learn to
work with engineers fro m an academic research environment.
Addressing this subject from the MIT perspective in July 1 953,
Forrester said, ''The difficulty in accomplishing the program on
time is increased inevitably by the fact that two organizations are
doing most of the work together. The relationship between them
is new, many of the people in the groups are new to the problem
and to the ways of attacking it."32
Indeed the two groups had rather different backgrounds and
initial expectations. Many MIT engineers believed Whirlwind
Copyrighted Maferial
 An Air Defense System 211

would require ver y little modification. They believed IBM engi­

neers would merely have to "do the production engineering,
whatever that was, and build the necessary quantity." By contrast
many of their counterparts at IBM believed MIT should simply
hand over a few pages of specifications so they could redesign the
computer the right way. Not surprisingly, "the first meetings of
these groups were loud and rancorous," Robert R. Everett (first
in command after Forrester) recalled. "Mter a while, as the two
groups began to know and respect each other, the arguments
became more cogent and took place between individuals instead
of between organizations. "33
The initial IBM group was drawn heavily from those who had
designed the 701 and 702 computers. From Januar y to June of
1953 they thrashed over architectural and hardware designs with
MIT engineers. Many meetings were held midway between
Poughkeepsie and Boston in a back room of the IBM branch
office in Hartford, Connecticut. The first of these meetings was
attended by twelve people from IBM and twenty from MIT.
Among their conclusions was that "final decisions regarding cir­
cuits cannot be made at this time." Of twenty-two different elec­
tronic computer circuits already identified for joint development,
little was known about ten of them and three of the circuits were
described as "complete unknowns."34
Decisions hammered out in subsequent meetings included
such fundamental things as electronic circuit packaging, the man­
ner in which arithmetic would be handled, and the provision of
index registers.35 These decisions were based on knowledgeable
input from both groups, but if agreement could not be reached,
the MIT engineers were authorized to make the final decision.
''This was a very irksome thing for many of our people," IBM's
first project manager recalled. "But it seemed to me that under
the urgency of the situation it couldn't have been any other
way."36
Many more technical decisions were reached at a series of seven
closely spaced meetings held during June and July. Known as
Project Grind, these meetings were structured to encourage par­
ticipants to grind away at urgent technical problems until deci­
sions were reached. Committees were formed to deal with specific
problems. Then as one engineer described it, 'We let the blood
flow on the floor until we came to an agreement."37
Achieving the required system reliability was particularly chal­
lenging, in spite of the substantial contributions already made by
-Copyngnrea Matenal
212 Building IBM

the MIT Whirlwind effort. All components were designed for

operation under "marginal conditions," that is, when all interact­
ing components were performing in their worst allowable man­
ner-for example, low power supply voltages or low signal levels.
The design of circuits to satisfy these conditions and the determi­
nation of "realistic" marginal conditions and test procedures re­
quired ex tensive engineering design, followed by lengthy
discussions and agreement of both groups.
Forrester believes his managemen t approach was crucial to
achieving system reliability. "When a mistake was recognized," he
asserts, "it was admi tted and fixed rather than evaded or denied."
Forrester credits this approach with forcing an early decision to
conduct "marginal checking" of critical circuits in service and
replacing marginal ones rather than waiting for failures during
normal operation. This improved circuit reliability by about a
factor of ten. When it was later recognized that the promised
reliability could still not be achieved with a single computer, the
air force was advised that the original assumptions were wrong.
Each computer would require an identical backup computer in
case the first one failed. This decision was not made until Novem­
ber 1953, after the air force had established a budge t for the
entire system. ''There was a lot of flak from that," Forrester
observes, "bu t our position was that it had to be done."38
To keep the backup computer ready to take over withou t delay,
i t had to carr y ou t many of the same tasks in real time as the
primary computer. The innovative "duplex" mode devised by the
IBM-MIT team permit ted the backup computer to carry out
certain ancillar y functions a t the same time, including au tomatic
marginal checking of circuits.
In September 1953 a contract for two prototype computers was
awarded to IBM. Shipment of the firs t prototype to the Lincoln
Laboratory was scheduled for August of the next year. Inevitably
there were delays. Many of the technical problems were more
complex than an ticipated. Consider just the tasks of individually
fabricating, testing, and wiring 589,824 tiny ferrite cores used by
the memory banks of the two computers. Semiautomatic fabrica­
tion and tes t equipment was devised to speed the process, but the
designs for these were newer than the designs for the pioneering
memory itself. Numerous modifications were required. Even the
electrical tests for distinguishing good cores from bad ones had
to be modified as more experience was gained.39

Copyrighted Material
 An Air Defense System 213

Remarkably, all of the technical problems were overcome in

time to ship the major units of the first prototype computer
(XD-l) to the lincoln Laboratory in Januar y 1955. It successfully
replaced Whirlwind in the air defense test system later that year.
The second prototype computer (XD-2) was retained by IBM to
provide a test facility for modified hardware and to support
programming systems development.40

The SAGE System

'The first large-scale computer for the nation's vast, new elec­
tronic air warning network is being shipped from here to
McGuire Air Force Base in New Jersey," the IBM Militar y Products
Division's plant in Kingston, New York, proudly announced in
June 1956.41 An acronym for Semi-Automatic Ground Environ­
ment, SAGE had been adopted in 1954 as the name for the
air-defense system. The name was chosen to signify the semiauto­
matic manner in which the system would provide the information
environment for operation of the Air Defense Command.42 Ob­
serving that the computer was designed by IBM in cooperation
with MIT and that the total "SAGE system was developed by
scientists and engineers at MIT's lincoln Laborator y under con­
tract with the Air Force," the IBM press release continued:

This extraordinary "electronic brain" will become the first of the giant
computers to fit into the integrated complex of radar, ships, jet aircraft,
communications networks, missiles, and people that is rapidly taking
shape as the supersensitive continental air defense system. T his im­
mense project is known as the Semi-Automatic Ground Environment
(SAGE) system .... With a knowledge of flight plans of friendly planes
available in the computer, hostile planes can be identified immediately
and the most effective defense action taken.
T he SAGE system starts with a radar ring-on land, on Navy picket
ships at sea, on offshore Texas Towers, and on airborne early warning
planes ranging far out over the ocean. These radars are linked by
telephone lines or ultrahigh-frequency radio directly to the high-speed
computer. Information about aircraft anywhere within the radar area is
relayed continuously and automatically to the computer. T his IBM-built
equipment, called the AN/FSQ-7, digests all of this information plus
Ground Observer reports, flight plans, and weather information as fast
as it is received and translates it into an over-all picture of the air
situation. T hese TV-like pictures show the air battle as it develops and
provide the basis for the necessary human judgments.
The computer automatically calculates for the operator the most
effective employment of such defensive weapons as guided missiles,
Copyrighted Material
214 Building IBM

Figure 15.2. SAGE

Top: Schematic reveals the basic elements of the SAGE Air Defense System.
Bottom: Operator sits at a display console using input facilities, including
the light gun held in his right hand. (Courtesy of the MITRE Corporation
Archives.)
Copyrighted Material
 An Air Defense System 215

antiaircraft batteries, and jet interceptors. In the case of the intercepting

jets, the aircraft is controlled by directions fed by radio directly from
the computer to the automatic pilot in the plane. Missiles are controlled
similarly. At any time, the air battle commander can have the computer
display the over-all air situation or whatever part of it he wishes to
monitor in detail. As the battle moves, information is transferred spon­
taneously to an adjacent computer.43

Because the SAGE operations seem almost routine in today's

world, it is hard to imagine the impact this announcement had
in 1956. It was the first time a computer had been used to control
a large, geographically distributed system. The size and scope of
SAGE was almost unbelievable. When fully deployed in 1 963,
there were twenty-three direction centers distributed near the
northern boundary and the east and west coasts of the United
States. Each of these contained an AN/FSQ-7 computer system
having almost 60,000 vacuum tubes, weighing 250 tons, occupying
an acre of floor space, and using up to 3 million watts of electric
power.
Each AN/FSQ-7 computer was actually two identical computers
that operated in a "duplexed" mode for greater reliability. They
used binary arithmetic, single address instructions, and 32-bit
words, and could process approximately 75,000 instructions per
second. Each of the two magnetic core memories had a capacity
of 8,192 words and had a read-write cycle of 6 microseconds.
(These memories were replaced by ones with 69,632 words begin­
ning late in 1 957.) Additional high-speed storage was provided
by 12 magnetic drums each with a capacity of 12,288 words. Each
central processor was capable of handling the operation of 100
display consoles while sending and receiving data from twelve
remote sites.44
There were also three Air Defense Division headquarters,
called combat centers, where military officers had responsibility
for supervising the activities of several direction centers. Each of
these contained a modified SAGE computer, called AN/FSQ-8,
which had fewer display consoles and required far less input
processing equipment because it received data previously proc­
essed by the direction centers. Its task was to summarize and
display air-defense information and to provide alternative actions
for the officers in charge.45
The first production system was tested and accepted in June
1 956 and was declared fully operational within the SAGE system
at McGuire Air Force Rase in New Tersev: in July 1958. The last of
CopyngnfeaMt:ltenal/
216 Building IBM

these systems was shut down in January 1984, after defending the
United States from air attack for more than a quarter centur y.46

Programming

The development, testing, and maintenance of computer pro­

grams for SAGE was an enormous task. Primary responsibility for
this was given to the Rand Corporation of Santa Monica, Califor­
nia , in par t because neither IBM, MIT, nor the Bell Telephone
Laboratories could envision using so many programmers once
SAGE was completed. Rand was also a ver y logical choice. Estab­
lished as a nonprofit corporation in 1948, its origins date back to
1945 when the air force created Project Rand to employ top-qual­
ity civilian scientists and advisers to conduct defense research.
A particularly significan t project, initiated by Rand in 1950, was
the study of group behavior in man-machine systems.47 For this
s tudy an existing air-defense radar station was replicated and
pioneering work in simulating bat tle conditions was undertaken.
Several IBM CPC electronic computers were used to perform the
millions of computations needed to create the simulated radar
scans presented to operators during these studies. The realism
achieved was found to improve dramatically the training of per­
sonnel within the air-defense system. Even greater realism was
achieved when an IBM 701 became available to replace the far
less powerful CPCs. In October 1954 the air force awarded Rand
a contrac t to develop, install, and maintain system training facili­
ties in 1 50 sites nationwide. These activities were excellent prepa­
ration for the organization tha t was to program the computers of
the new air-defense system.
Rand's programming activities for SAG E began in July 1955
when five of i ts programmers arrived at Lincoln Laboratory to
work in a "totally integrated, joint organization" with MIT person­
nel who had already wri t ten prototype programs for Whirlwind.
Near the end of the year Rand had grown its SAGE programming
and system training activities in to an independen t System Devel­
opment Division with 450 employees, 75 of whom were at the
Lincoln Laboratory. By the end of the next year t he division had
more than doubled in size and was spun off as an independent
company, t he System Development Corporation.4H
The programming of SAGE computers proved to be an even
grealer task lhan an ticipated. This is not surprising. The pro­
grams were, in aggregate,. ver y large,. and they controlled t he
Copynghted Malenal
 An Air Defense System 217

world's first on-line, real-time, geographically distributed system.

At the time, however, Bob Everett recalled, "we believed our own
myths about software-that one can do anything with software on
a general-pwpose computer; that software is easy to write, test,
and maintain; that it is easily replicated, doesn't wear out, and is
not subject to transient errors. We had a lot to learn. "49
Respect for the size of the programming job was nonetheless
evident in a bold decision to build programming tools before
starting development of the operational air-defense program it­
self. A set of programs named the Lincoln Utility System resulted
and served to mechanize program development and testing and
to impose systematic and efficient standards on those tasks. Con­
sisting of some 40,000 instructions, the system ensured that the
lOO,OOO-instruction SAGE operational program was developed in
a disciplined manner that facilitated operation, maintenance, and
extension. Al so crucial to success was the prototype air-defense
program already operational on the Whirlwind computer. The
SAGE program was, nevertheless, delivered a year later than
scheduled and required a far larger memory than initially
planned.
"The biggest mistake we made," according to one who worked
on both the Whirlwind and SAGE programs, "was that we at­
tempted to make too large a jump from the 35,000 instructions
we had operating on the much simpler Whirlwind I computer to
the more than 100,000 instructions on the much more powerful
IBM SAGE computer." Eventually close to 500,000 computer in­
structions were written, of which only about 25 percent actually
supported operational air-defense missions. The rest were needed
to test systems, help generate programs, document the process ,
and provide various prqject management functions. 50

The overall effectiveness of SAGE is hard to assess because it was

never used in real combat. Some might argue, however, that that
is the true measure of its success. It was a technological marvel,
a deterrent to potential aggressors, and a means for simulating
attacks and training U.S. military officers and personnel. Sitting
in the command-and-control room, armed only with light guns
( now called light p ens ) , military officers and personnel could
command the computer to provide information on simulated
attacks and calculateca}Jfn Hfl/d YFMNi�m,ive actions. The system
218 Building IBM

also served as a laboratory for evaluating new designs for com­

mand and control systems.51
But the importance of SAGE goes far beyond its impact on the
cold war. Its pioneering technological contributions were numer­
ous. Among these are magnetic-core memories, a large real-time
operating system, highly disciplined program structure and de­
velopment, overlapping of computation and I/O functions
through a rudimentary I/O channel, transmission of digital data
via telephone lines, cathode-ray-tube displays with light pens,
worst-case designs and marginal testing of circuits and compo­
nents for high reliability, and duplexed computer operation for
enhanced reliability. The system's overall reliability, a crucial de­
sign objective, is indicated by two-year performance data com­
piled beginning in 1978 when only seven centers were still in
operation. Both machines in any one center were typically down
simultaneously less than 24 hours per year, of which 20 hours
were attributed to failures in air conditioning and other noncom­
puter causes. 52
Without doubt the most significant technological achievement
of Project SAGE was design of the first geographically distributed,
on-line, real-time system implemented with digital computers.
Inspired by Perry Crawford's master's degree thesis, by the vast
number of electronic vacuum tube circuits successfully operated
in ENIAC, and by von Neumann's computer design concepts,Jay
Forrester moved quickly to use digital computers for system con­
trol functions. Development of this hitherto nonexistent field was
accelerated by the confluence of several fortuitous events. Par­
ticularly significant among these events was the 1 949 explosion
of the Soviet Union's first atomic bomb. This created an environ­
ment in which Forrester's project received almost unlimited tech­
nical and financial resources.
There are many examples of Forrester's engineering leader­
ship, but none is more striking than his persistent and successful
effort to develop a reliable and cost-effective computer memory.
Others had pioneered the use of magnetic cores, but they lacked
the insight, clarity of purpose, and challenging requirements that
helped Forrester create the first computer main memory imple­
mented with ferrite cores.
Numerous companies were involved in the development of
SAGE. Significant among these were the Western Electric Com­
pany, with responsibility for overall engineering management and
py;.
construction of the S1!J tgh�ff' M!Jte�N1tems Development Cor-
 An Air Defense System 21 9

pora tion (SDC ) , with responsibility for programming the SAGE

computers; the General Electric Company, with responsibility for
developing a variety of radar systems and data links ; and the IBM
Corporation, with responsibility for the design , manufacture , test­
ing , ins tallation , and maintenance of the SAGE compu ters .
Each of these companies profited from its involvement, but
with the possible exception of SDC whose very exis tence resulted
from SAGE, none profited more than I BM. Although SAGE pro­
vided less than 4 percent of I BM's total U.S. revenue from 1952
through 1955, it contributed almost 80 percent of the company's
revenue from stored-program computer systems . (See appendix
D . ) This was a critical period during which the company wrested
leadership from Remington Rand in the new market for large­
scale electronic stored-program computers. At the peak of its
activities in 1 95 7 and 1958, IBM employed over seven thousand
people to manufacture , install, service , and improve the SAGE
systems.53
The skills of its engineers were strengthened and refined
through close collaboration with engineers at the MIT Lincoln
Laboratory. Its manufacturing organization improved its ability to
mass-produce and test large computers, and its service organiza­
tion gained similar advan tages. The benefits of this experience
showed up rapidly in the IBM 700 series computers that followed
immediately after the 70 1 and 702 . The company also made use
of the technical skills of people trained in SAGE to develop, in
cooperation with American Airlines, the first commercial airline
reservation system , SABRE.

Copyrighted Material
16
Chasing NeuJ Technologies

Forrester's emphasis on developing reliable, high-speed memo­

ries for SAGE is hardly surprising. A means of storing information
for automatic manipulation and analysis by machine is basic to
any information processing system. Punched cards introduced by
Hollerith in the late 1880s served this function adequately for
more than half a century because the time required for arithme­
tic computations remained in reasonable balance with the time
required to handle punched cards.
This relationship was put severely out of balance by the intro­
duction of digital electronics in the 1940s. Electronic circuits
could perform arithmetic computations dramatically faster than
the data could be read from ·cards. Finding an adequate solu­
tion for the speed imbalance between logic circuits and memory
was the primary technological problem facing designers of early
electronic computers. Indeed it never ceased being a dominant
problem.
The designers of the IBM CPC solved this problem reasonably
well by providing a small amount of auxiliary electromechanical
memory that fell between vacuum tube and punched-card storage
in speed and cost, and by providing the ability to perform many
arithmetic operations electronically between the reading of data
and instructions from each card. The much larger Remington
Rand UNIVAC (operational about a year and a half after the
CPC) provided a better solution. Data and instructions were
stored on magnetic tape rather than on cards, and they were read
into a high-speed mercury delay-line memory before being exe­
cuted by electronic circuits. This stored-program approach, with
some important differences, was subsequently used in the IBM
701 as well as in later computers offered by IBM, Remington
Rand, and others.
Copyrighted Material
222 Building IBM

Among the mor e significant differences between the IBM 701

and UNIVAC was the 701's use of ligh tweight Mylar tape, coated
with m agnetic material, for information storage. Eckert and
M auchly had chosen heavy metallic tape bec ause ligh tweight
tapes broke too fre quently during stop-start operation . The break­
age pr oblem was overcome by IBM engineers who devised vac­
uum columns to create long loops of tape at either side of the
read-write head; the resulting slack in the tape dramaticall y re­
duced the stresses on the tape. This basic vacuum-column design
was adopted industrywide and remained the preferred design for
decades.
The IBM 70 1 also featured punched-card input-output capabil­
ity in addition to m agnetic tape storage, and it offered a high­
speed magnetic drum to supplement the main memory. The
cathode-ray-tube memor y used in the 701 was faster but less
reliable than UNIVAC's mercury delay-line memory. A truly sat­
isfactor y main m emor y for electronic computers was first pro­
vided by the ferrite-core devices pioneered in Pr�ject SAGE.
The rapidity with which these and many other technologies
were developed and used in products following World War II is
remarkable. Even more remarkable is the speed with wh ich yet
newer technologies threatened their supremacy. Most important
among these was the semiconductor transistor. Invented at the
Bell Telephone Laboratories at the end of 1947 and publicly
announced the f ollowing July, the transistor was devised spe­
cifically to replace vacuum tubes in electronic circuits. Next in
im por tance for the future of computers was magnetic disk stor­
age. Invented, implemented, and rapidly improved by IBM, disk
storage was destined to take over most functions of punched cards
and magn etic tape in computer systems.
Growth of the computer industry was driven for decades by
parallel imp rovements in magnetic disk storage and semiconduc­
tor device technologies. Leadership in the pro duct application of
these and other rapidly evolving computer tech nologie s was cru­
cial to IBM's continu ing success and is the suqject of this chap ter.

A New Laboratory

Creation of a new laborator y devoted to advanced technologies

was motivated in part by "ins i sten t demands of the Sales Depart­
ment" that had forced postponement of work on nearly all ad­
vanced com p onents ciJp�tefi�/laboratory. 'The only
 Chasing New Technologies 223

solution," Wally McDowell had advised John McPherson in 1950,

"is the establishment of a separate group or groups whose sole
function is to work on long range component development."l
Another motivating factor was the reluctance of many qualified
engineers, especially those from California, to relocate to
Poughkeepsie or Endicott, New York. This impediment came
forcibly to McDowell's attention early in 195 1 as he was attempt­
ing to double the engineering personnel in Poughkeepsie to 750
by the end of the year.2
Out of these considerations a plan evolved to establish a new
laboratory in San Jose, California, where the company had oper­
ated a card manufacturing plant since the early 1940s. Chosen to
head the laboratory was Reynold Johnson, the man Watson, Sr.,
had hired into the company in 1934 to develop the first automatic
test scoring machine (see chap. 6) . During the ensuing years,
Johnson had demonstrated his technical creativity and manage­
ment skills on many other projects as well.
Early in 1952 Johnson and four others took occupancy of a
small rented building at 99 Notre Dame Avenue and began the
task of transforming it into a research laboratory.3 To their sur­
prise and pleasure, a single advertisement placed in newspapers
in the West produced several hundred applicants. The applicants
were interviewed and many qualified engineers hired. By June
the new laboratory was reported to have a "smoothly functioning
organization" with approximately thirty employees.
Wishing to avoid the negative effects of the competition and
secrecy that had existed in the Endicott laboratory, Johnson in­
sisted that each engineer be conversant with all projects in the
laboratory. To emphasize his desire for a cooperative environ­
ment, he established the following priorities for each engineer:
"It is your most important assignment in this laboratory to give
assistance when you are asked to do so, by any other engineer of
this staff, in the form of consultation, experimentation, or sug­
gestions, and the second most important assignment is that of
carrying on the project to which you are assigned. "4
Many projects had been suggested for the laboratory even
before it was formally established . Others were identified by newly
hired staff members. Several of the more promising projects were
launched. Some of these were terminated following unsuccessful
initial experiments, and a couple were transferred to Endicott for
possible product development. But it took only one prqject to
make the entire acti�p}yli!f�te1i3Jlt project produced the
224 Building IBM

first magnetic disk storage unit, and it provided the primary

mission of the San Jose laboratory for years to come. It was
formally begun in September 1952 under Arthur J. Critchlow, a
1947 graduate of the California Institute of Technology. Accord­
ing to Critchlow, his initial assignment was to study the manner
in which information was organized, formatted, stored, and pro­
cessed using punched-card equipment-and to seek a better
method.5

Inventing Disk Storage

Urged by Johnson to learn as much as he could about customer

needs, Critchlow turned for help to a special representative in
IBM's branch office in San Francisco. Pleased to participate, the
representative gave private tutorials and arranged for visits to
customer installations.6 Critchlow learned that punched-card
equipment performed customer tasks especially well when proc­
essing could be done in batches on sequentially sorted informa­
tion. Serious problems arose when more nearly random access
was needed.
Inventory control was a particularly important application of
this latter type. In warehouse operations, for example, each order
typically required several cards to be manually located, removed
from a stack of cards, the inventory information updated, and
the updated cards returned to their original locations. To facili­
tate this clerical activity, drawers of cards were set out on work
tables so that several persons could access cards from the same
file. This arrangement was called a "tub file. "
Replacement of the tedious and error-prone work associated
with tub files soon became the project's objective. Initially con­
sidered was every storage configuration "that it was possible for
an ingenious staff to think up," Johnson asserts. These included
magnetic cards, plates, wires, bands, tapes, twirling rods, and
rotating drums and disks, as well as several optical storage con­
cepts.7 By April 1953 all configurations had been eliminated from
consideration except for continuously rotating magnetic disks,
spaced along a straight shaft. Proposed by a newly hired engineer
only three months earlier, this proposal may seem obvious today,
but a related prior proposal lacked its crucial simplicity.8
Published in the August 1952 issue of an engineering journal
and brought to the attention of Critchlow in December, the
Copyrighted Material
 Chasing New Technologies 225

earlier proposal had been conceived by a researcher at the Na­

tional Bureau of Standards. Known as the "notched-disk mem­
ory," it stored digital information on magnetically coated disks
with pie-shaped notches cut out of them. The disks were mounted
on a ring-shaped (rather than straight) axle with their notches
normally aligned toward the center of the ring. Read-write heads,
mounted on an arm that pivoted about an axis at the center of
the ring, could be rotated through the notches of the disks to any
desired disk. Information was then written on, or read from, the
selected disk by rotating it one full revolution past the read-write
heads. Disks not selected for the read-write operation remained
stationary.9 Despite this peculiar arrangement (by today's per­
spective) , the article did help stimulate the IBM engineers to
conceive their own successful design.
The new IBM storage unit was publicly demonstrated in May
1955. A press release said in part: 'The new electronic device will
permit mechanization of accountin� and record-keeping pre­
viously found impractical owing to costs or procedural problems.
Sorting-one of the most costly and time-consuming office ma­
chine processes-will be greatly reduced and in some cases actu­
ally eliminated. "10
More technical information was presented at the Western Joint
Computer Conference early in 1956: "The information is stored,
magnetically, on 50 rotating disks. These disks are mounted, so
as to rotate about a vertical axis, with spacing between disks of
0.3 inch. This spacing permits magnetic heads to be positioned
to any of the 100 concentric tracks which are available on each
side of each disk. Each of these tracks contains 500 alphanumeric
characters. Thus, the total storage capacity is 5,000,000 charac­
ters. "ll This seemed like an enormous storage capacity at the
time, but by the early 1990s two tiny silicon chips could provide
as much.12
The difficulty of rapidly accelerating and decelerating selected
disks, as necessary for the Bureau of Standards design, was
avoided by keeping all disks spinning continuously at 1,200 revo­
lutions per minute. Access to both sides of a selected disk was
achieved by two read-write heads that straddled the disk. Mounted
on a single mechanism, the heads could be withdrawn and rein­
serted elsewhere in the stack of disks in less than one second.
Possible damage to the recording surfaces was avoided by forcing
compressed air through orifices in the read-write heads to create
Copyrighted Material
226 Building IBM

a cushion of air between them and the disk surfaces. The resul­
tant head-to-surface spacing was about one-thousandth of an
inch.
Known as the IBM 350 Disk Storage Unit, its first use was on
the small IBM 305 RAMAC (Random Access Memory Accounting
Machine) . Fourteen of these were available for field tests begin­
ning in June 1956, and the production version was announced
that September. The monthly rental for a basic RAMAC was
$3,200, of which $650 was for the disk storage unit, $1,625 for the
processing unit and power supply, and $925 for the the console,
printer, and card punch. More than a thousand of these vacuum­
tube-based computers were built before production ended in
1961.13
Important to the success of RAMAC was a novel method for
finding stored information when its physical location on the disks
was unknown. This first known use of "randomization and chain­
ing" for storing and retrieving records was devised by H. Peter
Luhn, who was already an established inventor when hired by
Watson, Sr., in 194 1. Luhn's application for a patent on the
method was rejected by the company's patent review board in
Poughkeepsie in 1956 because the concepts were not embodied
in hardware. The company had made a policy decision that
computer programs and procedures were not patentable.14
The most significant competition to RAMAC was provided by
the Univac File Computer, which was also first delivered in 1956.
The Univac system was equipped with magnetic drum storage that
utilized one read-write head per track of information. Thus it
could provide far faster access to information than RAMAC,
which had only two read-write heads (on one access mechanism)
for the entire storage unit. Because information was stored on
the cylindrical surfaces of the drums, however, its volumetric
storage efficiency was far poorer than for disks. Even when
equipped with its maximum of ten drums, the Univac File Com­
puter offered only one-third the storage capacity of RAMAC with
its single storage unit. 15 The cost-performance superiority of disks
over drums became increasingly evident during the ensuing
years, making drum storage obsolete in little over a decade.
Designed as an extension of IBM's punched-card equipment
and first offered on the small RAMAC, the IBM 350 Disk Storage
Device was the first of a long series of disk storage devices that
helped to build the computer industry. By driving the cost of
information storage � fIf�&�i�tk9-mfr with each succeeding
 Chasing New Technologies 227

Figure 16.1. RAMAC and Its Development Leaders

Top: Pictured (from left) are William A. Goddard, who led the development
of the disk storage system beginning in April 1953; Reynold B. Johnson, labo­
ratory manager; Louis B. Stevens, who took over development responsibility
in November 1953 and put RAMAC into production; Arthur J. Critchlow,
who did early studies leading to the RAMAC development; and John W.
Haanstra, who had initial responsibility for magnetic recording and electron­
ics. Bottom left: T he RAMAC disk storage system, revealing the read-write ac­
cess mechanism. Bottom right: RAMAC systems undergoing final tests in the
new San Jose manufacturing plant in 1957. (Courtesy of the American Soci­
ety of Mechanical Engineers.)

Copyrighted Material
228 Building IBM

disk product, the San Jose engineers made major contributions

to IBM's success and to the growth of the industr y. Among their
many improvements to magnetic disk storage, the following are
particularly significant: a dedicated read-write head for each disk
surface and self-acting air bearings in 1961, removable disk packs
in 1962, voice-coil actuators with servoing to selected tracks in
1970, flexible (floppy) disks in 1971 , the "Winchester file" with
its lightweight heads in 1973, and thin-film heads in 1979.16
In 1 984 the IBM 350 Disk Storage Device was designated "an
International Historic Mechanical Engineering Landmark by the
American Society of Mechanical Engineers. As reported in their
brochure, "It is difficult to overstate the impact the 35O's disk
technology has had upon the world in the years since its an­
nouncement. ... it is sufficient to say that fixed and flexible disk
drives alone-all derivatives of the basic 350 technology-gener­
ated an estimated $12.5 billion in sales worldwide in 1983 for the
72 manufacturers of fixed disk drives and the 52 manufacturers
of flexible disk drives."17

The Solid-State aallenge

"A device, called a transistor, .which has several applications in

radio where a vacuum tube ordinarily is employed, was demon­
strated for the first time yesterday at Bell Telephone Laboratories
... where it was invented." Thus began a brief story in the 1 July
1948 issue of the New York Times.1s T he invention had been
achieved the previous December by John Bardeen and Walter H.
Brattain, following years of work with William Shockley and oth­
ers at the Bell Telephone Laboratories. In 1956 Bardeen, Brattain,
and Shockley were awarded the Nobel Prize for their contribu­
tions. Except for the brief story in the New York Times, the 1 948
announcement of the invention was largely ignored by the news
media. It was not ignored by scientists working in solid-state
physics, however.They avidly read the one-and-one-half-page ar­
ticle published in the July 1948 issue of the Physical Review.19
Halsey Dickinson, who had succeeded James Bryce as head of
IBM's Patent Development Department at World Headquarters,
was also keenly interested in the invention.Although he had no
prior experience in solid-state theory or devices, he obtained
sample transistors from Bell Laboratories within weeks after the
announcement. Soon he and others in his department were
making their own tr�JfflfntJw ��FAtjng additional electrical
 Chasing NI'W Technologies 229

contacts to commercially available germanium diodes. They de­

signed and patented many transistor circuits, but most of their
designs incorporated vacuum tubes as well, to overcome the
shortcomings of early transistors.2o
By the spring of 1950 Tom Watson's newly appointed director
of engineering, Wally McDowell, had concluded that the methods
of IBM's old-time inventors would not suffice. Believing that
modern research methods were needed, he placed all work on
solid-state devices under Ralph Palmer in the Poughkeepsie labo­
ratory. Palmer's job was to build the company's expertise in
transistors and other solid-state technologies as rapidly as possible
while continuing to accelerate the pace of development of new
products using vacuum tubes.21
Palmer initiated classes to train engineers and hired people
with advanced degrees or practical experience in relevant tech­
nologies. The world-renowned solid-state physicist, Leon Bril­
louin, was among those hired to teach courses to the
Poughkeepsie engineers. 22 Most courses, however, were taught by
people already in the laboratory. Eager to increase their knowl­
edge, engineers spent long hours in the laboratory doing their
normal tasks while also attending classes several hours each day.
Homework was done at night.
A particularly significant course was conducted in the summer
of 1953. Twenty-eight engineers participated. Mter spending
three weeks studying theory, the engineers were divided into six
groups for six weeks of laboratory work. Each group was asked to
design a practical machine with transistor circuits. One of the
groups redesigned the Type 604 Electronic Calculating Punch,
using transistor circuits to replace the vacuum tube circuits.23 So
promising was their design that a transistorized 604 was built and
demonstrated in October 1954 at the opening of the company's
new Research Laboratory in Poughkeepsie. Some 2,200 transis­
tors had been used to do the job of 1,250 vacuum tubes, but the
resulting machine occupied less than half the volume and used
only 5 percent as much power.24
The transistorized 604 became the basis of the IBM 608 Tran­
sistor Calculator. First shipped in December 1957, the 608 was the
first all solid-state computing machine to be manufactured for
the commercial market. The phrase "all solid-state" indicates the
machine contained no vacuum tubes or electromechanical de­
vices in its logic, memory, and power-supply circuitry. Not only
were its logic circui ���twMb transistors, but so were
230 Building IBM

the support circuits for its small ferrite-core memory. Its advanced
circuit technology stood in sharp contrast to its architecture,
which was more akin to the EAM equipment it was replacing than
to a stored-program computer.25
Long before the 608 was completed, Tom Watson became
concerned that the use of transistors in company products was
progressing too slowly. If each engineering manager was permit­
ted to make his own technology decisions, the more familiar­
and still cheaper and more reliable-vacuum tubes would likely
remain the technology of choice for many years. Improved reli­
ability and lower costs for transistors could be achieved only
through higher manufacturing volumes, which in turn could be
achieved only if product development programs were committed
to their use.
To force this to happen, Watson took dramatic action two
months before shipment of the 608 began. At his behest a corpo­
rate policy was established with the following key provisions: "It
shall be the policy of IBM to use solid-state circuitry in all ma­
chine developments. Furthermore, no new commercial machines
or devices shall be announced which make primary use of tube
circuitry. "26 The policy was vigorously enforced and achieved the
desired results.

Competition in Supercomputers

The engineering resources of the IBM Poughkeepsie laboratory

were taxed to their limit during the early 1950s. In rapid succes­
sion there had been frantic efforts to gain expertise in vacuum
tube circuits, magnetic tape storage, magnetic cores, and transis­
tors. The 700 series of large computers was being developed, and
a massive effort had been initiated on Project SAGE.
In spite of these efforts, Ralph Palmer believed his overworked
laboratory was at risk of losing a desperate race with engineers in
other laboratories. He was only beginning to have the capabilities
to push the state of the art in a few technologies. What he wanted
was a project with demanding objectives like SAGE, but dedicated
to solid-state rather than to vacuum tube electronics. The right
project would provide a desirable focus for his research activities
as well as essential funding.
Tom Watson was also not satisfied, but for different reasons.
He continued to be rankled by Remington Rand's public image.
Its UNIVAC had bec.!)me .so I- clps�lv identified with large-scale
c;opyngn,ea Matenal
 Chasing New Technologies 231

electronic computers that the ver y name, UNNAC, was often

used synonymously for "electronic computer." The public image
of Remington Rand as the leading computer company did not
change in December 1954 when IBM announced completion of
NORC, the most powerful supercomputer of the time. Nor did it
change one year later when the number of large-scale computers
installed by IBM exceeded the number installed by Remington
Rand.
Watson's frustrations were shared by others, among them Steve
Dunwell. Returning to IBM from the Signal Corps in mid-1946,
he had been influential in the specification and design of many
of the company's electronic products, including the CPC, 650,
and 702 computers. During part of 1953 and 1954 he had been
on special assignment in Washington, D.C., to assess the com­
puter requirements of federal agencies, among them the Federal
Bureau of Investigation, National Security Agency, Social Security
Administration Treasur y Department, and Weather Bureau.27
,

Teaming up with Ralph Palmer in the summer of 1954, Dunwell

was enthusiastic about designing a supercomputer for govern­
ment applications that would also provide Palmer with the focus
he wanted for advancing the art in solid-state technologies.
The problem of funding the effort appeared to have a solution
in early 1955. The University of California Radiation Laboratory
(operated in Livermore, California, for the U.S. Atomic Energy
Commission) expressed interest in obtaining a "superspeed com­
puter. " The task of seeking the contract fell on Cuthbert Hurd,
who had replaced Yin Learson as director of EDPM. Hurd's
counterpart in the negotiations was Edward Teller, the physicist
who later became known as the "father of the hydrogen bomb."
Both men were well supported by technical experts and admin­
istrative personnel.
When the two groups met in late January 1 955, Teller let it be
known that he was serious and in a hurry All basic decisions
.

about the computer's specifications were to be made and distrib­

uted by March. Bids would be due in April, and the winning
contractor selected in May. A fixed-price contract in the vicinity
of $2.5 million was contemplated for the proposed computer,
called LARC, an acronym for Livermore Automatic Research
Computer 28
.

The schedule and conditions proposed for LARC precluded

the development of significantly improved solid-state devices.
Thus what had app��g MDtemportunity to achieve two
232 Building IBM

objectives with one contract suddenly became a divisive issue.

Watson, Learson, and Hurd were eager to proceed, whereas Pal­
mer and McDowell were reluctant to commit their limited engi­
neering resources to a project that would not allow time to
develop substantially improved device technologies.
A compromise was finally reached that favored the view of
Palmer and McDowell. The proposal Hurd submitted in April
offered to build and deliver a supercomputer at the desired price
in forty-two months. More significantly, the proposal urged that
negotiations then be undertaken to modify the contract to sub­
stitute a far more powerful computer, on the same schedule, but
at a higher price. Not surprisingly, the bid was rejected in favor
of Remington Rand's, which promised delivery of LARC in
twenty-nine months. Teller had let it be known from the begin*
ning that an early delivery date was of the highest priority.29
Loss of the contract was a major disappointment for Palmer.
Not only had he lost an opportunity for his laboratory to design
and build a supercomputer, but the winner of the contract was
IBM's archrival. Unless he could obtain an equivalent or better
contract quickly, Remington Rand's leadership in supercomput­
ers would be assured. This outcome would not be acceptable to
Tom Watson. Thus began several months of high stress for Palmer
as he frantically sought government funding.
Finally in August the National Security Agency (NSA) offered
to fund the development of two ferrite-core memories with speeds
three and twelve times faster, respectively, than the memory used
in SAGE. Funding was also offered for an exploratory study of
supercomputer designs to satisfy special requirements of the NSA.
"They now believe that we have something spectacular in core
memories and very bold and courageous thinking in machine
organizations," Palmer enthusiastically reported. He had not
been successful, however, in obtaining support for research on
transistors. The evidence he offered that IBM ''would have a large
commercial usage of transistors in the field before the telephone
company"-which had invented them-had failed to convince
the agency to bet its money on a "newcomer" to the field of
transistors.3o

Pniject Stretch

The $1.1 million offered by the NSA for eighteen months of

research was far too ImJ1)1fign�tW:ta,development of a super-
 Chasing New Technologies 233

computer, but it was an important endorsement of the the com­

pany's technical capabilities. It helped Palmer gain approval to
establish Project Stretch with considerable IBM funding in Au­
gust 1 955. The name of the project had been selected to connote
the intent of Palmer and McDowell to undertake research that
would "stretch" all aspects of computer technologies. In January
1956 Steve Dunwell was formally put in charge of the new
project.31
To obtain additional funding, the project quickly became in­
volved in proposing "the development and construction of a
high-speed general purpose computer for the United States
Atomic Energy Commission, for installation at the Los Alamos
Scientific Laboratory." 32 The proposal called for delivery of a
supercomputer for a fixed price of $4.3 million within forty-two
months after the contract was signed. Although the AEC ap­
peared reluctant to pay so much for the computer, the cost of its
development and construction was estimated by IBM to be over
three times higher. The higher costs would have to be recovered
from the sale of copies of the supercomputer, research contracts
with other government agencies, and use of Stretch-developed
technologies in commercial products. 33
Invitations to submit proposals had been sent by the AEC to
twenty firms, of which four were said to be interested. The great­
est concern for IBM was that Los Alamos would join its sister
laboratory in purchasing a rARC from Sperry Rand. (Remington
Rand had merged with Sperry to form the Sperry Rand Corpo­
ration in June 1955.) This anxiety vanished in April when IBM
representatives were advised informally of the acceptance of their
proposal. Formal contract signing did not occur until November,
giving Dunwell a much needed extra half a year to get started.34
Rapid staffing of the project with skilled technical people was a
major effort. The number of employees grew from fewer than
fifty in mid-1956 to two hundred one year later. It then remained
relatively constant until another one hundred were added in 1959
to begin machine construction. 35
Despite many problems, Project Stretch achieved its primary
objective of advancing computer technologies. Among its contri­
butions was a greatly improved ferrite-core memory with a capac­
ity of 131 ,072 bytes and a read-write cycle of 2.2 microseconds.
Equipped with transistor support circuits and other improve­
ments, it was nearly three times faster than the memories used in
SAGE, yet less expen_fdghtEgivVat�city.36
234 Building IBM

Another contribution was the "drift transistor," which was in­

trinsically faster than the "surface-barrier transistor" used in
tARC. By 1958 IBM's "newcomers" to the field of transistors had
placed their drift transistors in production and had made ar­
rangements for the Texas Instruments Company to produce ad­
ditional quantities. Also important was invention of the
current-switch, emitter-coupled logic (ECL) circuit by IBM engi­
neer Hannon S. Yourke in August 1956. Quickly adopted for use
in Stretch, ECL became the preferred circuit for high-speed
switching applications throughout the industry.37
The transistor circuits and the cards on which they were
mounted were part of a new Standard Modular System (SMS) of

Figure 16.2. Circuit Cards

The circuit cards shown here were developed as part of Project Stretch and
were used in the IBM 7000 series of computers. Six transistors (in cylindrical
metal cans) and assorted resistors and jumper wires were mounted on the
card (right). Electrical interconnections among these components were
achieved by soldering the component leads to copper lines preprinted on
the back of the card (left). The cards were known as SMS cards because they
were part of IBM's proprietar y Standard Modular System of circuit packag­
ing. The cards were 2.5 inches x 4.5 inches and were equipped with sixteen
gold-plated tabs for making external electrical contact.
Copyrighted Material
 Chasing New Technologies 235

circuit packaging developed in Project Stretch. Designed for auto­

mated mass production and low-cost maintenance, SMS was so
reliable and economical that it was used in all IBM electronic
products introduced during the early 1960s and continued to be
used in certain types of new products long after integrated circuits
were introduced.38
An unanticipated event caused Stretch-developed technologies
to be used first in a computer other than one for which they had
been developed. This shift in priority followed the October 1957
launching of the first manmade earth satellite, Sputnik-a feat
that demonstrated the Soviet Union's ability to deliver atomic
bombs to any target in the United States using rocket-powered
ballistic missiles. To defend against this threat, the U.S. Air Force
undertook development of the Ballistic Missile Early Warning
System (BMEWS) . Like SAGE, it would rely on a number of
interconnected computers to receive and analyze incoming infor­
mation and to determine the appropriate response.
Learning in the spring of 1958 that the air force had mandated
the use of transistorized computers in BMEWS, Ralph Palmer
responded rapidly. He arranged to propose a transistorized ver­
sion of the IBM 709 computer, which some of his engineers had
already designed in an experimental study. The acceptance of
IBM's proposal in October was heavily influenced by two factors:
first, the components and system design were available; second,
programs could be written for BMEWS and tested on a 709
computer before the first transistorized version was available.
Well before the contract was signed, Learson decided to an­
nounce the same transistorized 709 computer in the commercial
market. Christened the IBM 7090 Data Processing System, its
compatibility with the 709 facilitated its rapid acceptance in the
commercial market, as did a performance six times higher and a
price only one-third higher than the 709. The first 7090 was
delivered in November 1959 to Sylvania Electric Products, Inc.,
at Needham, Massachusetts, for incorporation in BMEWS. It was
the first of the highly successful 7000 series of computers.39
The Stretch supercomputer (named after Project Stretch) was
not delivered until seventeen months later in April 1961. It was
fully operational the following month-just one year later than
originally specified in the contract. Only one unscheduled main­
tenance period (which lasted four minutes) was required during
the three-day, forty-hour acceptance test.4 0 Subsequently desig­
nated the IBM 7030c<S}lyri��mcft'en�rcomputer was approxi-
236 Building IBM

mately two hundred times faster than the IBM 701, forty times
faster than the IBM 709, and seven times faster than the IBM
7090. It was also substantially faster than LARC, which had been
delivered about a year earlier.41
Although the 7030 reigned as the world's fastest computer for
over a year, its performance was only about half as good as
predicted. 42 Performance estimates had been based on the speed
of individual components and a system design that provided for
a high degree of concurrent operation among internal units.
There was considerable uncertainty, however, as to how much
concurrency might be achieved in actual operation. To obtain
this information, a pioneering simulator had been devised to run
on an IBM 704. The simulator proved to be helpful in choosing
among different system designs, but its ability to predict actual
performance on a customer's job was known to be limited.43 Even
today the prediction of system performance is not a precise
science.
The actual performance of the supercomputer was therefore
not known until it began to run customer programs, about two
months before its shipment to Los Alamos. 'With respect to raw
operation speeds . . . it appears that the machine executes accord­
ing to specifications," one observer reported. "However, when the
machine executes all of its instructions in complicated programs,
the actual efficiency of the system falls below what these raw
speeds might indicate."44
Tom Watson was embarrassed to announce the less than antici­
pated performance of the 7030 supercomputer during his key­
note address at the May 1961 Western Joint Computer
Conference. Worse yet, the performance projections had been
given to potential customers. Watson felt obliged to compensate
for the lower performance by reducing the computer's price from
$13.5 million to $7. 8 million. "If we get enough orders at this
price, we could go out of business," Watson wryly observed. Be­
cause IBM would lose money on each one sold at the lower price,
the 7030 was not offered beyond the eight customers with whom
negotiations were already underway.45
It was several years before Watson could overcome his preoc­
cupation with the failures of Stretch and thereby appreciate its
many technological contributions. In addition to advances in
solid-state technologies, Stretch pioneered many concepts that
influenced the design of later computers, especially the IBM
System/360, announ�jfJgAnRf1M�g these were multipro-
 Chasing New Technologies 237

gramming, memory protect, generalized interrupt, interleaving

of memories, lookahead, the memory bus, a standard interface
for input-output equipment, and the eight-bit character called the
byte.46 Although acknowledging the importance of these contri­
butions in a 1964 letter to Dunwell, Watson reiterated his distress
with Project Stretch, saying, "Our customers and the top manage­
ment of this corporation were led to believe that the performance
objective was being achieved when, in fact, we were falling short
of it. "47
Two years later IBM's difficulty in competing with large com­
puters offered by the Control Data Corporation caused Watson
to reconsider his position, which had discouraged engineers from
taking the necessary risks to compete in large computers. Seeking
to right a wrong and to bolster the morale of engineers "working
in the big machine area," Watson took the unusual step of devot­
ing most of his remarks at the 1 966 IBM Annual Awards Dinner
to one award winner. Applauding Steve Dunwell's appointment
as an IBM Fellow-the highest recognition for technical achieve­
ment in the company-Watson said Dunwell's work on Stretch
and many earlier machines had made him "a major contributor
to the success of this business." Then noting that "things" had
happened during the last few years "that haven't been as fair to
Red Dunwell as they might," Watson said, "I just thought I would
take the opportunity of publicly trying to correct the record. "48
It was vintage Watson. Quick to anger and quick to punish, he
was also prepared to acknowledge an error and take corrective
action.

Reorganizing Research

To carry out the company's rapidly growing research and devel­

opment activities, the number of employees so engaged had
grown from 600 to 3,000 (from 3 to 9 percent of IBM's domestic
employees) between 1950 and 1 954. Much of this growth had
been in Poughkeepsie, where a sevenfold increase from 200 to
well over 1,400 had occurred.49 Although this rapid growth had
been managed by McDowell under direct instructions from Wat­
son,Jr., the senior Watson had closely monitored and influenced
it. The confidence both Watsons had in McDowell had been
crucial to his success in making these dramatic changes.50
In recognition of McDowell's achievements as director of engi­
neering and the inc� MMWapf his assignment, Tom
238 Building IBM

Watson promoted him to IBM vice president in July 1 954. Re­

sponding to the younger Watson's challenge that he substantially
strengthen the company's research and development programs,
McDowell promoted four engineering managers to executive po­
sitions at World Headquarters to assist him in managing all cor­
porate technical activities. Chief among those promoted was
Ralph Palmer, who was named director of engineering.51 McDow­
ell also established a task force of about a dozen technical leaders
from the Endicott, Poughkeepsie, and Watson laboratories to
consider how research should be managed.
Meeting for several weeks at the Fox Hill estate in Connecticut
in the spring of 1955, the task force reviewed and compared the
research management methods of IBM with those of leading
companies such as AT&T, GE, RCA, and Westinghouse. Following
its deliberations, the task force recommended that a separate
research organization be created, that it report directly to corpo­
rate management, and that it be led by a distinguished research
executive hired from outside the company.52
Ralph Palmer was not sanguine about the proposal. He be­
lieved the Poughkeepsie laboratory's success had been achieved
in part through close interactions between research and develop­
ment groups. Development activities provided much needed fo­
cus for research, and good research provided a ready source of
new ideas and alternative solutions for development. But Palmer
was in the minority. Wallace Eckert, director of the Watson Labo­
ratory, strongly favored a separate research organization with
close ties to government and university laboratories. McDowell
also favored the recommendation. His experience in the Endicott
laboratory had caused him to conclude, some years earlier, that
research should be separated from development to protect
against near-term product pressures.
McDowell's acceptance of the recommendations was signaled
by the following announcement in the January 1956 Pough­
keepsie Engineering Newsletter: "IBM has established a new in­
dependent research organization to develop the business ma­
chines of the future. The new group will remain within the
company's research and engineering department, but will oper­
ate independently of its parent. It will be headed by Mr. R. L.
Palmer."53
Among the anticipated benefits of a separate research organi­
zation, with greater emphasis on basic research, was the ability to
attract outstanding cnll��.Jrr.arlu.a tes with advanced degrees in
Copyngflrea lVIarenal
 Chasing New Technologies 239

science and engineering. Typically these young people were im­

bued by their professors with the belief that product developmen t
was an inferior activity to basic research. As reported elsewhere
in the newsletter, ''The problem of recruiting top-caliber engi­
neers in sufficient quantity to meet the company's ever-increasing
needs is probably the most important single problem facing us
today." More than 1 ,300 additional engineers were to be hired
during the year.54
What the newsletter failed to reveal was that Ralph Palmer's
assignment as head of Research was only temporary. His primary
task was to separate research activities in Poughkeepsie from
development, thus establishing an interim research organization
to help attract a distinguished research executive from outside
the company. In September McDowell announced that Emanuel
R. Piore had accepted the position of IBM director of research.
Palmer's previous position was now officially renamed "acting
director." Piore's responsibilities would encompass the research
activities Palmer had separated from development in Poughkeep­
sie; advanced development activities in San Jose that Reynold
Johnson had separated from the RAMAC development effort; the
Watson Laboratory in New York City; and a laboratory established
the previous year in Zurich, Switzerland.
An immigrant from Russia at age nine, "Mannie" Piore had
obtained a Ph.D. in physics at the University of Wisconsin in 1935.
During World War II he had served as a lieutenant commander
on the staff of the deputy chief of naval operations for air. Mter
the war he distinguished himself with the Office of Naval Re­
search (ONR), serving as its chief scientist from ] 951 to 1955.
Piore was particularly respected for his role in continuing ONR
support for research at universities after hostilities ceased, and
for shifting the emphasis from short-term military requirement..
to more basic research. This policy contributed to the preemi­
nence of U.S. universities in science and engineering and
strengthened the country's economy in the postwar era.55
More a research administrator than a scientist, Piore had a
good understanding of strategy and the politics of organizations.
At IBM he developed close working relationships with Tom Wat­
son and other key executives. From the Research organization,
he initially selected three bright young technical leaders to review
ongoing projects. Soon labeled "the three wise men" by their
colleagues, they were to help Piore select projects to be dropped
}n
or moved out of Rest"l:fp � cf«MW6JMjfy fields of research the
240 Building IBM

company should enter. Piore also initiated annual research plan­

ning conferences to which senior Research members and a few
technical leaders from the development laboratories were in­
vited.56
An early decision was to move Research from Poughkeepsie to
a place with easier access to major universities, cultural centers,
and travel facilities. The site Piore selected was Yorktown Heights
in Westchester County, approximately twenty-five miles from the
northern border of New York City. Here a modern research
facility, designed by Eero Saarinen, was built on a 240-acre site
overlooking rolling, wooded hills. In honor of Watson, Sr., it was
dedicated in April 1961 as the Thomas J. Watson Research Cen­
ter.57
For five years Piore was given a nearly free hand to set the
direction of Research. By emphasizing basic research, he attracted
an outstanding cadre of bright young scientists with doctor's
degrees from leading universities. The newly hired researchers
and some of the old-timers rapidly established a good reputation
for the laboratory in fields as diverse as solid-state lasers, language
translation, integer linear programming, and artificial intelli­
gence. In recognition of these successes, the Research organiza­
tion was elevated to the status of an IBM division in November
1963.58
The new organization's attempts to contribute to the product
line were less successful. Thin magnetic films, for example, were
vigorously pursued and promoted as replacements for ferrite
cores. Transferred to a development laboratory in 1962, magnetic
films were successfully used in only one IBM computer model
before they became obsolete, because of semiconductor memo­
ries. Even more disappointing was the outcome of work on
cryotrons. Research on these superconducting devices (which
operated at very low temperatures) was begun in the hope that
they would provide superior performance to semiconductors.
Only after the project was transferred to a development labora­
tory in 1962 were the real limitations of the technology under­
stood. Cryotrons had no chance cf replacing semiconductors in
any application.59
Exciting achievements of the late 1950s in language translation
and artificial intelligence were found to be illusory. More than
three decades of research-and dramatic improvements in hard­
ware technologies-were required before artificial intelligence
or language - translation �ystem� b�1lan to arrive in the market.
(jopyngnreu Marenal
 Chasing New Technologies 241

Figure 1 6.3. Piore and the Watson Research Center

Top: Emanuel R. Piore is shown in his office soon after being appointed
IBM director of research in 1 956. Bottom: The IBM Thomas ]. Watson Re­
search Center ( named after Watson, Sr. ) is shown here about the time it
was dedicated in April 1 96 1 . Designed by Eero Saarinen, it is l ocated in
Westchester County about forty miles north of the center of New York City.
Extensions to either end in 1 9 79 and 1984 increased the length of the front
of the building from 1 09 1 to 1,769 feet and its floor space fro m 459,000 to
,

757,000 square feet.

Copyrighted Material
242 Building IBM

Even more frustrating was the difficulty of achieving any prod­

uct advantage from outstanding accomplishments of IBM Re­
search in timely fields such as solid-state lasers and integer linear
programming.
By the end of the 1 960s it had become evident that a change
in the emphasis of the Research Division was needed. Contrary
to earlier expectations, product improvements were less likely to
result from radically new technologies (as had occurred after
World War II) than from evolutionary advances in existing tech­
nologies. To contribute effectively to this process, the Research
Division would have to allocate more of its resources to improving
existing technologies, and it needed closer coupling with the
product development laboratories. Creating this coupling and
nurturing means for transferring improved technologies from
research to development were challenging tasks for future direc­
tors of Research.60
The success of the Research Division should not, however, be
measured solely by its contributions to product technologies.
From its inception, the luster of its basic science helped promote
IBM products and made it easier to hire top-quality technical
people throughout the company. Also of importance was the
influence the Research director could exert on company-wide
product development plans. Piore was particularly adept at this.
With his carefully nurtured ties to top corporate executives and
easy access to the expertise of Research people, his opinions on
the technical adequacy of product development programs could
not be ignored.

Copyrighted Material
17

The year 1956 marked many significant events for IBM. Among
these, the first SAGE computer was installed, the first computer
with magnetic disk storage was introduced, the first all solid-state
computing machine was readied for production, and research was
separated from development to create an independent corpo­
ratewide research organization. An antitrust suit was settled with
the government, and the senior Watson yielded the post of chief
executive officer to his son.
For forty-two years Watson, Sr., had guided the growth of a
company and an industry whose origins lay in the inventive and
entrepreneurial genius of Herman Hollerith. By 1956 he had
achieved his final goal: IBM's future was secure in the hands of
his older son, and the company's wholly owned World Trade
Corporation was growing rapidly under the leadership of his
younger son. Proud of his accomplishments, the eighty-two-year­
old ThomasJ. Watson, Sr., died of a heart attack on 19 June 1956.1
During the three-quarters of a century since Hollerith first
sought solutions for handling census data, his inventions and
entrepreneurial activities had led to remarkable advances in stor­
ing and processing information. Performing mundane record
keeping as well as sophisticated engineering and scientific calcu­
lations, the progeny of his machines had markedly altered the
operations of governments, businesses, and academic institutions
throughout the world. The company he founded to tabulate data
from the census of 1 890 had evolved into the industry-dominating
IBM Corporation. This chapter examines IBM's competitive po­
sition and reviews some of the events and decisions that had
shaped the company Watson now entrusted to his son.

Copyrighted Material
244 Building IBM

Figure 17.1. Transfer of Authority

Thomas J. Watson. Jr., is congratulated by his father upon his promotion to
chief executive officer in May 1956.

Naming the Industry

Somewhat surprisingly a fully satisfactory name for his equipment

and the industry he created has never been found. Hollerith
referred to his equipment as tabulating machines, and he named
his business the Tabulating Machine Company.2 Others were
likely to speak of punched-card machines or Hollerith machines,
and the cards were widely known as Hollerith cards.
When IBM introduced its 80-column card with rectangular
holes in 1928, Watson insisted that they be called IBM cards. This
marketing maneuver was so successful that users of Powers equip­
ment might inadvertently refer to their cards with round holes as
Copyrighted Material
 Legacy 245

IBM cards. In the late 1930s, customers and then IBM itself began
referring to the equipment as EAM , the initials for the company's
Electric Accounting Machine Division. The term EAM helped
distinguish IBM equipment from the mechanical equipment of
the company's only significant competitor, Remington Rand. The
generic names used for the equipment were punched-card, tabulat­
ing, or accounting, even though its applications were already ex­
panding well beyond tabulating and accounting.
Rapid advances in electronics during World War II and the
need to perform vast computations led to the development of
electronic computing machines toward the end of the war. These
machines were called computers because they carried out tasks
previously handled by people with the job title of "computer."
Using pencil, paper, and mechanical calculating machines, these
human computers (mostly women) had carried out lengthy com­
putations that could now be assigned to electronic machines.
Watson, Sr., insisted that the word computer should not be used
to designate any IBM equipment. He believed that the word had
a job-threatening connotation and should be avoided. It was one
battle he finally lost. The name computer became the primary
designation for IBM's equipment and the industry. The names
that IBM initially employed, such as automatic sequence control­
led calculator, have long been forgotten. Even the carefully cho­
sen name EDPM (Electronic Data Processing Machine) survived
only as the formal designation for the IBM 701 computer and a
few subsequent offerings of the company. The phrase data pro­
cessing has, however, been widely used to connote the industry
and its products.
It is ironic that UNIVAC was officially called a computer even
though it was built and marketed primarily for business and
government data processing, whereas the IBM 701 was officially
called a data processing machine even though it was built and
marketed for scientific and engineering computations .3 It is also
ironic that computing has long been a minor application for
computers. Most computers are used to perform tasks involving
computations that are transparent to the user-one of the more
popular tasks being text processing.
General purpose computers are better described as informa­
tion processing machines. Fundamental to information process­
ing equipment is a means for storing information so it can be
processed with as little human intervention as possible. That
Copyrighted Material
246 Building IBM

function was originally handled by punched cards. By the end of

the 1970s it was handled primarily by magnetic disks and semi­
conductor memories.
The phrase information processing is widely used and describes
today's industry better than does the single word computer. Infor­
mation processing is also an appropriate name for the business
Hollerith established in 1889. Nevertheless, it seems likely that
the briefer appellation, computer industry, will retain its greater
popularity for the foreseeable future.

Some Early Business Practices

Although Hollerith should be remembered primarily as an inven­

tor and entrepreneur, he made many management decisions that
survived long after he died in 1929. Perhaps the most important
of these was to rent, rather than to sell, his equipment. The
renting of capital goods was not a new business concept, but it
was particularly appropriate in the embryonic information proc­
essing business. Most customers were unwilling to install Hol­
lerith's novel equipment until they were shown how to make
effective use of it, and they needed help in adapting their business
procedures to it.
These services were provided at no extra charge by Hollerith,
who recovered his costs from the rental of installed machines and
the sale of punched cards. He also serviced his machines at no
extra charge. Indeed his contract with the Census Bureau re­
quired him to pay a substantial penalty if the equipment failed to
function properly. Partly because of this contract provision, but
primarily because of pride in his creation, Hollerith was meticu­
lous in specifying his equipment for manufacture to ensure its
reliable performance. He also responded rapidly to any reported
malfunctions.
Renting equipment and providing customer service at no extra
charge became a cornerstone of Watson's policies after he took
over the company in 1914. Many a story is told of IBM servicemen
working around the clock to get equipment running. If a fire,
flood, or other disaster occurred, IBM was ready to help. Using
equipment loaned by another customer, or available at a branch
office, or trucked in from a warehouse, the company's personnel
were expected to get the customer's work done on time. Reliable
service was probably IBM's most valued product, but we can only
Copyrighted Material
 Legacy 247

speculate whether customers would have been willing to pay for

it had they known its cost. Equipment was available only for rent,
and customers were required to have it serviced by IBM.
The rental-only policy had several other advantages for the
manufacturer. First, it assured a nearly undiminished revenue
stream in bad economic times when little new equipment was
being installed. Watson became acutely aware of this advantage
during the brief recession of 1922, and he benefited from it even
more during the Great Depression that began in 1929. Second,
it was more difficult for others to enter the market because of the
longer time required to recover the coste; of developing and
manufacturing equipment. Third, it was possible for the owner
to protect proprietary equipment designs through appropriately
devised rental agreements. It will be recalled that Hollerith re­
moved his equipment from the Census Bureau when he learned
that the bureau was using it for experimental studies to design its
own equipment. Fourth, rental customers were more likely to buy
supplies from the manufacturer.
To ensure all of these advantages for themselves, IBM and
Remington Rand entered into an agreement in 1931 that (among
other things) committed the two companies "to lease and not sell
tabulators and sorters" and "to require customers to purchase
their card requirements from the lessor or pay a higher price for
the rental of machines." Challenging these provisions in 1932,
the government filed a complaint charging both companies with
violating the Sherman and Clayton acts. The companies voluntar­
ily canceled these agreements in 1934 before the case came to
trial, but IBM continued to offer its equipment for rent only-a
policy the government later alleged contributed to its dominance
of the industry. The requirement that customers pay a higher
rental fee if they purchased cards from other suppliers was ad­
judged to be illegal in 1936.4
Although Watson followed many of Hollerith's business prac­
tices, his approach to managing the business was quite different.
Whereas Hollerith had difficulty attracting and managing top­
quality people, Watson was a master at it. Perhaps nowhere was
this difference between the two men more evident than in their
approaches to selling. Believing the equipment should sell itself,
Hollerith relied heavily on word-of-mouth recommendations of
satisfied customers and the efforts of independent sales agents.
Watson also believed in the importance of satisfied customers, but
Copyrighted Material
248 Building IBM

he was convinced that a well-managed and properly motivated

sales force could sell equipment more effectively and keep cus­
tomers better satisfied.
Borrowing from his experience at NCR, where he had worked
for John Patterson, "father of modern salesmanship," Watson
created the legendary IBM sales force. Two practices pioneered
by Patterson were fundamental. First, each salesman was given
exclusive rights to his territory. This increased his commitment
to customer satisfaction and to marketing efforts in his territory.
It also eliminated disputes among salesmen as to who earned a
commission. Second, each salesman was given a sales quota that
"had to be met." Based on the estimated number of potential
customers and previous sales records for the territory, quotas
provided, according to Patterson, "a definite measure of sales
efficiency and accomplishment." Quotas also provided a basis for
"friendly competition" among salesmen. Those who met their
quotas were rewarded by membership in the Hundred Per­
cent Club and an invitation to its emotionally charged annual
meeting.5
Watson additionally emphasized formal training for his sales­
men, who were expected to understand the operation and appli­
cations of each machine they sold. He supported them with
extensive advertising, and he motivated them with company songs
and a sense of being part of the "IBM family." Wearing the
mandatory blue suits, white shirts, and neckties, IBM salesmen
were taught to look, speak, and feel like winners. Under Watson's
tutelage, they became winners.
Salesmen worked closely with customers to ensure that installed
equipment performed its tasks as effectively as possible and to
identify new sales opportunities. They were instructed to report
back to headquarters any information about desirable improve­
ment'> or competitive innovations. In an important departure
from Hollerith's practice, Watson established engineering and
manufacturing organizations to respond rapidly to new market
requirements, to ensure the quality of products, and to protect
proprietar y information.
During the depression years of the 1930s, Watson maintained
full employment in his engineering and manufacturing organiza­
tions by increasing the time employees spent in educational pro­
grams and by increasing the IBM content in engineering tooling,
,

and manufacturing. Consolidating engineering and m anu factu r ­

ing ill Endicott, he used lhes� activities as a test-bed for the

c.;opynghteu Mstenst
 Legacy 249

application of punched-card equipment to the monitoring and

contfol of engineering and manufacturing operations. As a result,
the company's equipment and methods were in high demand to
help facilitate the country's rapid buildup in the manufacture of
war materiel for World War II. Further honing its own manufac­
turing capabilities through the production of various munitions,
IBM had established itself as a premier engineering and manu­
facturing organization by the end of the war. 6

Patent Policies

The shock troops of Watson's army were his salesmen. They

served in the front lines, representing IBM to its customers. They
were the first to learn of, and respond to, new competitive threats.
Watson was especially proud of their performance, but of all the
things that helped Watson dominate the industr y, none was more
important than his patent policies and associated research and
development activities.
Patents were, of course, the basis of Hollerith's early monopoly
of the information processing (punched-card equipment) busi­
ness. When Watson took over management of the business in
1914, Hollerith's earliest patents had expired, and actions of the
Census Bureau had further weakened his patent position. Never­
theless Hollerith still held many critically important patents. Wat­
son planned to use these patents and add to them to strengthen
the company's position. He was encouraged in this plan by a
consent decree entered into by the General Electric Company in
1912, settling an antitrust case concerning its dominance of the
incandescent lamp business. While ruling against mergers as a
legal means of gaining a near monopoly position, the court
affirmed the legality of competitive advantages based on patents.'
When representatives of the Powers Company requested a li­
cense to Hollerith's patents in 1914, Watson chose to accommo­
date them. He liked to represent this decision as magnanimous.8
But in truth he had good reason to believe that some competition
was necessary to avoid antitrust action by the government. Fur­
thermore the licensing agreement with the Powers Company
included a royalty payment of 25 percent of gross revenue from
equipment sales and rentals and 18 percent of the gross revenue
from the sale of cards. Because these royalty payments had to be
recovered by Powers in higher prices to customers, Watson en­
joyed an enormous ��riJIe could use the large
250 Building IBM

potential profit margin to pay stockholders larger dividends, or

to lower prices (and drive Powers out of business), or to invest in
his own business by purchasing patents, increasing research and
development, and financing equipment on rental . Watson chose
to invest in his business.
With the help ofJames Bryce, he built a strong patent portfolio.
He identified and purchased relevant patents. He hired inventors
and engineers to create new products and to file for patents.
During the brief recession of 1922 Watson took advantage of an
opportunity to buy the patents and research facilities, and to hire
key personnel, of his most innovative competitor, the J. R. Peirce
Company. He also purchased important patents of C. A. Tripp.
Then in 1930 he encouraged Bryce to establish the small Patent
Development Department to create new ideas and patents in
areas likely to be important in future products.
During the recession of 1922 Watson also reduced by half the
royalties charged to Powers to help that company out of the
receivership into which it had fallen. Once again he chose to help
his primary competitor survive rather than risk an antitrust suit.
Nevertheless he retained an important provision of the original
licensing agreement that diminished that company's prospects.
Patents were licensed only for use in mechanical systems. Electri­
cal detection of punched holes, which was basic to Hollerith
equipment, was reserved for IBM. The greater flexibility of elec­
trical detection provided a substantial advantage in scientific cal­
culations and in some business applications.

Dominating the Industry

So successful was Watson in dominating the information process­

ing industry, that the U.S. Department of Justice filed an antitrust
suit in January 1952 that asserted in part: "IBM now owns all
electrical tabulating machines in use in the United States, which
comprise 90 percent of all tabulating machines now in use in this
country. The remaining 10 percent of such tabulating machines
are mechanical machines manufactured by Remington Rand and
are either leased or owned by their users . ... More than 95
percent of the tabulating machines used by the United States
Government are owned by IBM. "9
Deeply distressed by the government's suit, the senior Watson
was also proud of the evidence it gave of his own success. When
fired from NCR by JohBpYftW7M8IMIDeH1J4, Watson had vowed to
 Legacy 251

build a larger and more successful company than Patterson.

When Watson left, NCR had approximately 90 percent of the
domestic cash register business. Now IBM was reported to have
90 percent of the domestic punched-card equipment business,
and its gross revenues and net earnings were substantially larger
than those of NCR.
The Department of Justice alleged that IBM was monopolizing
"interstate and foreign trade and commerce in the tabulating
industry, including new and used machines, machine parts and
service, cards and service bureaus." It further asserted that IBM
had "entered into contracts, agreements and understandings in
unreasonable restraint of the aforesaid interstate and foreign
trade and commerce in tabulating machines and tabulating
cards." These alleged acts violated Sections 1 and 2 of the Act of
Congress of July 1890, "An Act to Protect Trade and Commerce
against Unlawful Restraints and Monopolies," commonly called
the Sherman Act.IO
During the six decades since the Sherman Act was passed, its
primary function had shifted from protecting the public against
"flagrantly anti-competitive activities of the trusts" to an economic
policy tool for maintaining competition. A reading of past court
decisions revealed that a company was unlikely to prevail against
a government suit based on the Sherman Act unless the company
could demonstrate either that it did not possess monopoly power
or that its monopoly power had been achieved unintentionally­
for example, that it had "become a monopolist by force of
accident. "11
Because of the difficulty of convincing anyone that IBM had
achieved its market position "accidentally" rather than "intention­
ally," its first line of defense was to define its industry as one in
which it had a relatively small percentage of the market. Defining
its industry as the automation of accounting and record keeping,
the company said its products competed with adding machines,
desk calculators, typewriters, key-operated accounting machines,
punched-card accounting machines, addressing machines, mar­
ginally punched cards, pegboards and other form alignment de­
vices, visible index equipment, and large-capacity electronic
machines.
It was a reasonable position to take. The company's punched­
card equipment competed against these products for every new
installation and whenever established installations were ex­
panded. Principal m�MafetfBtse products, with which
252 Building IBM

IBM said it competed, ranged in size from Remington Rand and

the National Cash Register Company, with gross revenues of $236
million and $227 million, respectively, down to R. C. Allen Busi­
ness Machines, with revenues just over $8 million. Numerous
smaller companies were also active in the industry. Listing its own
domestic revenues as $334 million, IBM claimed it had less than
20 percent of the domestic market.12
Consistent with IBM's position, an independent study done for
the Burroughs Corporation revealed that the percentage of sales
in the "office machines industry" in 1951 were divided as follows:
IBM, 24; Remington Rand, 20; NCR, 19; Burroughs, 11; and all
others, 26.13 Clearly the share of the market held by any company
depended on which market was considered and precisely how it
was defined.
Adhering to a much narrower market definition that excluded
all but punched-card tabulating equipment and related supplies
and services, the Department of Justice claimed IBM had 90
percent of the domestic market. Even IBM admitted to having
over 80 percent of this more narrowly defined market.14
A careful reading of the government's complaint suggests that
much of IBM's success, and specifically its gain in market share,
resulted from technologically superior products, protected by a
strong patent position. The complaint alleged, however, that the
company's patent position had been achieved and maintained in
part through illegal means and that it was improperly used to
restrain trade.
According to the complaint, IBM had (among other things)
"systematically acquired developments, inventions, and patents
made or owned by others relating in any way to tabulating ma­
chines"; "prevented, by its leasing system, the experimental use
of IBM leased machines and required compulsory grant backs of
inventions which might result from a breach of this condition";
"used its control over experimental use of its tabulating machines
to obtain inventions relating to the electrical tabulating system
made by others engaged in joint developments with IBM"; and
"opposed the procurement by others of patents relating to the
electrical tabulating system, by ... systematically preempting the
services of inventors active in the tabulating field by employing
them on a long term exclusive retainer basis." 15
Among allegedly offensive uses of its patent position was intro­
duction of the patented 80-column card with rectangular holes
Copyrighted Material
 Legacy 2.53

that made IBM equipment incompatible with that pf Remington

Rand. No longer could equipment of one company handle cards
of the other. An organization wishing to standardize its operations
now had to select either a fully IBM or a fully Remington Rand
system. As reported in the complaint: 'The necessity of having
uniform records of accounting and statistical data within the
Department of National Defense has led the Department to
standardize on the electrical tabulating system .... Other govern­
ment agencies and large businesses have, for similar reasons,
standardized on the electrical tabulating system." Also presum­
ably offensive was IBM's refusal to license any other company to
use its patents for the manufacture and distribution of tabulating
equipment using electrical detection of holes punched in cards.
Remington Rand, for example, was licensed to use IBM patents
only in equipment that detected holes mechanically.16
Not all issues raised by the government related to patent prac­
tices. For example, the company was said to have acquired foreign
businesses that were potential competitors of IBM in the United
States; "granted discriminatory concessions to certain of its lessees
to forestall the acquisition of business by potential competitors";
and "operated service bureaus to preempt the available demand
for tabulating service. "17
Most of the allegations against IBM, even if proven, would have
been legal if undertaken by a minor company in the industry. But
in the murky field of antitrust law, actions judged to be legal when
performed by a small company could be judged illegal if per­
formed by a larger company. Thus there was considerable uncer­
tainty as to the outcome of the case. IS

Settling the Antitrust Suit

Incensed that the government was attempting to destroy his hard­

won success and convinced he could win the case, the senior
Watson insisted that a vigorous defense be mounted. As the
pretrial proceedings dragged on, however, the antitrust suit be­
came a heavy burden for his son, who had been promoted to
president of IBM just six days before the Justice Department filed
the suit.19 Every business decision now had to be judged by its
effect on the legal case.
For reasons similar to those that provoked Hollerith to give up
fighting the government in 1910 to concentrate on new business,
Copyrighted Material
254 Building IBM

Tom Watson now wanted to settle the antitrust suit quickly. But
his father, who as chief executive officer still had ultimate author­
ity, was adamant that the case be fought and won. He believed
IBM was innocent of any wrongdoing and deserved a complete
exoneration by the Justice Department. A settlement out of court
would appear to be an admission of guilt. Angry arguments
erupted between father and son over this issue-as they fre­
quently had over other issues. Finally in late 1955 the senior
Watson capitulated. 2o
The case was settled by a consent decree inJanuary 1956. Titled
"Final Judgment," the document asserts that both parties to the
suit had "consented to the entry of this Final Judgment, without
trial or adjudication of any issue of fact or law herein and without
any admission by either party with respect to any such issue."
Using such phrases as "IBM is hereby ordered and directed" and
"IBM is hereby enjoined and restrained," the document set
forth rules under which the company would have to conduct it')
business. 21
The most onerous provisions for Watson, Sr., probably related
to the manufacture and sale of punched cards.22 This was the
heart of the business he had taken over from Hollerith. It still
accounted for approximately 20 percent of the company's gross
revenue and 30 percent of its net profits. To help others enter
and succeed in the manufacture and sale of punched cards, IBM
was to be constrained in its pricing policies, its contracts with
suppliers of paper stock, and the type of specifications it could
establish for cards used with its equipment. It was further ordered
to offer to sell to competing manufacturers "rotary presses in
good condition" and "paper suitable for the manufacture of
tabulating cards . .. not required for the reasonably anticipated
needs of IBM." Finally, if truly competitive conditions had not
been achieved within seven years, IBM was to divest itself "of such
part of its then existing capacity for the manufacture of tabulating
cards as may then be in excess of 50% of the total capacity for
the manufacture of tabulating cards in the United States."23
The most important provisions for the Justice Department's
attorneys appear to have related to IBM's policy of renting rather
than selling its equipment. In the future the company would be
required to offer to sell its equipment "upon terms and condi­
tions which shall not be substantially more advantageous to IBM
than the lease charges." Further, IBM was required to "afford to
Bm
its salesmen compenseB�Pri eW 1M9Ierial' which shall be not less
 ugacy 255

favorable to them than their compensation for leasing the same

machines. "24
By requiring IBM to offer to sell its equipment, the government
expected to increase competition in many business areas, such as
repair service, secondhand equipment sales, and service bureaus.
To facilitate this process, IBM was required (among other things)
to provide outside service organizations with service manuals,
replacement parts, and personnel training equivalent to that
provided to IBM's own employees.25 The company's service bu­
reau, which offered information processing services to customers,
was to be known as the Service Bureau Corporation without any
evident tie to IBM. It would not be permitted to use the IBM
name, and all dealings with IBM would have to be on the same
basis as were available to any other service bureau. Finally, the
Service Bureau Corporation would not be permitted to employ
people who were also employed by IBM. This final constraint
created a serious problem. The company's service bureau em­
ployees would have to leave IBM or find entirely new assign­
ments.26
A particularly controversial provision of the consent decree was
the requirement that IBM "grant to each person making written
application therefor an unrestricted, n onexclusive license ... for
the full unexpired term of, any, some or all IBM existing and
future patents." This provision applied to patents on electronic
data processing machines (computers) as well as to the company's
more traditional products. A reasonable royalty could be charged,
except that no royalty could be charged for already existing
patents on tabulating machinery and cards. Thepatent licensing
requirements were made more palatable by a provision that per­
mitted IBM to refuse to offer a license to any company that would
not, in return, offer a license to IBM for its relevant patents.27
The requirement that IBM give up its constitutional right to
profit from its patents seemed extreme to many. One could argue,
however, that if patents were acquired improperly, or used to
restrain trade excessively, then that right should be lost. A more
pragmatic reason for accepting the government's terms was the
political climate. Only one day earlier, the American Telephone
and Telegraph Company had agreed to license its past, present,
and future patents as part of a settlement of its antitrust suit with
the government. In discussing the antitrust settlements with
AT&T and IBM, the chief of the Justice Department's antitrust
division , said they "suoolement t:aJ,:h o.ther." These suits and one
�opyngnrea Mafenaf
256 Building IBM

pending with the Radio Corporation of America should be

viewed, he asserted, "as part of one program to open up the
electronics field."28
The impact of the antitrust settlement on IBM was hard to
predict, but it was equally difficult to predict the future without
it. Some of the mandated changes were likely to occur with or
without the settlement-for example, allowing customers to buy
as well as lease equipment. Concerning the patent licensing re­
quirements, the introduction of electronics had already cost the
company its dominant patent position. Its patents on magnetic
tape and disk drives were more valuable than yet realized, and
the company badly underestimated the value of its SSEC patents.
Nevertheless IBM would need licenses to patents held by many
others to compete successfully in the rapidly evolving field of
electronics.
Events soon vindicated Tom Watson's decision to suffer the
many constraints of the settlement rather than fight the case in
court. He correctly believed the company's future was in electron­
ics and that magnetic tape and disk storage would replace
punched cards. His management team had to be free to concen­
trate on the future and to adapt the company's products and
business methods to the onrush of new technologies. The com­
pany could not afford to sit still while waiting to be judged.

Domestic Competition

The January 1956 consent decree provided new guidelines and

administrative burdens for the conduct of IBM's business, but
Tom Watson believed the primary challenge remained un­
changed. That challenge was to shift the product line from what
the court defined as "tabulators" to what it defined as "electronic
data processing machines."
In making this technological transition, IBM's primary com­
petitor would continue to be Remington Rand, which had be­
come a division of Sperry Rand during a June 1955 merger. Tom
Watson already knew, however, that the number of installations
using large-scale computers manufactured by IBM had finally
surpassed the number using Remington Rand computers.29 In
addition to thirty completed installations of IBM 700 series com­
puters at the end of 1955, there were five times that many new
orders. Poor cooperation between the ERA and Eckert-Mauchly
divisions, poor coupling between engineering and manufactur-
Copyrighted Material
 Legacy 257

ing, a weak marketing effort, and limited customer support had

cost Remington Rand its once preeminent position. But with
gross revenues 50 percent larger than those of IBM, the recently
formed Sperry Rand Corporation had the financial resources to
overcome these problems if its management had the capability
and motivation.30
A new entrant to the market was the Radio Corporation of
America (RCA), which had installed its first BIZMAC in 1955.
This large-scale electronic computer in a "typical" configuration
was priced at about $1.5 million, although a specially configured
system then under construction was reported to have been sold
for $4 million. With gross revenues twice those of IBM and
substantial experience in military and consumer electronics. RCA
was a feared competitor. Another significant entrant was the
Datamatic Corporation, which had been formed as a jointly
owned subsidiary of the Minneapolis-Honeywell Regulator Com­
pany and the Raytheon Manufacturing Company. The combined
revenues of these two companies nearly equaled IBM's. Data­
matic's first product offering was the DATAMATIC-I000, selling
for about $1.8 million. Like recent large-scale computer offerings
of IBM, Sperry Rand, and RCA, it was equipped with ferrite-core
main memories.
In smaller stored-program computers selling under $300,000,
IBM's primary competitors were NCR with 27 machines installed
and the Electrodata Corporation with 14 installed at the begin­
ning of 1956. Other entrants with low-end machines were the
Bendix Aviation Corporation, Burroughs Corporation, Hogan
Laboratories, Inc., Logistics Research, Inc., Marchant Calculators,
Inc., Monroe Calculating Machine Company, Sperry Rand Cor­
poration, Technitrol Engineering Company. and Underwood
Corporation. These other companies collectively had fewer than
20 operational installations as 1956 began, but the number of
machines on order was considerably higher. All but one of the
smaller machines had a magnetic drum memory as did the IBM
650, of which over 100 had been installed and 750 were on
order.31
This brief summary of the domestic competition in electronic
data processing machines does not suggest IBM was in difficulty,
but it does portray a far more competitive market place than
existed for the company's traditional EAM equipment. With so
many new competitors and a rapidly evolving market, the out­
come was far from ptedetermineti. � .

c.;opyngnrea MalenaI
258 Building IBM

The World Trade Corporation

By far the largest part of IBM's revenue from outside North

America came from its European operations.32 Here the World
Trade Corporation was busy facilitating the activities of its wholly
owned operating companies in each major country. Its tasks in­
cluded the selection of sites for manufacturing plants and devel­
opment laboratories to make IBM products available throughout
Europe as economically as possible. The patch-work quilt of cus­
toms regulations, national standards, and taxes that governed
commerce among the many independent countries made this
task particularly challenging. Obtaining patent protection in each
country was another important responsibility that the World
Trade Corporation shared with its national companies.
Particularly important was IBM Germany. This national com­
pany had grown out of Dehomag (Deutsche Hollerith Machinen
Gesellschaft), which had been established by a German business­
man in 1910 to distribute and service Hollerith equipment in
Germany and certain other European countries. In 1922 IBM
(then called CTR) purchased a controlling interest in the com­
pany through a sale by its founder.33 During World War II the
business was devastated, as were most other businesses in Ger­
many. As soon as hostilities ended, the senior Watson arranged
to send food to its employees, as well as tools to work with and
used equipment to recondition and place on rental. As a result
IBM's subsidiar y was among the first German businesses to return
to full operation after the war.
The situation in Great Britain was quite different and had
presented a serious barrier to the creation of the World Trade
Corporation in October 1949. It will be recalled that the British
Tabulating Machine Company (BTM) had acquired exclusive
rights from Hollerith to manufacture and distribute the com­
pany's products in all countries of the British Commonwealth
except Canada. Because this contract could not be terminated
without the consent of BTM, the senior Watson had been forced
to offer a nonexclusive license, free of charge, on all existing
products as well as on all issued and pending patents. These terms
were attractive to BTM. Without the large royalties it was pre­
viously obliged to pay, the British company expected to be more
profitable and better able to develop new products. All that IBM
got in return was the right to compete with BTM.34
Copyrighted Material
 Legacy 259

It was a significant gamble. IBM would now have to compete

in Britain (and elsewhere) against its own products and technolo­
gies, manufactured and marketed by an established British firm.
Among the factors that pushed IBM toward this decision was the
threat of antitrust action. The U.S. Justice Department had begun
preparing its antitrust case with a series of inquiries at least five
years before it filed the lawsuit in 1952. It was known in IBM by
1949 that its agreement with BTM was under scrutiny. Ofparticu­
lar concern was the provision that the two companies not com­
pete in each other's territories.35
Beyond the antitrust considerations, the senior Watson had
long been distressed by the slow rate at which BTM was develop­
ing its mar ket . There had also been continual disputes over
royalty payments and other aspects of their relationship.36 Above
all Watson believed his concept of a World Trade Corporation
would provide a more effective means for developing and distrib­
uting products worldwide.37

Competition Abroad

In 1949 when IBM and BTM were reconsidering their a greemen t,

the British-based Powers-Samas company was renegotiating a simi­
lar agreement with Remington Rand. Powers-Samas had its ori­
gins in the Accounting and Tabulating Machine Company of
Great Britain Limited, which was established in November 1915
as a subsidiary of the American-based Powers Accounting Ma­
chine Company. Like BTM it began manufacturin g some of its
own equipment during the 1920s.
In 1929 it eliminated its dependence on sales agencies by
establishing its own sales organization under the name, Powers­
Samas Accounting Machines Limited . The entire company and
its products were soon known by this name. The name Samas had
previously been used to designate the company's machines in
France, where Remington Rand had refused to allow it to use the
registered name Powers. 38
A clause in their agreement that prevented either company
from competing in the other's territory had come to be viewed
more as an impediment than an advantage by the Power-Samas
company. Among other things it prevented the company from
selling equipment to Canada, whose large exports of wheat to the
United Kingdom made it a highly desirable and willing customer.
Copyrighted Material
260 Building IBM

Another disadvantage was the relatively poor performance of

Remington Rand's Tabulating Machine Division in competition
with IBM. "One wonders how long the unequal struggle can
continue," the general manager of Powers-Samas pondered.39
Spurred to action by IBM's termination of its agreement with
BTM, the Powers-Sam as company notified Remington Rand of its
intention to end their agreement as well. For Remington Rand
there was at least one major incentive for acquiescing, namely,
the antitrust implications. Although an agreement not to com­
pete in each other's territories was not illegal per se, Remington
Rand had been advised that such an agreement would figure
prominently in any antitrust action brought by the government.
For this reason, among others, the separation proceeded with
little resistance by either party.40
In addition to the four punched-card equipment companies
already identified in Europe, there was another fully integrated
company. Machines Bull of France had grown out of the efforts
of Frederik R. Bull, a Norwegian engineer, who undertook in
19 18 to make equipment less expensive and more reliable than
Hollerith's. More than a quarter of a century after Hollerith's
equipment began tabulating the U. S. census of 1890, Bull devised
his first tabulating machine. It was a combined tabulator and
sorter. When neither function operated efficiently, he began de­
signing separate units. Like Hollerith, he chose electrical rather
than mechanical detection of the holes in cards. Following Bull's
untimely death in 1925 at age forty-three, his work was continued
by another engineer.
Finding a suitable manufacturing location was difficult, in part
because the equipment used concepts covered by IBM patents.
Switzerland was selected because IBM held no patents there. In
1931 the operation was moved to France (where IBM also held
no patents) to take advantage of its much larger domestic market.
In 1933 the company took the name Compagnie des Machines
Bull.41 Three years later, it had seventy-two installations in France
and six others elsewhere in Europe. It continued to expand its
operations until World War II and moved aggressively after the
war. 42
Competing against Remington Rand, Machines Bull, BTM, and
Powers-Samas, IBM gained market share at home and abroad.
Domestic and foreign revenues both increased an average of 22
percent per year during the 1950s. The United Kingdom was the
�� �
only significant geogreB h h�WMa��}la'Yhich IBM had less than
 Legacy 261

half of the business:13 In 1954 when Dick Watson was elected

president of the IBM World Trade Corporation, it still had fewer
than 600 employees in the UK, whereas BTM and Powers-Samas
each had over 5,000.44
Both of these British companies gave primary emphasis to
strengthening their lines of punched-card equipment, thus leav­
ing the European market for electronic stored-program comput­
ers largely to IBM and Remington Rand. The slowness of BTM
and Powers-Samas to exploit electronic computer technology is
surprising in view of Britain's leadership in electronics during
World War II and its subsequent pioneering contributions to
electronic computers.45
Remarkably the J. Lyons bakery was the first British company
to build an electronic computer for business data processing. In
1951 it built the LEO computer in collaboration with Cambridge
University, and in 1955 it formed Leo Computer Limited to
exploit the market. By then Ferranti, Elliott Brothers, and English
Electric had also entered the embryonic computer field. The
British computer industry did not begin to take form, however,
until BTM and Powers-Samas merged to form International Com­
puters and Tabulators Limited in 1958.46
Thus the legacy of Thomas J. Watson, Sr., at the time of his
death in 1956, was a company whose primar y punched-card
equipment business dominated the domestic market and led in
all significant foreign markets except Great Britain. In electronic
computers, IBM faced stiff competition in the United States, but
it had more small-to-medium-sized computers installed by far
than all of its competitors combined. Even in large computers,
the number of its installations exceeded the number of its nearest
competitor, Sperry Rand. Outside the United States there was
little prospect of significant near-term competition in electronic
computers, except for that offered by U.S. companies and any
alliances they might make with European companies. IBM's
World Trade Corporation, therefore, faced relatively weak com­
petition and a vast and rapidly growing market for its products.
The senior Watson had provided the initial technical and mar­
ket base for the company's entry into electronic computers, but
the success achieved by 1956 was largely attributable to the team
Watson, Jr. , had assembled and led. Indeed an important part of
the legacy of Watson, Sr., was his older son, whom he had
groomed for leadership.
Copyrighted Material
 18
Gambling un System/360

Tom Watson took little comfort in IBM's relatively strong position

in punched-card equipment and electronic data processing sys­
tems. He knew that many companies were being lured into the
industry by prospects of profiting from electronics expertise,
acquired through military work. Adding to his concern was the
antitrust settlement, which had placed enormous constraints and
requirements on IBM's normal business operations. Driving him­
self and his associates to overcome these problems, Watson
launched IBM on a development effort that transformed the
company and shaped the information processing industry for
years to come.

Organizing for Growth

Modernizing the company's archaic organizational structure and

putting his most trusted associates in key positions was an early
priority. In November 1956 Watson assembled over one hun­
dred IBM executives to announce a companywide reorganization.
The structure in which all major decisions had funneled to his
father was being replaced by one that emphasized decentralized
responsibility. '
Some decentralization had already begun with creation of the
Electric Typewriter and Military Products divisions in late 1955
and a Supplies Division in early 1956. Now he proposed to con­
solidate and extend these earlier steps. Particularly appealing to
Watson was a reduction in the number of individuals who re­
ported directly to him from seventeen (as it had been in 1954)
to only five. These five individuals, who retained their titles of
vice president or executive vice president, were given the addi­
tional title of "group executive" to signify they reported directly
Copyrighted Material
264 Building IBM

to the president and typically had responsibility for more than

one functional unit.
The group executives with assignments most clearly illustrating
the new structure were Al Williams, under whom all corporate
staff functions had been collected, and Vin Learson, who now
had responsibility for the Military Products, Time Equipment,
and Special Engineering Products divisions, as well as the pro­
posed Service Bureau Corporation.2 The other three group ex­
ecutives were L. H. LaMotte, responsible for the newly formed
Data Processing Division; H. W. Miller, Jr., responsible for the
Electric Typewriter and Supplies divisions; and the younger Wat­
son brother, Dick, who retained responsibility for the World Trade
Corporation. Finally a Corporate Management Committee was
established, consisting of the five group executives plus Tom
Watson as chairman.3
Although the broad outlines of this structure survived for many
years, a specific failing was observed. Continuing rapid changes
in technologies and market requirements were putting major
stresses on the Data Processing Division. The organizational struc­
ture that treated this business like any other business was not
adequate. More significantly, the one person Watson believed
could understand the complex technical and marketing issues of
the EDPM business had been assigned elsewhere. That person
was Vin Learson, the man to whom he had turned for help in
1954 to get the EDPM business started.
Watson's solution was to put Learson in charge of defining,
developing, and manufacturing all products for the Data Process­
ing Division. Implementing this decision in May 1959, Watson
removed the product development and manufacturing functions
from the Data Processing Division and split them into two new
divisions. Each of the new divisions was responsible for develop­
ing and manufacturing roughly half of the EDPM product line.
The Data Systems Division (with plants and laboratories in
Poughkeepsie) was responsible for large computer systems, typi­
cally renting for more than $10,000 per month. The General
Products Division (with plants and laboratories in Endicott, New
York; San Jose, California; Burlington, Vermont; and Rochester,
Minnesota) was responsible for small systems, typically renting for
less than $10,000 per month. The current product offerings of
GPD included EAM equipment, RAMAC, and the 650 and 1401
computers; those of DSD included the 700 and 7000 series. For
Copyrighted Material
 Gambling on System/360 265

historical reasons DSD had responsibility for magnetic tape stor­

age and GPD for magnetic disk storage.
These two divisions were placed under Group Executive Lear­
son, who was also given responsibility for a newly created Ad­
vanced Systems Development Division. This division was to
specialize in systems work that was too applied for Research but
too exploratory for product development. All divisions that pre­
viously reported to Learson were placed under other group ex­
ecutives, except for the Time Equipment Division, which had
been sold in 1958. The Data Processing Division, which now had
responsibility for sales and service of EAM and EDPM products,
was also placed under one of the other group executives.4

Impact of the IBM 1401

The new structure put Learson in the middle of a growing rivalry

between two laboratories. Engineers in Poughkeepsie had created
the company's first electronic products, and they had bested the
highly touted Remington Rand engineers in large computers. It
was the engineers in Endicott, however, who had built the IBM
650. Although disparaged by Tom Watson as "a bunch of mon­
key-wrench engineers" and required to compete against ERA for
the right to develop the 650, the Endicott engineers had pro­
duced the most popular computer of the 1950s.5
Now at the time Learson acquired responsibility for defining,
developing, and manufacturing all electronic data processing
systems, the Endicott laboratory was completing another small
computer system. Announced in October 1959 as the IBM 1401
Data Processing System, it used SMS circuit packaging technology
developed in Project Stretch. Equipped with ferrite-<:ore memo­
ries having capacities of 1400 to 4000 characters, the 1401 could
be configured to use punched cards and magnetic tape, and it
could be used either as a stand-alone computer or as a peripheral
system for larger computers.6
The timing of its development was fortuitous. Several impor­
tant IBM-developed components had just been improved or were
available for the first time. In the latter category was the newly
devised chain printer. It provided unusually good print quality at
a speed of 600 lines per minute. This was four times faster than
most printers then available. The chain printer helped to make
the 1401 the most popular computer of the early 1960s. The
Copyrighted Material
266 Building IBM

Figure 18.1. IBM 1401 Computer

The smallest version of the IBM 1401 Data Processing System is shown here
with Type 1402 card read-punch (left), Type 1401 processing unit (center),
and Type 1403 chain printer (right). Announced in October 1959, the 1401
was the most widely used computer for many years. Magnetic tape, and later
magnetic disk, storage units were available as optional equipment.

monthly rental for a 1401 system was $2,500 and up, depending
on the configuration chosen. By the end of 1961 the number
installed in the United States had reached 2,000. This was about
25 percent of all electronic stored-program computers installed
by all manufacturers to that time. The number of 1401s installed
peaked at more than 10,000 in the mid-1960s.7
The IBM 1401 announcement sent shock waves through the
information processing industry. In the United States many new
entrants were forced to drop out. Although able to build the basic
electronic central processing unit, they were not prepared to
develop and manufacture low-cost ferrite-core memories or to
match IBM's engineering and manufacturing capability in
punched-card equipment, magnetic tape and disk storage, and
printers. They also lacked appropriate marketing expertise.
Sperry Rand was among the survivors. Its most successful small
computer in the late 1950s and early 1960s was its Solid State
Computer. Using an IBM-compatible 80-column card, this com­
puter was marketed at first only in Europe to avoid making
obsolete the company's 90-column-card equipment in the United
States. When finally released in the United States as the S5-80
computer, it was outclassed by the newer IBM 1401 in cost and
performance and in the variety of configurations available.
Rather than investing in a replacement for the S5-80, Sperry
Copyrighted Material
 Gambling on System/360 267

Rand chose to upgrade the eXlstmg product during the early

1960s. The result was not satisfactory. The number installed
peaked at about 600, which was only 6 percent of the number of
1401s.8
The number of European companies offering computers was
smaller than the number of U.S. companies. A lower level of
government funding and a less well developed market for
punched-card equipment were primary reasons. These factors
probably also accounted for the slowness with which Machines
Bull of France and the newly formed International Computers
and Tabulators Limited (IC T) in Britain entered the field. For
example, IC T did not deliver its first electronic computer until a
year after the 1401 was available, and significant quantities were
not available for yet another year. Thus it was primarily the IBM
1401 that effectively terminated the sale of conventional
punched-card equipment there just as it had in the United
States.9
The overwhelmingly good acceptance of the 1401 increased
the rivalry between the IBM laboratories in Endicott and
Poughkeepsie. So intense was it that sometimes it seemed to
exceed the rivalr y with external competitors. Channeling this
along beneficial lines was a major challenge for Learson.
The success of the 1401 also increased the need to reassess the
company's EDPM strategy. With the 1401 rapidly replacing con­
ventional punched-card equipment, the products of the Endicott
and Poughkeepsie laboratories would soon differ primarily in
their cost and performance. It should be possible for customers
to migrate easily from low-cost to high-cost equipment, but none
of the Poughkeepsie machines could run programs written for
Endicott machines. In fact, none of the first six IBM computers
with transistor circuits could run programs written for another. IO
The lack of compatibility was serious for customers, but it was
even more of a problem for IBM. The company had to train sales
and service personnel and provide programming support for
each of the incompatible systems. The process was costly and
chaotic. Economies of scale in engineering and manufacturing
were reduced. For example, peripheral equipment designed for
one computer could not be used on any other without modifica­
tion. As Learson fully understood, these were not simple prob­
lems to solve. The engineers who had designed each of the
computers and the customers who used them were committed to
them and to their ex tc8�Wf(ihte d Material
268 Building IBM

Seeking Compatibility

In September 1960 the 1401-compatible (and more powerful)

IBM 14lO was announced. Immediately successful in the market­
place, it demonstrated how valuable compatibility was. Meanwhile
the Poughkeepsie engineers were nearing completion of a plan
to create a series of new computers. The series consisted of four
separate processors that offered two levels of performance for
business applications and two levels for scientific and engineering
computing. The most powerful computer was to match the per­
formance of Stretch. Known as the 8000 series and intended to
replace the 7000 series, the Poughkeepsie proposal was presented
to the Corporate Management Committee in January 1961. The
enthusiasm engendered by the presentation convinced Learson
he would have to act quickly to achieve compatibility in future
products. If he failed, the two laboratories might well battle each
other in the marketplace with incompatible product lines. 11
Only a few days after the 8000 series presentation, Learson
acted. He removed the systems development manager in
Poughkeepsie and replaced him with Bob O. Evans, who had
been engineering manager for the 1401 and 14lO projects in the
Endicott laboratory. Evans was thus already committed to com­
patibility, and he had no vested interest in the Poughkeepsie
proposal. Three months after his appointment, Evans recom­
mended that work on the 8000 series be terminated. In its place
he urged that an effort be undertaken, with interdivisional par­
ticipation, "to develop a total cohesive product line."12
Acceptance of his recommendation would necessarily delay the
availability of new IBM systems. To prevent competitors from
replacing IBM's aging equipment, Evans proposed a "temporiz­
ing" plan in which the older machines would be upgraded. Fur­
ther support for a delay in developing new systems was provided
by a corporate study on components. It recommended that im­
proved semiconductor devices and circuit packaging technologies
be developed to replace SMS. Known as Solid Logic Technology
(SLT) , the new circuit technologies could not be ready until at
least a year later than required to meet the 8000 series schedule.
Delay of a year or more in developing new systems could improve
their competitiveness through use of SLT.
The proposed change in plan was vigorously opposed by
Frederick P. Brooks, Jr., lead designer of the 8000 series. With a
1956 Ph.D. from HO� �� 9'MBj9�ering computer science
 Gambling on Systern/360 269

Figure 18.2. Bob O. Evans

Bob Evans is shown demonstrating a System/360 model at the time of its an­
nouncement in April 1964. Three years earlier he had persuaded manage­
ment to abandon other product plans in favor of this compatible line.

program at Harvard and experience as a Stretch designer, Fred

Brooks was an articulate and knowledgeable opponent. The vigor
with which he pressed his views gave no hint of his subordinate
position to Evans in the company 's hierarchy. To break the stale­
mate, Learson replaced Brooks's manager withJerrier A. Haddad,
who had most recently served for two years as head of the Ad­
vanced Systems Development Division. Previously he had been
chief engineer for the IBM 701 development. Haddad's first task
in his new assignment was to review the two proposals. He did so
and sided with Evans. In May 1961 Learson terminated the 8000
series project. 13
Bob Evans immediately asked Brooks to lead a search for the
"ultimate" family of systems to serve all customers. "To my utter
amazement, Bob asked me to take charge of that job after we had
been fighting for six months," Fred Brooks recalls. "I was dumb­
struck. "\4 Brooks eagerly accepted, and his infectious enthusiasm
and competence soon won the support of others. 15
The plan to create an interdivisional product line was not
without opposition. The mpst influential opponent was . John W.
Copynghted Material
270 Building IBM

Figure 18.3. Frederick P. Brooks, Jr.

Mter serving as project manager for System/360 from its inception to near
its announcement, Fred Brooks took over responsibility for development of
the critical operating system, OS/360.

Haanstra, president of the General Products Division. Firmly

committed to the independence of IBM divisions, Haanstra was
insistent on retaining self-determination for his division to de­
velop and manufacture its own products. With the highly success­
ful 1400 series developed in Endicott and the pioneering
magnetic disk storage products developed in San Jose, he was
confident of the capability of his division. His enthusiasm for disk
storage was heightened by his own early experience, working on
disk storage in San Jose during the 1950s. He believed that
developments underway on a low-cost 1401-like processor and on
a low-cost file, with the first removable disk packs, would greatly
expand IBM's customer base.

The SPREAD Task Group Report

To overcome Haanstra's opposition and to secure the coopera­

tion of others, Learson established a corporatewide task group
with Haanstra as chairman and Evans as vice chairman. Code­
named SPREAD (an acronym for Systems Programming, Re­
search, Engineering, And Development), the task group was "to
establish an overall IBM plan for data processor products. "16 Fred
Brooks was one of eleven other members. The SPREAD Task
Copyrighted Material
 Gambling on System/360 271

Group issued its final report at the end of December 1961. One
week later it was presented to the company's top executives and
their staffs at the new T. J. Watson Research Center. The reaction
was mixed, but Learson concluded the report offered the best
available plan and ordered that it be implemented.'7
The report's twenty-six pages contained product recommenda­
tions, analyses of their likely impact, and supporting technical
and market information. Its primary recommendation was that a
new line of "compatible processors " be developed that would
span a broad range of performance. A processor was defined to be
the central processing unit plus main memory and channels for
attaching peripheral equipment. Compatible processors were
defined as ones "capable of operating correctly all valid machine­
language programs of all processors with the same or smaller 1/0
and memory configuration. "18
The proposed product line was to consist of five processors
spanning a 200-fold range in performance. This range was to be
achieved by using SLT circuits of differing speeds, data paths of
differing widths, memories of different sizes and speeds, and a
variety of other engineering trade-offs. These trade-offs would be
the responsibility of each engineering team assigned to develop
one of the processors. Each engineering team was also mandated
to develop an "economically competitive" processor, rather than
depending on the rest of the line to ensure its market success.19
Use of a small high-speed memory with permanently stored
information to control the action of logic circuits in the processor
was strongly urged. Known as a "read-only control store," such a
device would obviate the need to wire into the machine the
capability to perform each of the programming instructions re­
quired of all processors in the compatible line. In Hursley, Eng­
land, IBM's laboratory had conducted extensive studies of control
stores, based on a 1951 proposal by M. V. Wilkes at Cambridge
University.2o Promoted vigorously by the representative to
SPREAD from the Hursley laboratory, the proposed use of con­
trol stores would be their first commercial application.
In operation, each programming instruction (such as add,
multiply, or conditional branch) was equated to an address in
control store. When the word at this address was read out, its
permanently stored Is and Os were used to open or close associ­
ated computer logic gates as required to execute the instruction.
If an instruction execution required a sequence of such settings
(i.e., more than one machine cy�le), the first word addressed
Copyrighted Material
272 Building IBM

carried the address of the next control-store word needed to

complete the instruction execution.21
Noting that many customers were using the same computer for
scientific and business applications, the task group carried out a
study to determine what instructions (if any) were used by one
type of application but not by the other. Only floating-point
instructions (used in engineering and scientific computations)
were found to be of this type. It was therefore decided that each
system in the line would be marketed for both types of applica­
tions. At the discretion of the engineering design team for each
processor, however, floating-point capability could be offered
either as a standard feature or as a field-installable option.22
A fundamental and contentious issue related to character size.
Prime contenders were 6-bit and 8-bit characters. The 6-bit char­
acter was favored because it could represent all 26 letters of the
alphabet plus punctuation symbols and because it was used in
most computers, including the popular 1401. It was also used to
record information in IBM's magnetic tape and disk storage
products. It was opposed because it had 2 bits more than needed
to represent a decimal digit and thus wasted expensive memory
capacity. The 8-bit character was favored because it could be used
to store 2 decimal digits efficiendy and because the 2 bits more
than needed for alphabetic symbols would provide for a wide
range of special symbols in the future. The 8-bit character-which
had been christened the "byte" when used in Stretch-was se­
lected. Experience indicates it was a good decision. The 8-bit byte
continues to be used in almost all computers.23
Among other significant architectural decisions was the re­
quirement that standard interfaces be used for input-output
(I/O) equipment such as tape and disk storage, card readers,
printers, and terminals. As stated in the report, ''When one proc­
essor is substituted for a slower one, the I/O gear shall not need
to be changed. "24 This provision would not only permit customers
to upgrade their systems gradually, but it would reduce engineer­
ing design and field service costs.
Although supportive of much of the SPREAD report, Haanstra
and his Endicott engineers were apprehensive about the lack of
compatibility of the proposed computers with any of the com­
pany's established products. As stated in the report, "Since such
processors must have capabilities not now present in any IBM
processor product, the new family of products will not be compat­
ible with our existing Qrocessors. "25 All members of the task group
"Copyrighted Material
 Gambling on System/360 273

knew that incompatibility with existing products would be a bar­

rier to sales, but the advantages of the new line were expected to
overcome this barrier. It was not until early 1963 that the possi­
bility of competitors marketing machines compatible with IBM's
present ones, but having superior cost and performance, appears
to have surfaced as a serious concern.26

Saved by Emulation

The competitive action that Haanstra feared occurred in Decem­

ber 1963 when Honeywell announced its low-cost H-200 com­
puter. A program, named "Liberator," was announced for use in
translating 1401 programs to programs capable of running on
the H-200. Prices were not quoted in the announcement, but up
to a fivefold improvement in cost-performance over the four-year­
old 1401 was indicated. Within two months the IBM sales force
had reported 196 losses to the H-200.27
The possibility that translators, such as Honeywell's Liberator,
could be devised to translate programs from IBM's current ma­
chines to run on its new line had been considered and aban­
doned in IBM well before the Honeywell announcement was
made. Studies had revealed that code produced by automatic
translation would be inefficient to run unless the two machines
were quite similar. as were the H-200 and 140 1. Semiautomatic
translation was also attempted, but achieving the desired code
efficiency required too much human effort.
The well-proven method of simulating one computer with an­
other was also explored. In simulation, the "host" computer is
programmed to run programs written for the "target" computer
by simulating its architectural features. Again the achievable per­
formance was shown to be acceptable only if the two machines
were quite similar. Development of a cost-effective means for the
conversion of programs written for current machines, to run on
those of the new product line, was proving to be technologically
more difficult than anticipated.
For some time Haanstra had been quietly working to create
his own response to the type of threat Honeywell posed with its
H- 200. Rather than solving the conversion problem, his approach
was to build HOI-compatible computers using SLT. The effort
began unintentionally-or perhaps deliberately-in late 1960
with a decision to design a 1401 processor in SLT to provide a
"feasibility test" of th't��f6HiM�'6JMlology. With its schedule
274 Building IBM

tied to the development of SLT, the feasibility model was not fully
operational until early 1963. Then, when no means for helping
customers convert to the new product line had yet been found,
Haanstra transformed the feasibility-model effort into a full­
fledged product program. By using a relatively small number of
additional circuits in the model, the designers were able to cause
it to run programs for two other members of the 1400 series as
well. The proposed product, tentatively designated the IBM 1470,
was scheduled for announcement in February 1964. It was a
blatant challenge to the new product line being developed under
the leadership of Evans and Brooks.28
Learning of this decision early in January, Bob Evans warned
that marketing a 1401-compatible computer, built with SLT and
improved memory and storage devices, would inevitably lead to
a decision to respond to other competitive threats by taking
similar action for 7000 series computers. His vision of a company­
wide family of compatible computers seemed to be nearing its
end. "I am sad at the consequences," he lamented in a memo that
suggested a helplessness uncommon for him.29
Almost miraculously, his vision of the new product line was
saved by a last-minute technical accomplishment. In mid-1963
engineers in the Poughkeepsie and Endicott laboratories had
begun exploring the possibility of adding special microcode to
the control stores of computers to improve their performance
when simulating earlier IBM computers. The first result of the
Poughkeepsie group was exciting. The addition of one particu­
larly well chosen word of microcode had speeded up a frequently
used simulation loop by almost a factor of fifty. Additional micro­
code produced significant improvements in other parts of the
simulation process. So dramatic was the improvement in perfor­
mance of a simulator with microcode (and selected circuitry)
enhancements that the Poughkeepsie engineers coined the term
emulator to describe the concept.30
Meanwhile some engineers in Endicott were moving toward an
even more ambitious achievement. Because the 1401 instruction
set was relatively simple, they planned to implement all of it in
an expanded control store of the smallest processor of the new
line. To further speed up the simulation, they added a few circuits
and made use of high-speed registers (already in the computer)
to help translate decimal memory addresses of the 1401 into
binary. The result was an emulator with no software portion.
Using their emulator'CBpi������Jt�?lMleW processors could be
 Gambling on S.vstem/360 275

run in "1401 mode" to execute programs written for the 1401

more rapidly than the 1401 itself.31
It was a salesman's dream come true. The new product line
now offered computers that could execute a customer's current
programs faster than his present computer and that provided a
path into the new world of compatible computers, capable of
applications not possible with older equipment. Facing increasing
competitive pressures, the IBM sales organization demanded an
early product announcement.

Announcing System/360

On 7 April 1964, the new product line was publicly announced

as the IBM System/360. To his "Fellow IBMers," Tom Watson said:
"Today we are making the most significant product an­
nouncement in IBM history-the System/360. In Poughkeepsie,
scores of reporters are attending a press conference to hear the
news about this new generation of computers. In branch office
cities throughout the country, many thousands of IBM customers
and prospects have gathered to hear presentations on this new
system. The day so many of you have worked so hard for has now
arrived. "32
About two hundred writers and editors had been brought to
Poughkeepsie on a train hired by IBM. At the press conference,
Watson said: "System/360 represents a sharp departure from
concepts of the past in designing and building computers. It is
the product of an international effort in IBM's laboratories and
plants. . . . This is the beginning of a new generation-not only
of computers-but of their application in business, science and
government. "33
The word system was chosen to signify that the offering was not
just a group of processors with peripheral equipment but rather
an aggregation of interchangeable hardware units with program
compatibility from top to bottom. The number "360," which is
the number of degrees in a circle, was chosen to represent the
ability of each computer to handle all types of applications. Ex­
cept for different costs and performances, all processors in the
System/360 line were equivalent and presumably able to handle
any anticipated information processing task.
The announcement was unprecedented in its scope. Six proc­
essor models, designated the 30, 40, 50, 60, 62, and 70, were
announced. Models �t) �gm9(FMar�a/ replaced by superior 65
276 Building IBM

Figure 18.4. A System/360 Installation

A typical configuration of a relatively small System/360 computer includes
(from left) a card read-punch, four magnetic tape units and controller, two
magnetic disk storage units, and the processor (here a Model 40) with con­
trol panel. The operator sits at the printer keyboard within reach of the con­
trol panel.

and 75 models before any of the original three were shipped.

Each processor had its own range of memory capacities: 8K to
64K bytes on the Model 30, and 256K to 1024K bytes on the
top-of-the-line Model 75. (K is a binary-oriented multiplier equal
to 210 = 1024. Thus 8K bytes is 8192 bytes.)
A fiftyfold performance range in processors was achieved, pri­
marily by two design factors. First, data path widths provided a
factor of eight. From one byte wide in the Model 30, the data
path was successively increased, reaching eight bytes in the Model
65 and above. Second, SLT circuits provided almost a factor of
four with operational times ranging from about 8 nanoseconds
to 30 nanoseconds. A variety of other design techniques achieved
an additional factor of about two in performance.34
Forty-four different peripheral devices were also announced.
These included several types of magnetic storage devices, visual
display units, communication equipment, card readers and
punches, printers, and an optical character reader. Monthly sys­
tem rentals ranged "from $2,700 for a basic configuration to
$115,000 for a typical large multisystem configuration." Compa­
rable purchase prices were said to be $133,000 to $5,500,000.35
The same general information was provided to an estimated
100,000 customers and prospects at special branch office meet­
Copyrighted Material
 Gambling on Systnn/360 277

ings held in 165 cities in the United States. The response was as
unprecedented as the announcement. Orders rapidly exceeded
forecasts. Well over one thousand orders were received during
the first four weeks, and most orders specified more memory and
I/O than anticipated.s6
Watson had gambled almost all of the company's development
resources for more than two years on a single product line. Now
the rewards projected by the SPREAD Task Group were at hand,
but the early order rate did not ensure success. Another two years
of concerted engineering and manufacturing effort were needed
before confirmed orders and successfully delivered products
would guarantee fulfillment of the promise of System/360.

Copyrighted Material
19
Commitment and Delivery

Successful emulation of a 1401 by the smallest computer of the

new product line had failed to change John Haanstra's plans.
Believing his division best understood the needs of its customers,
he continued preparing to announce a 1401-compatible com­
puter built with SLT. He urged that the marketplace be allowed
to decide between it and System/360.
Haanstra was one of IBM's finest engineering managers. Bril­
liant and self-assured, he pursued his goals with unrelenting vigor.
His resistance to the new product line had emphasized the need
for a cost-effective means for converting customers to the new
system, and he had provided a viable alternative for the low end
of the line. Had emulation not been successful, his alternative
might have been used. But once the decision was made to an­
nounce System/360, opposition had no further value. One
month before the product announcement, Haanstra was removed
as president of the General Products Division.' The message this
action delivered was unambiguous: managers and their organiza­
tions would be judged primarily by their commitment and con­
tributions to System/360.

Entering the Components Business

Of all the hardware commitments, none posed greater opportu­

nity and risk than IBM's decision to develop and manufacture its
own semiconductor devices. It was not an easy decision. Battles
had raged for years over whether it was better to make or buy
components. Tom Watson, Jim Birkenstock, and other corporate
leaders contended that the components business had a lower
profit margin than the systems business and should be left to
others. Ralph Palmer, Wally McDowell, and several other key
engineers believed t�p.� ",aWrJalPmponents was essential
280 Building IBM

to leadership in information processing systems. They believed

that advances in components would lead to new system concepts
and unanticipated product opportunities.
Because the general argument was never resolved, decisions
were made on a case-by-case basis. The company had manufac­
tured its own cathode-ray tubes for use in Williams-tube memories
because suppliers failed to provide the necessar y quality, but it
had elected not to manufacture more conventional tubes. Failing
to obtain satisfactor y patent licensing terms from its primar y
supplier of ferrite cores, IBM had acquired patent rights to a
different process and manufactured its own.2
In December 1 957 the company had entered into an important
agreement with the Texas Instruments Company (TI). The con­
tract called for "exchange. of patent licenses, purchasing arrange­
ments, interchange of technical information, and joint
development" of transistors and diodes. It made TI the primar y
supplier of semiconductor devices to IBM.3 Palmer's initial dislike
of the agreement was increased in 1959 when the world's first
fully automatic manufacturing line for transistors-developed by
his engineers-was taken apart and shipped to TI. Initially used
to manufacture transistors only for IBM, the line was soon repli­
cated and used to produce transistors for TI's other customers as
well.
There was growing dissatisfaction with the agreement at the
corporate level as well. Technical information exchange provided
TI with improved devices and manufacturing methods, but the
purchasing agreements prevented IBM from buying freely from
other vendors to encourage lower prices from TI. Citing these
and other problems, Palmer began pressing for a decision to
manufacture all critical components in IBM.
To help the corporation address this and other technical issues,
a corporate-level Research and Development Board was estab­
lished in May 1959. Chaired by McDowell until Mannie Piore took
over in 1960, the board's nine members included McDowell,
Palmer, Piore, and several laborator y managers. Making techno­
logical choices critical to IBM and then "deciding the best way to
implement these choices" was the board's primar y mission.
High on the agenda for the board's monthly meetings were
semiconductor device technologies. Almost immediately the
board assigned an engineer the task of combining several small
projects to establish a companywide program to develop semicon­
ductor circuits and p� ��McUDlii/Products. By the end of
 Commitment and Delivery 281

1959 the program had made many technical decisions and was
projecting 149 employee-years of development work during the
next year.4
It was a critical time for this focus. Technical advisers to V.S.
military organizations were placing increasing emphasis on the
need for smaller and more rugged circuits. Funded by the V.S.
Army's Micromodule Program, for example, RCA had made a
proposal for achieving over two hundred components per cubic
inch. This was more than ten times the density of IBM's SMS
technology. The RCA proposal called for mounting discrete tran­
sistors, resistors, capacitors, and other electric components on
thin ceramic modules a few inches on a side and interconnecting
the circuits among many modules in a three-dimensional array.
The IBM Research and Development Board viewed the pro­
posal as a serious threat. Not only was RCA well established in
electronics, but it had announced its first fully transistorized
computer (the 501) in late 1958 and had installed the first one
a year later. Because it was also known to be developing more
advanced computers,5 IBM could not afford to fall behind RCA
in circuit technologies.
Adding to the sense of urgency were technical advances
throughout the semiconductor industr y. In Februar y 1959 jack
S. Kilby filed a patent describing a way to combine all circuit
elements and their interconnecting wiring on a single silicon
chip. It was the first patent filed for an integrated circuit, as that
term came to be defined. (Prior to the mid-1960s the term
integrated circuit was used to describe any densely packed assembly
of electric components that formed one or more circuits.)
Kilby was not alone in his ideas. Robert N. Noyce, a founder of
the Fairchild Semiconductor Company, was working rapidly to
document his ideas. Five months after Kilby filed for a patent and
his invention was publicly announced, Noyce filed for his patent.
By incorporating advanced processing techniques developed
throughout the industry, Noyce in his patent described inte­
grated-circuit structures similar to those that were later devel­
oped. In particular it made use of planar processing, whereby
photographic images created on the flat surface of a silicon wafer
delineated where chemical processing steps would create con­
ducting lines and more complex device structures.6
These advances toward integrating entire circuits in compact
structures supported Palmer's view that IBM could no longer buy
individual componcntSopvJli�",�Mf8,tn as it had for SMS. The
282 Building IBM

automated manufacturing of SMS had been achieved by design­

ing circuit cards with the manufacturing process in mind. Palmer
believed this concept should now be extended down to the most
basic circuit elements. His views were endorsed by the new direc­
tor of research, Mannie Piore, who established a small task group
to study the issue. An important ally and task-group member was
Mervin J. Kelly, retired chairman of the Bell Telephone Labora­
tories, who had been hired by Tom Watson as a "consultant on
research and engineering matters."7
Within the year the decision was made that IBM would enter
the components business to supply most, if not all, of its compo­
nent needs. A components organization was established in
Poughkeepsie in October 1 960. Several existing groups were
placed in the new organization, and additional people recruited.
In July 196 1 the new organization became the IBM Components
Division. Its manager reported to group executive Vin Learson.8

Developing SLT

Technical leaders in the new Components Division had been

following developments at TI, Fairchild, and elsewhere. In their
judgment, the type of integrated circuits described by Kilby and
Noyce were years away from a cost-effective implementation. As
events unfolded at IBM, there would not be time for that. The
triggering event was approval by the Corporate Management
Committee in March 1961 of the 8000-series plan. Learson was
opposed to the decision, but he had no alternative to offer. In a
first step toward finding an alternative, he asked Bob Evans to
establish an interdivisional task group to study "the feasibility
of using an advanced [circuit] technology in the 8000-series
machines. "9
Chosen by Evans to lead the task group was Erich Bloch, a
native of Germany who had studied at the Swiss Federal Institute
of Technology before immigrating to the United States. Working
days and studying nights, Bloch had obtained a bachelor's degree
in electrical engineering from the University of Buffalo before
joining IBM in 1952. Respected for his practical and thorough
approach to engineering problems, he was an ideal choice. While
serving as engineering manager for Project Stretch, he had
gained firsthand knowledge of the development and use of SMS
circuits and packaging. 10
Copyrighted Material
 Commil1TU!Tl1 and Delivery 283

Figure 19.1. Erich Bloch

Erich Bloch led the efforts that defined and developed SLT. He is shown
here in 1970, reviewing semiconductor manufacturing facilities.

The task force, consisting of Bloch plus eight other highly

qualified engineers, issued their findings within two weeks. 'The
TI approach to microminiaturization using the solid circuit is
deemed to have an economical disadvantage, because of the
specifications that must be achieved simultaneously on many
diverse devices," the report opined. "At the present time, it is
contemplated that one transistor per chip will yield optimum
cost." To keep transistor costs low and reliability high, the "use of
planar, glass passivated chip transistors" was recommended. A key
feature of the report, for Learson, was the conclusion that the
tightest possible development schedule would would add an ad­
ditional year to the development schedule of the proposed 8000
series.lJ
Within three months after Bloch's report was issued, Learson
had terminated the 8000 series, the components group had be­
come an IBM division, and Erich Bloch had been selected to lead
the development of the proposed circuit technology. Early in

Copyrighted Material
284 Building IBM

August Bloch submitted a forty-two-page report titled "Solid Logic

Technology Program-Objectives and Directions. "12 It provided
a remarkably accurate description of the semiconductor devices
soon to be manufactured. Five months later, in Januar y 1 962,
Learson endorsed the SPREAD report as the basis for company­
wide development of a series of compatible processors. Intrinsic
to that plan was SLT.
The SLT chips used with System/360 were little changed in
appearance from those proposed in Bloch's report of August
196 1 . Many novel ideas, patentable and unpatentable, however,
were needed to transform the ideas of 1 961 into the realities of
1964. Among the patentable ideas was a technique for putting a
thin layer of glass over the processed chip surface to protect it
from contaminants during subsequent processing steps and years
of use in customer facilities. Three tiny holes were etched through
this glass layer to permit three small copper spheres to be sol­
dered to conducting lines on the silicon chip. These balls pro­
vided the means for making electrical contacts with the next level
of the circuit package.
That level was called an SLT module. Its fabrication began with
a white ceramic substrate, just under a half inch square. Copper
pins were inserted through twelve holes around the perimeter to
make electrical connections with the next level of packaging.
Conducting lines and resistors were printed on the surface of
the ceramic substrate. The ability to fabricate resistors with bet­
ter precision and lower cost than was then possible on silicon
was a key reason for choosing SLT hybrid integrated-circuit
technology. 13
All processes and device designs were intended to facilitate
economical mass production. The initial plan for manufacturing
SLT modules assumed no human intervention from beginning to
end. Ceramic blanks were to enter the line at one end and exit
from the other-with lines and resistors fabricated, silicon chips
mounted, electrical characteristics tested, capped, and labeled­
all at the rate of thirty modules per minute. The design philoso­
phy was similar to that used in the pioneering manufacturing line
shipped to TI in 1959.
It soon became evident, however, that greater flexibility was
needed. The fabrication of SLT modules required far more proc­
essing steps than existed in the previous line. Rather than inter­
connecting all process steps together in lock-step fashion, it was
decided to make eac Jc �IMf9litfJ8WfGtJR[}fpendently, with its out-
 Commitment and Delivery 285

Figure 19.2. Electronic Circuitry Evolution

Top: An SLT card with six attached modules, as announced with System/360,
is pictured with an SMS circuit card used on 7000 series computers and the
eight-tube pluggable unit introduced on the IBM 701. Bottom: Close-up of
an SLT module, without its protective cap. reveals semiconductor chips (1),
printed lines (2), and printed resistors (3). The tops of twelve pins (for plug­
ging the module into a card) are evident around the module's periphery.

Copyrighted Material
286 Building IBM

put stored in racks and manually moved to the next stage. This
permitted each process unit to be optimized independently, with
several slower units for one process step keeping up with a single
faster unit for another.14
The next level of SLT packaging was a printed circuit card, the
smallest version of which was 1.6 inches square and could hold
up to six modules. The cards in turn were plugged into large
printed circuit boards. Devising the wiring pattern for thousands
of printed lines to interconnect hundreds of circuits on a board
was a time-consuming and error-prone activity. Many different
board types were needed, and engineering changes would be
frequent in the early stages of production.
To automate both the design and manufacturing processes, a
computer program was written for the IBM 7090 that generated
an engineering description of each board type. This description,
which included a detailed wiring layout, was transmitted to an
IBM 1401 that converted it into specific commands that operated
production equipment. It was just one of many pioneering proc­
ess tools available when the manufacture of SLT cards and boards
was moved into the newly completed Process Technology Build­
ing in Endicott in the fall of 1964}5

Memory and Storage

The availability of large amounts of reliable main memory on

each processor was a major selling point. Equally important was
the ability to attach magnetic tape and disk storage via a standard
channel interface. Operating systems were provided to allocate
memory space among various tasks and to provide standard pro­
cedures for moving information between memory and storage
and for operating a variety of other peripheral devices.
The choice of ferrite cores for System/360 main memories was
obvious. No other memory technology offered such good perfor­
mance and reliability, and the company had achieved technical
leadership through Project SAGE. In November 1961, two
months before the SPREAD report was adopted, a task force was
established to seek improved memory designs for low-cost manu­
facture. Members included engineers from development and
manufacturing organizations. Among its accomplishments was
the design of a wiring pattern that reduced unwanted electrical
noise and a manufacturing tool to accomplish the pattern in
production. Copyrighted Material
 Commitrrumt and Delivery 287

An inescapable requirement for control stores was speed: they

had to supply control words more rapidly than was possible with
conventional ferrite-core memories. The solution was to take
advantage of the read-only characteristic of control stores. By
representing each logical 1 or 0 by the presence or absence of a
tiny electrical element, such as a capacitor or inductor, both the
time required for a write-back cycle and the cost of write circuits
were eliminated. Information could also be read out more rap­
idly. Corporate responsibility for control store designs was as­
signed to the laborator y in Hursley, England, where pioneering
design work had been done.
By the fall of 1963 the development of main memories and
control stores for the smaller processors was progressing reason­
ably well, but there were severe problems in satisfying the needs
of higher-performance processors. To solve these problems, Erich
Bloch was given responsibility for the development of computer
memories in addition to SLT. By the end of the year, he had
authorized several significant changes. A novel "load-sharing
matrix switch" was implemented to reduce the current levels
required in memory drive circuits. Special SLT circuits were de­
signed for memory use. And a ferrite core with a 30 percent
smaller hole diameter was adopted after an engineer, working on
his own initiative, demonstrated that improved manufacturing
methods would be able to wire the smaller cores into memory
arrays.16
The redesigned memories achieved read-write cycles even
shorter than the targeted 1.0 microsecond. When ways were
found to speed up the related control store and logic circuitry as
well, processor Models 60, 62, and 70 were withdrawn in April
1965 and replaced by higher performance Models 65 and 75 with
0.75-microsecond cycle memories. Around-the-clock work by en­
gineers made it possible to ship the first Model 65 in November
1965, only three months later than originally scheduled for the
earlier models.17
Also important to the success of System/360 was the availability
of improved magnetic disk storage. The most important technical
advance since the 1955 announcement of RAMAC was the self­
acting air bearing. Rather than using a jet of forced air to float
the read-write head above the disk surface, use was made of a
dense layer of air that formed naturally between the rapidly
moving disk surface and a properly shaped head structure. An
t
important improvemgeJ:> J:lgM€8ljt1alW18/was was achieved by re-
288 Building IBM

placing the single actuator arm (having two read-write heads)

with enough arms ganged together to provide a dedicated head
for each disk surface.
Both of these advances were introduced with the IBM 1301 Disk
Storage in 1961. Compared with RAMAC, it provided a thirteen­
fold increase in storage density and three times faster average
access to information. An improved version of the 1301, with
another fourfold increase in storage density, was announced in
September 1963 as the 1302. With little change other than to the
attachment interface, it was announced seven months later with
the rest of System/360 as the IBM 2302 Disk Storage.IS
Also available with System/360 was a disk storage system with
interchangeable disk packs. The primar y motivation for this in­
novation was to reduce information storage costs by making it
possible to store information on disks off-line. System designers,
however, also wanted to be able to write information on disks with
one system and then read the stored information with another. This
required much tighter control of mechanical tolerances in a tech­
nology in which tolerance control was already a severe problem.
The first product to provide interchangeability of disk packs
was the IBM 1311 Disk Storage Drive, announced in October
1962. Each pack contained six 14-inch-diameter disks in a 4-inch­
high stack that weighed ten pounds and held up to 2 million
characters. An average access time of 150 milliseconds to infor­
mation was achieved. The IBM 1440 computer, on which this
storage was first offered, did poorly in the marketplace because
it was initially offered without the ability to attach magnetic tape
units as well.
This mistake was not repeated with System/360. The two types
of storage media coexisted on that system for many years, with
magnetic tape providing shelf storage at a lower cost than mag­
netic disks. An improved version of the interchangeable disk
storage, having twice the capacity and half the access time, was
announced with the rest of System/360 in April 1964 as the IBM
2311 Disk Storage Drive.19
The initial System/360 development plan called for no im­
provement of tape storage products. There were several reasons
for this decision: the belief that tape storage would soon be
displaced by disk storage, IBM's presumably secure leadership in
tape-drive products, and the need to conser ve engineering re­
sources in Poughkeepsie. To handle the planned 8-bit byte that
was to replace the tbgpy�J6-bit character, a "tape
 Commitment and Delivery 289

adapter unit" was proposed. It would convert three 8-bit bytes

from the channel into four 6-bit characters for the tape unit, and
vice versa, with logic to create appropriate parity bits.
The efforts of a few self-motivated engineers in manufacturing
changed this plan in late 1963. They built a head capable of
reading and writing nine tracks on conventional half-inch tape
and demonstrated that it could be manufactured at reasonable
cost. A variety of other improvements were added during product
testing in 1964 to meet perceived competitive threats. especially
from Honeywell and Sperry Rand. Among these improvements
was a read-backward capability that made it possible to sort re­
cords more rapidly. Long offered by UNIVAC, this capability
previously had been partially compensated for by the superior
stop-start and high-speed rewind capabilities of IBM's tape units.
Now IBM provided both features.2o
Again and again during the development of System/360, engi­
neers provided unanticipated solutions for critical problems.
Ralph Palmer had motivated this type of activity by challenging
more than one engineering group to create "the best solution"
for a specific problem and by allowing engineers to spend part
time pursuing their own ideas. Many engineers labored long
hours on their own time, seeking ways to make better products.
The competition between engineers, which Watson, Sr., had
used so effectively in the past, was a key element in the success
of System/360. There was, however, an important difference.
Competition was no longer "blind." Now information on engi­
neering plans, assignments, and requirements was broadly avail­
able. Individual engineers could readily learn what approaches
were being tried and what opportunities there might be for
improvement.

Achieving Manufacturing Goals

Orders for System/360 computers quickly exceeded the forecast.

More than one thousand orders were received in the first month
after announcement, and another one thousand were received
during the next four months. Adding to production require­
ments. most orders requested more memory and storage capacity
than IBM's product planners had anticipated.21
Construction of a 100,OOO-square-foot manufacturing plant for
SLTwas begun in September 1962. Located on a newly purchased
450-acre site in Ea¢o';'�d �BI twelve miles south of
290 Building IBM

Poughkeepsie), the plant was expanded to three and one-half

times its initial size within three years.22 The number of SLT
modules manufactured in East Fishkill in 1963 was 0.5 million. A
twelvefold production increase in 1964 resulted in 6 million mod­
ules, and plans called for the production of 28 million modules
in 1965.
Asked to increase production by yet another factor of two in
1965, the plant manager explained why it could not be done
without excessive risk. Ordered to double production anyway, he
refused and was replaced by a manager willing to accept the
higher commitment. The result was a near disaster. Routine qual­
ity control activities during 1965 revealed poor quality in many
modules. By September more than 25 percent of the year's pro­
duction had been impounded by the Quality Control Depart­
ment, and a decision was made to stop production.
Scientists and engineers from research and development were
brought in to help solve the problem. Within a month it was
discovered that the higher-volume production equipment caused
an undesirable chrome oxide to form when a chrome layer was
being deposited on the wafers. Once the problem was under­
stood, it was easily corrected. Production was resumed at the
higher level, but IBM had been forced to announce an embar­
rassing two- to four-month delay in product shipments.23
The production of SLT modules reached 36 million in 1965
despite the near disaster, and 90 million were produced in 1966.
By then IBM's East Fishkill plant was producing more semicon­
ductor devices than were produced by all other companies in the
world combined. Meanwhile production in newly established IBM
plants in Burlington, Vermont, and Essonnes, France, had been
ramped up to 27 and 26 million modules, respectively.24
So frantic was the development of high-volume production
equipment for semiconductor devices that little thought was given
to protecting proprietar y designs. Product innovations were rou­
tinely covered by patents, but production equipment was not. As
a result much of the production equipment designed by IBM
found its way into the general market when vendors, hired to
build equipment for IBM, broadened their business by selling
essentially the same equipment to others.25
The problems encountered in manufacturing ferrite-core
memories were less severe because of years of experience. The
production of ferrite cores, for example, reached one billion per
year in 1963, and t��yWghtlW:MaWHMulty when the rate was
 Commitment and Delivery 291

increased tenfold during the next four years. The problems of

wiring the cores into memory arrays were not so easily solved. So
many wire feeders were needed to wire cores into arrays that a
department was created in Kingston just to produce them. As
quickly as wire feeders were built, they were put on the manufac­
turing floor. By the end of 1964 there was no more floor space
for wire feeders and there were insufficient people to operate
them.
A plant was built in Boulder, Colorado, in mid-1965 to alleviate
this problem as well as a similar problem in manufacturing tape
drives. When this action proved inadequate, an experiment was
undertaken to have people in Japan wire core frames by hand.
So careful and meticulous were the Japanese workers, and so low
were their wages, that memor y arrays wired by hand inJapan had
the same quality and a lower cost than those produced with
automated wire-feed equipment in Kingston.
Core wiring operations were quickly expanded throughout the
Orient into what became known as the "Hong Kong Core House. "
Competitors also established core wiring facilities in the Orient,
causing IBM to lose much of the advantage it had gained through
automated manufacture. Its memor y designs did, hovever, con­
tinue to have fewer semiconductor circuits and lower final assem­
bly costs than those of its competitors.26

Stress at the Top

The logistical problems of manufacturing System/360 compo­

nents and products in plants throughout the world were enor­
mous and so was the stress on managers. This can be appreciated
best by viewing the scene at the top. To ensure a successful
introduction of the new product line, Tom Watson promoted Vin
Learson to senior vice president with responsibility for sales. The
sales results attest to his success. Watson also promoted his
brother to senior vice president with responsibility for research,
development, and manufacturing. In preparation for this assign­
ment, Dick Watson had left his World Trade Corporation position
and spent half a year as head of the Corporate Staff. Now he and
Learson had the two top positions under Tom Watson.27
Dick Watson's assignment was intended to prepare him to
become the company president at some future date, but its effect
was quite different. W ith little relevant experience, he was unable
to cope with all the �cf18IMlfacturing and shipping
292 Building IBM

System/360. He also had to contend with rising fears of his

brother that SLT would rapidly become obsolete because of inte­
grated circuits and that competitive product announcements
were undercutting System/360 in the marketplace. Responding
to these fears, Learson demanded that new product features be
added immediately. even though there were already more prob­
lems than could be handled by the engineering groups for which
Dick Watson was then responsible. 'The friction between Dick
and Yin got completely out of hand," Tom Watson said.28 Rather
than creating two complementary positions, Tom Watson had put
his brother and Learson into a competition-and Learson was
not one who lost easily.
Once the decision was made to announce product shipment
delays in October 1964, Tom Watson felt he had no choice but
to turn once again to Learson for help. W ithout formal an­
nouncement, Learson returned from sales to replace Dick Watson
in running the development and manufacturing organizations.
Choosing the head of the corporate manufacturing staff as his
top assistant, Learson drafted four senior engineering managers
and gave each of them responsibility for operations at selected
laboratories and plants. These "four horsemen," as they came to
be called, were given companywide authority to decide how Sys­
tem/360 commitments would be met. All other activities had
lower priority. Citing the "myriad of typical early manufacturing
problems on every box," one of the four horsemen said the
situation was "an absolute nightmare." He recalled his own in­
volvement as "a gray blur of twenty-hour days, seven days a week­
never being home." But in less than five months, Learson's team
had System/360 shipments back on schedule.29
In January 1966 Tom Watson advised the board of directors
that Al Williams would step down as president at the end of
March. Learson was elected president, effective the first of April.
Dick Watson was elected to a newly created position of vice
chairman of the board. The traumatic events of System/360 that
propelled Learson to the presidency had effectively ended Dick
Watson's IBM career. In 1970 he undertook a new assignment as
the United States ambassador to France.30

Software Support

Of all the challenges presented by the System/360 an­

nouncement, none w �oWfrfgh?Jm��H�fccomplish than provid-
 Commitment and Delivery 293

ing the promised software support. Its importance was never in

doubt. General purpose programs, provided "free of charge" by
manufacturers, were considered essential to the marketing of
computers . 31
Most comp uter processing was then done in a stacked-job mode
in which one program was fully processed before another one was
started . Interest was growing, however, in a multiprogramming
mode in which two or more programs were processed more or
less simultaneously. Multiprogramming could improve the overall
efficiency of a system by causing the processor to work on one
program while waiting for information (from tape or disk stor­
age) needed by another.
Interest was also growing in applications in which people could
interact with a co mputer using terminals at remote locations. This
capability had been demonstrated in a few systems. For example ,
the SABRE airline reservation system, jointly developed by Ameri­
can Airlines and IBM, offered extensive access via remote termi­
nals. But providing such capability for a single system, dedicated
as SABRE was to a specific task, was far simpler than providing it
as part of a general-purpose operating system. Furthermore each
processor of the new line offered many configurations of main
memor y, magnetic tape and disk storage, and other input-output
equipment that had to be supported by the operating system.
Just what type of programming support should be offered was
heavily debated following adoption of the SPREAD report in
Januar y 1962. It was readily agreed that the smaller processors in
the line, especially those ordered with minimum memor y capac­
ity, could be given only a rudimentar y stackedjob capability. The
type of scaled-up support that should (and could) be offered on
intermediate to large systems was more difficult to j u dge, how­
ever, because of limited experience with the newer programming
concepts.32
In January 1963, by which time four levels of programming
support had been defined, Fred Brooks established the firm
guideline that any application program compiled or assembled
at any level must be able to run at any higher level as well. "I
should like to be notified at once in the event of any proposed
amendment to, or departure from , these ground rules," Brooks
admonished the responsible managers.33 The constraint was es­
sential to full realization of the benefits of a compatible line, bu t
its immediate impact was that no part of an operating system for
any one of the processors could be �de.fin ed without considering
c.;opyngntea MaLenal
294 Building IBM

its relationship with operating systems for the others. Decision

making was impeded.
As 1963 drew to a close, programming activities were in suffi­
cient disarray that Bob Evans requested a technical audit. While
highly critical of the "lack of management decision making," the
audit report placed some of the blame on the heavy burdens of
programming groups that also had to develop software support
for the "temporizing" products. Another obstacle had been Haan­
stra's use of programmers in Endicott to support his proposed
1401-compatible processor rather than the new product line. Of
the report's many recommendations, the most important was that
a design task force be established to distill months of deliberation
into a new set of programming objectives.34
Less than one month after the design task force was established,
it had hammered out decisions that became the basis for OS/360,
the new product line's operating system. Rejecting the plan for a
graded set of compatible operating systems, the task force recom­
mended development of a single, full-function operating system,
suitable for the largest processor. Its control program was to be
comprised of parts that could be selectively omitted to allow for
a variety of reduced-function versions for use on processors with
smaller memory capacities, down to as little as 16K bytes.35
The recommendations of the design task force became the
basis for OS/360 as announced with System/360 hardware in
April 1 964. A stackedjob control-program version of OS/360 was
to be available by the end of 1965 and a multiprogramming
version was promised for delivery six months later. Also an­
nounced were a variety of "special" programming support pack­
ages for small systems with as little as 8K bytes of memor y.36
To help ensure success of the programming effort, Fred Brooks
offered to supervise it much as he had coordinated the engineer­
ing designs of the processors. His offer was eagerly accepted even
though it was understood that he planned to leave IBM in the
fall to establish a computer science department at the University
of North Carolina.37
The magnitude of the task of developing the proposed operat­
ing system was grossly underestimated. Over one thousand people
(including managers, programmers, manual writers, and those in
all support functions) were employed during the peak year. More
money was spent to develop System/360 software support during
that one year than had been planned for the entire project. At

Copyrighted Material
 Commitment and Delivery 295

the personal request of Tom Watson, Brooks agreed to remain in

charge of the effort for an extra year, while spending one week
per month at the University of North Carolina.38
Despite a sustained intensive effort, the various parts of OS/360
were delivered months later than announced; they generally used
more memory capacity than planned; and their processing times
were longer than desired. Improving the performance of critical
parts of OS/360 was a key objective of subsequent releases.
Brooks attributes the lack of elegance of OS/360 to inadequate
vision by its designers. to the difficulty of achieving conceptual
integrity in so vast a project, and to the need to subdivide the
implementation among so many groups. Concerning his own
inability to get the system completed on time despite the use of
vast resources, he wryly proclaimed what he called Brooks's
Law: "Adding manpower to a late software project makes it
later. "39
The inability to achieve satisfactor y performance of OS/360 on
processors with less than 32K bytes of memor y was particularly
serious. This deficiency was not recognized until several months
after the April 1964 announcement of System/360. During the
intervening months, customers who ordered the System/360
Model 30 with 1 6K bytes of memory had been assured that they
could use the new operating system. The failure to achieve this
objective-or perhaps the error in committing to it-resulted
from the limited experience of programmers in Poughkeepsie
with small systems. In part because of Haanstra's resistance to the
new line, the views of the small systems designers in Endicott had
not received sufficient consideration.
The situation was so serious in the fall of 1964 that the General
Products Division (headquartered in Endicott) was given respon­
sibility for all support required by System/360 processors with 8K
to 16K bytes of memory. It was a daunting responsibility. Hard­
ware deliver y schedules demanded that all of this support be
available within a year. On the last day of 1964 a firm commitment
was made, and IBM formally announced three additional operat­
ing systems, all of which were to be available by the end of 1965.40
The names subsequently given to them were BOS (Basic Oper­
ating System), TOS (Tape Operating System), and DOS (Disk
Operating System) . The minimum storage and memor y require­
ments for BOS were disk plus 8K bytes of memor y; for TOS, tape
plus 16K bytes of memory; and for DOS, disk plus 16K bytes of
Copyrighted Material
296 Building IBM

memor y. The special programming support offered for systems

with only 8K bytes of memory became known as BPS (Basic
Programming Support).
Long hours of dedicated work by programmers in Endicott and
San Jose produced this programming support more or less on
schedule-except for DOS, which was finally released in June
1966. Despite the limited time available to define and complete
DOS, and despite its late delivery, it became the most widely used
operating system in the world. The quality of its conception and
implementation contributed to the increasing popularity of disk
storage. A multiprogramming version was subsequently released
for systems with at least 32K bytes of memor y. Some twenty-six
releases of DOS, each with enhanced function or performance,
had been issued by IBM by the time a decision was made in 1971
to create no more new versionsY

Market Impact

When System/360 was announced in 1964, the major companies

of the computer industr y were often referred to as Snow White
and the Seven Dwarfs. Snow White was IBM. The Seven Dwarfs
were Burroughs, Control Data, General Electric, Honeywell,
NCR, RCA, and Sperry Rand. Of the estimated $10 billion world­
wide inventory of installed computers in 1964, the Dwarfs had
produced about 30 percent, and IBM had produced the rest. Five
years later IBM's worldwide inventor y had increased more than
threefold to $24 billion and that of the Dwarfs had increased by
about the same ratio to $9 billion.42
This apparent lack of change, other than dramatic growth, is
misleading. The popularity of System/360 made it difficult for
others to compete across the broad spectrum of the general
purpose computer market. Two of the Dwarfs that attempted to
do so failed to make a profit and dropped out of the business in
the early 1970s. At the same time, weaknesses of System/360
provided lucrative opportunities in certain parts of the market,
and compatibility among processors and modularity of its com­
ponent parts created attractive opportunities for companies to
provide plug-compatible products for attachment to the system.
Among the Seven Dwarfs, RCA pursued the most aggressive
and risky strategy. In December 1964 it announced the Spectra
70 product line with four processors covering almost as wide a
Copyrighted Material
 Commitment and Delivery 297

performance range as System/360. Not only were the four pro­

cessors compatible with each other, they were designed to be
compatible with System/360 as well-a feat presumably accom­
plished through careful study of IBM's publicly available docu­
mentation. Spectra 70 customers would thus be able to use
software written for System/360.43
Because price was the only obvious differentiation between the
two competing systems, RCA planned to offer a 1 5 to 20 percent
price-performance advantage over IBM equipment. Customer re­
sponse was reasonably good, and RCA's computer revenues dur­
ing the 1960s increased fifteenfold to $211 million. This was only
3 percent as large as IBM's revenues, however, and represented
a ver y small part of RCA's total revenues as well. In an attempt
to preempt IBM's System/370 announcement, RCA announced
its RCA series, a price-reduced version of Spectra 70. The result
was disastrous. Rather than displacing IBM equipment, the new
series primarily displaced RCA's own Spectra 70 computers. This
event, coupled with accumulating losses, persuaded RCA to exit
the business in 1971 by selling its computer division to Sperr y
Rand.44
General Electric (GE) also moved aggressively in the 1960s to
capture a larger share of the rapidly growing computer market.
Like Honeywell, its primar y strategy was to facilitate the migration
to its systems by users of IBM's older computers. General Electric
did particularly well with systems for the banking industry, and it
expanded in Europe through acquisitions of Machines Bull and
the Olivetti computer division. But it had difficulty operating at
a profit. Concluding the cost of achieving profitability was too
high, GE sold most of its computer business to Honeywell in
1970.45 With GE and RCA out of the general purpose computer
business, five of the Seven Dwarfs remained. These were dubbed
the BUNCH: Burroughs, the Univac Division of Sperr y Rand,
NCR, Control Data, and Honeywell.
The Control Data Corporation (CDC) was one of the more
successful companies in the computer industr y. Founded by Wil­
liam Norris with other former Sperry Rand employees in 1957,
its computers accounted for nearly 5 percent of the value of all
computers installed worldwide during the late 1960s and early
1970s.46 The company's success was based on building computers
for scientific and engineering computations. Behind that success
was Seymour Cray-a computer designer who became a legend.
Copyrighted Material
298 Building IBM

A key event was CDC's 1960 proposal to build a Stretch-class

supercomputer for the Lawrence Radiation Laboratory. In De­
cember 1964 this supercomputer was announced as the midrange
computer in the CDC 6000 series, which was to consist of the
6400, 6600, and 6800 computers.
Unwilling to be bested by this tiny company, Tom Watson
exhorted his engineers to use the most advanced technologies in
the yet-to-be announced System/360 Model 90. The sales organi­
zation was similarly exhorted not to lose sales to CDC. Burdened
by requirements of compatibility with System/360 and by exces­
sively high development costs, the Model 90 (of which there were
several versions) proved no match for the more elegant creations
of Seymour Cray.47
Fearful of the wrath of Watson should they fail, IBM's market­
ing organizations used unusually aggressive tactics to sell Model
90s. Alleging that these tactics had been predatory and intended
to monopolize the market, the Control Data Corporation filed an
antitrust suit in December 1968 against IBM, which viewed the
case as risky, given the vagaries of the antitrust law, and settled it
before it came to trial. The terms included the sale to CDC of
IBM's Service Bureau Corporation at its book value of $16 mil­
lion-probahly only a third of its market value. Other features of
the settlement provided CDC with cash and contracts worth $ 101
million.48
At the same time that CDC was exploiting the weakness of
System/360 at the top of the line, the Digital Equipment Corpo­
ration (DEC) was exploiting opportunities below the bottom of
the new line. Kenneth H. Olsen, who founded DEC in 1957, had
worked closely with Jay Forrester on Project SAGE at MIT. While
assisting IBM engineers with Project SAGE in Poughkeepsie, Ol­
sen had learned a lot about the company's manufacturing and
management methods. He began his own business quite modestly
by making circuit modules for other organizations. By 1960 these
modules had become the building blocks for the company's first
"programmed data processor," or PDP-I.
As semiconductor circuit technology improved, DEC was able
to provide greater function at lower cost. The PDP-8, introduced
in 1965 at a price of only $18,000, was ver y successful. Stimulated
by the then-popular miniskirt, someone called the PDP-8 a "mini­
computer," and the name stuck. The company's minicomputers
were initially used in scientific apparatus to collect data and to
Copyrighted Material
 Commitment and Delivery 299

control experiments. Later uses were in process control and

communication networks. Gradually minicomputers took their
place as general purpose computers. From 1 963 to 1 969 DEC's
worldwide revenues increased ninefold to $9 1 million. In 1 980 its
revenues reached $2.4 billion, which was almost 10 percent as
large as IBM's $26 billion.49
Pressure from the BUNCH and from rapidly rising systems
suppliers such as DEC, Scientific Data Systems, and Data General
increased as these companies positioned themselves to exploit
weaknesses of System/360. But it was price competition from
plug-compatible manufacturers of peripheral products and from
computer leasing companies that gave IBM its greatest challenge.
The Telex Corporation was the first company to provide plug­
compatible products for System/360. Already in the business of
manufacturing tape drives, Telex offered its first System/360 tape
drives in July 1967. Sold at a 50 percent price discount over
equivalent IBM equipment, more than three thousand Telex tape
drives had been installed on System/360 computers by the end
of 1 970. The second company to enter the business was Memorex.
Its opportunity was provided by IBM's inability to manufacture
the newly invented removable disk packs rapidly enough to satisfy
customer demand. Already a leading supplier of high-quality
magnetic tape for computers, Memorex began manufacturing the
more profitable disk packs in 1966. In mid-1 968 it began shipping
plug-compatible disk drives as weI1.50
Many IBM employees were lured into the plug-compatible busi­
ness by companies willing to pay a premium for their knowledge
of IBM technology and product plans. Others chose to profit
more directly by establishing their own companies. For example,
Information Storage Systems was founded by twelve IBM employ­
ees in San Jose who resigned en masse in December 1967, and
the Storage Technology Corporation was founded by four IBM
employees in Boulder who resigned in July 1 969. By the end of
1970 the companies supplying plug-compatible products had pro­
duced considerably less than 1 0 percent of the tape drives and
only 3 percent of the disk drives attached to System/360 comput­
ers, but these percentages were growing rapidlyY
Among the noteworthy IBM alumni was Alan F. Shugart. Join­
ing the Memorex Corporation as vice president of product devel­
opment in 1 969, he persuaded more than one hundred IBM
employees to join him over the years in keeping Memorex's
Copyrighted Material
300 Building IBM

storage products technologically abreast of IBM's. Shugart sub­

sequently founded two successful storage-product companies:
Shugart Associates in 1 973 and Seagate Technology in 1978.52
The existence of computer leasing companies added to IBM's
competitive problems. A typical mode of operation was to offer
to buy installed System/360 equipment from IBM and lease it
back to the user at a lower rental than was being charged by IBM.
To show a profit, leasing companies typically assumed a five- to
ten-year life for leased equipment, in contrast to the four-year life
IBM assumed. If IBM's estimate proved to be more correct, the
leasing companies would be in serious financial difficulty in little
more than four years-as indeed many were. Leasing companies
also provided a ready market for the lower-priced plug-compatible
peripheral products. By early 1968 IBM had come to regard
leasing companies collectively as potentially its biggest domestic
competitor.53
If IBM responded to this competition by cutting its own prices,
it would be the primary loser because of its larger market share.
But if it failed to respond, the leasing companies and plug-com­
patible manufacturers would continue to gain market share. The
company chose a mixed response-an important element of
which was to outrun the competition with technological improve­
ments. Although often impeded by the loss of key employees or
by outright theft of proprietary information, the strategy never­
theless helped IBM maintain its high profit levels and was a
driving force behind dramatic improvements in computer cost­
performance.

Copyrighted Material
20
Onrush of Technology

By the end of the 1960s, the dramatic success of System/360 had

realigned the computer industr y and had virtually assured IBM's
dominance for years to come. The modern IBM Corporation had
been built.
The introduction of System/360 had been so fraught with
difficulties, however, that Tom Watson and his engineering man­
agers had little time to celebrate its successes. There had been
delays in shipping hardware and software, the performance of
the software was below expectations, and competitive an­
nouncements threatened the company's installed computers and
raised doubts about its new product line.
Nothing caused greater concern than the negative reception
accorded SLT. Government agencies that were funding the devel­
opment of monolithic integrated circuits were quick to belittle SLT.
Monolithic integrated circuits had one or more circuits fabricated
on a single silicon chip, whereas SLT used several chips to make
one circuit. Component vendors, whose sales were diminished by
IBM's decision to manufacture its own components, were also
critical. Gleefully reported by the trade press, these denigrations
of SLT unnerved many of IBM's executives.
Concerns that SLT would not be competitive proved to be
unfounded. Several years of research and development by many
companies would be required before monolithic integrated cir­
cuits became as cost effective as SLT. In the long run, however,
the concern was valid. Dramatic progress in achieving ever lower
costs with integrated circuits was a dominant factor behind dec­
ades of growth of the computer industr y. Continued success for
IBM would depend on its ability to remain at the leading edge of
technology-and to adjust its products and marketing methods
as new business opportunities were created and old ones were
diminished. Copyrighted Material
302 Building IBM

Going after Monolithics

A year before the first System/360 computers were shipped, John

Haanstra had already concluded that progress in monolithic in­
tegrated circuits would rapidly make obsolete the company's new
computers. Ousted as president of the General Products Division
for opposing System/360, Haanstra saw integrated circuits as an
opportunity to regain a leadership position. In a report to man­
agement issued in September 1964, he asserted that IBM was
more than two years behind TI, Motorola, and Fairchild. He
urged an aggressive program to catch up and get ahead. At a time
when the leading semiconductor companies were struggling to
fabricate fewer than five circuits on a chip, Haanstra boldly set a
goal of ten to one hundred.1
Pressured by his brother to "make sure the 360 wasn't going to
be obsolete before it was even delivered," Dick Watson ordered
the technology development divisions to begin working together
"to design the next generation machine." Monolithic integrated
circuits and other advanced technologies were to be used.2 To
pursue this effort, while at the same time introducing Sys­
tem/360, Dick Watson restructured the company's development
and manufacturing organizations in January 1965. The Compo­
nents Division was abolished; its manufacturing activities were
incorporated in a new Systems Manufacturing Division and its
development activities, in a new Systems Development Division
(SDD). The manufacturing division was to concentrate on pro­
ducing System/360 while SDD was to create the next generation
of machines.3
John Haanstra was appointed president of SDD. System and
component development had been combined in one division
because Haanstra and others believed the ability to put many
circuits on a single chip required that systems and components
be designed together. Haanstra envisioned the next generation
machines to be at least as great a leap forward as System/360 was
from its predecessors. To accomplish this, the new machines were
to have an entirely new operating system as well.
A firm believer in the "needs-capability" management concept,
Haanstra asked his system designers to specify the component
needs and the component developers to provide the capability.
To minimize the difficulty of achieving their own objectives, the
system designers set very aggressive goals for new components.
When the hapless c<m}�c:PNtiRJn.JlIFrs complained that the
 Onrush of Technology 303

goals set for them could not be achieved, Haanstra was uncon­
vinced. ''You are telling me what you are capable of rather than
responding to the needs," he admonished them. Managers who
expressed reluctance to accept the goals were advised that others
would be happy to have their jobs.4
Semiconductor technology groups were additionally under
pressure to help solve problems in manufacturing components
for System/360. When failure to solve manufacturing problems
fast enough forced Tom Watson to announce a two- to four­
month delay in product shipments in October 1965, he took
quick action. He relieved his brother of responsibility for devel­
opment and manufacturing and put Yin Learson in charge. By
the time Learson had gotten System/360 shipments back on
schedule, Haanstra had been removed as president of SDD. The
effort to achieve his unrealistic product development objectives
had contributed to the difficulties in getting System/360 products
manufactured and shipped on time.5
Operating under more realistic guidelines, the development
groups created integrated circuits that fell far short of the objec­
tives originally set by Haanstra. The circuits were, nevertheless,
equivalent to the best available from outside the company. First
used on the System/360 Model 85 announced in January 1968,
the IBM integrated circuits had one to four circuits on each
silicon chip. Up to four chips were mounted on a half-inch-square
ceramic module similar to an SLT module. The copper balls that
electrically coupled SLT chips to printed lines on ceramic mod­
ules had been replaced by a proprietary solder-ball technology
that allowed for higher circuit densities.6
The most significant technological advance on the Model 85
was introduction of the first cache. This was a small high-speed
buffer (made of monolithic semiconductor memory elements)
that was coupled to the main memory in such a manner that the
combination achieved almost the high speed of semiconductor
devices with the low cost of ferrite-core memories. Whenever a
word was requested from memory during the execution of a
program, a block of many words (containing the requested word)
was moved into the high-speed cache.
For a well-designed cache and memory combination, the fre­
quency with which needed words were found to be already in the
high-speed cache was typically greater than 95 percent-resulting
in an effective memory speed nearly as fast as the cache. The use
of a cache to impr�opJWgl.tte�narmance soon became a
304 Building IBM

Figure 20.1. System/360 Model 85

Announced in January 1968, the Model 85 was the first computer equipped
with a cache. It also had IBM's first processor to be built entirely with mono­
lithic integrated circuits. The configuration pictured here shows the proces­
sor, in which the cache is located (center). Surrounding it (clockwise from
lower left) are data channels , tape drives, printers and control units, disk
drives, card input-output units, and main memory.

standard feature of almost all computers. Even after semiconduc­

tor devices replaced ferrite cores as the primary memory ele­
ments, computers typically featured memory structures in which
a small high-speed cache enhanced the performance of a lower­
cost main memory.7

System/370

The first two of the next-generation computers were announced

in June 1970 as the System/370 Models 155 and 165. The desig­
nation 370 was chosen to suggest it was a system for the decade
of the 1970s. Rather than being the dramatic leap forward envi­
sioned by Haanstra, System/370 was an evolutionary extension of
System/360. The use of monolithic integrated circuits and a
cache nevertheless gave the processors a fourfold improvement
Copyrighted Material
 Onrush of Technology 305

in performance over the models they replaced. The reliability of

the circuits was ten times better than SLT, which was itself nearly
one hundred times more reliable than the SMS it had replaced.
Crucial to the success of the new systems was a two-times faster
(2,000 lines per minute) printer and a disk storage unit that
offered twice the access speed and three times the storage capac­
ity per disk pack. With eight disk packs per unit, the new IBM
3330 Disk Storage, announced with System/370 in June 1970,
offered 30-millisecond average access time to 800 megabytes of
on-line storage.8
An important architectural change was the addition of 6 in­
structions to the 143 available in System/360. The new instruc­
tions provided better performance in certain operations. Their
greater significance, however, was from a marketing perspective.
They would hasten the obsolescence of the older systems. Sys­
tem/360 processors would not be able to execute System/370
programs that incorporated any of the new instructions, whereas
the new processors would be able to execute any old or new
program written for System/360.9
The two technological advances of System/370 that actually did
the most to make earlier computers obsolete were not initially
available. The first of these was the industry's first all-semiconduc­
tor main memory. Its introduction on the System/370 Model 145
in September 1970 signaled the beginning of years of IBM lead­
ership in semiconductor memories. This leadership was even
more important to IBM's success than had been its earlier lead­
ership in ferrite-core memories. Haanstra's early emphasis on
developing semiconductor memories during his brief tenure as
president of SDD deserves considerable credit. Another impor­
tant factor was invention of the cache. This gave IBM practical
experience in manufacturing semiconductor memory devices
(for the cache) long before the low costs needed for main memo­
ries could have been achieved. 10
The second major advance was "dynamic relocation." This per­
mitted programs to be written with "virtual addresses" that were
converted to real memory addresses at execution time. A trans­
lator built into the system, with a combination of hardware and
software, converted the addresses automatically. As described at
the time, this provided a ''virtual memory that, as far as the
programmer is concerned, is directly addressable as if it were
main memory. The size of this virtual memory is limited only by
the address field an�cJ/>% JlNemlliHmn economics."ll If the
306 Building IBM

Figure 20.2. Semiconductor Memory

The two-level ceramic module holds four semiconductor memory chips.
Each chip contains 128 bits of memory. These modules were used in the
industry's first all-semiconductor main memory, announced on the
System/370 Model 145 computer in September 1970. Twenty years later
IBM announced its RISe System/6000 workstations with semiconductor
memory chips, having a capacity of 4 million bits each.

virtual memory size required by a program exceeded the amount

of real memory. the system automatically moved "pages" of infor­
mation between disk storage and main memory to satisfy the
needs of the program.
The concept of enhancing the apparent size of main memory
by "paging" information automatically between it and a slower
storage device was not new with System/370. The first practical
implementation of this was on the Atlas computer, developed by
Manchester University and Ferranti Limited in England. First
operational in a customer's facility in late 1962, Atlas automat-

Copyrighted Material
 Onrush of Technology 307

ically paged information between a ferrite-core memory and a

larger-capacity rotating magnetic drum storage device.12
Although IBM system designers were well aware of these tech­
niques, they chose not to implement them in System/360 because
there was litde information on the performance of paging systems
on general applications. But when MIT and the Bell Telephone
Laboratories began working with General Electric to modify one
of its computers to provide this capability for terminal-oriented
applications, Learson insisted that IBM also provide it. As a result,
the System/360 Model 67 was developed. Delivered to customers
beginning nine months after it was announced in August 1965,
the Model 67 provided automatic paging between main memory
and a magnetic disk storage device. Performance data from
Model 67s, handling customer applications, was used in the de­
sign of dynamic relocation techniques that implemented virtual
memory on System/370.
Bob Evans's role in putting virtual memory on System/370
computers-after the first two were announced-is revealing of
the management processes and turmoil in IBM during the 1960s.
Evans says he was removed from System/360 leadership in Janu­
ar y 1965 primarily because of his failure to offer this capability.
An era of conservatism then swept through the development
organization following Haanstra's removal as president of SOD.
Virtual memory was once again judged to be too speculative for
use in mainline products. When Evans returned to the company's
mainline activities as president of SOD in October 1969, he "de­
cided not to be punished twice for the same mistake" and imme­
diately reinstated virtual memor y in the product plans for
System/370.13

The FS Failure

The announcement of System/370 in June 1970 coincided with

an industrywide business decline. Orders failed to achieve projec­
tions. Concern was growing in IBM and the rest of the industr y
that the financial decline was indicative of more serious long-term
problems. In particular, the cost of computer hardware was drop­
ping so rapidly that new customers and applications would have
to be found at an alarming rate just to maintain the same dollar
value of sales each year.
The situation looked particularly grim in IBM because its tech­
nologists were overly 6pp�e<1\\faterfature achievements. With
308 Building IBM

the first all-semiconductor main memory about to be announced,

the research and development groups were projecting the cost
per bit would drop thirtyfold during the next five years, and
another thirtyfold by the end of the decade. By these projections
IBM would have to install 900 times as much computer memory
capacity in 1980 as it did in 1970 just to maintain its current
revenue. To achieve its traditional 15 percent annual revenue
growth rate in memor y devices, IBM would have to ship 3,600
times as much memor y capacity during the last year of the decade
as it did during the first. A similar line of reasoning indicated that
40 times as much disk storage capacity would have to be installed
per year at the end of the decade.
Although the cost-reduction projection for memory was about
ten times higher than actually occurred, the qualitative message
was correct. T he industry's success in lowering the costs of its
products would greatly reduce its future revenues unless innova­
tive new uses were found for computer power, and especially for
memory capacity.14
Assessing this situation in December 1970, the Corporate Tech­
nical Committee urged that a SPREAD-like task force be estab­
lished to set goals for what was already being called the Future
System, or simply FS. In August 1971 the proposed task force was
established with John R. Opel in charge. Opel, who became IBM
president in 1974, had joined the company as a sales repre­
sentative in 1949 immediately after receiving his master's degree
in business administration from the University of Chicago. Se­
lected as Tom Watson's administrative assistant in 1959, he rose
rapidly and was in charge of the corporate finance and planning
staffs when he was put in charge of the FS task force. Other task
force members included several of the company's top technical
leaders.
A major goal for FS was to reduce the cost of developing
application programs. This was considered essential because of
the growing shortage of programmers and the high costs in­
volved. Customers at the time were spending twice as much
money on programmers and other computing-center personnel
as they were on hardware. If FS could be structured to reduce the
number of people required to program and operate computers,
IBM would receive a higher fraction of the money spent by
customers for information processing. 15
The most promising proposal considered by the task force was
based on an explorawpy� _fill/study at the company's
 Onrush of Technology 309

T. J. Watson Research Center in Yorktown Heights. Users of the

proposed system would not have to concern themselves with the
amount or type of main memory and storage or where any
particular information might be stored. All physical memory and
storage was to be represented to the user as if it were a single-level
store. Addressing was to be object-oriented rather than location-ori­
ented, and information stored anywhere in the system would be
addressed by the user in the same manner.16
There were insufficient experimental data on automatic, sys­
tem-managed storage hierarchies to guarantee success in this
endeavor. The success of the cache, however, which moved infor­
mation automatically between main memory and a high-speed
buffer, and the success of automatic paging between disk and
main memory on the System/360 Model 67, gave considerable
encouragement. The potential benefits were sufficiently great
that the risk of a major development effort seemed warranted.
The single-level store became a central feature of FSP
Development resources were increasingly committed to FS dur­
ing 1972 in an effort reminiscent of the one that produced
System/360. The increased complexities of computers and of the
international market they served were reflected in the size and
complexity of the FS effort. Intended to provide user-friendly
interfaces for professional programmers and novices alike, the
internal structure of FS was far from simple. Conflicting views of
the many development groups were difficult to resolve because
few, if any, understood the entire FS architecture. New problems
were uncovered as rapidly as old ones were solved. Schedules
slipped so badly that the number of months projected before
announcement of FS was forty-five in the spring of 1974. just as
it had been in the fall of 1971 when the Opel task force completed
its work.18
By the fall of 1974 it was clear that the company's System/370
computers would be in serious jeopardy in the marketplace long
before FS could be shipped. Following extensive consultations
with key technical leaders and top corporate executives, Bob
Evans undertook the painful task of seeking the best way to
terminate FS and to "extend the architecture of System
360/370. "19 In Februar y 1975 the FS effort was terminated. Al­
though much of the hardware developed for FS could be
modified for use in any new computer, most of the system design,
microcode, and software were discarded. It was the most expen­
sive development fail�.i'riritted:�'s history.2o
310 Building IBM

Many causes for the FS failure have been suggested. Frequendy

cited is the attempt to achieve so many technological advances in
a project in which all advances had to be accomplished on the
same schedule and had to function compatibly with one another.
Other frequendy cited causes are the complexity of the FS archi­
tecture itself and the complexity of coordinating a development
effort among laboratories throughout the world. Returning to
IBM from the University of North Carolina as a part-time consult­
ant in June 1973, Fred Brooks summarized the documented
description of FS architecture with one word, "Incomprehensi­
ble." Concerning the overall project, he observed, "Complexity is
the fatal foe. "21
Another important factor was the change in market conditions.
Beginning as a proprietar y development, 360-370 architecture
had become such a dominant worldwide standard that many of
IBM's competitors had a vested interest in perpetuating it. Among
these were leasing companies whose profitability depended upon
keeping 360-370 computers (purchased from IBM) installed as
long as possible. Had IBM developed and introduced FS, the
leasing companies would have vigorously resisted its installation
by their customers.
Leasing companies were already cutting their costs and upgrad­
ing the systems they rented to customers by purchasing magnetic
tape and disk storage devices from plug-compatible manufac­
turers. Yet greater improvements in the performance of these
systems could be achieved with the addition of more plug­
compatible memories and peripheral devices. Several engineer­
ing firms and independent software developers were ready to
help leasing companies upgrade their systems. Customers that
had chosen to purchase rather than rent systems from IBM would
also carefully weigh the benefits of new FS functions against
the more easily understood option of upgrading their installed
systems.
These impediments to a successful introduction of FS-no
matter how good it might have been-were exacerbated by inde­
pendent manufacturers of 360-370 compatible processors.
Among these was Gene M. Amdahl, a key architect of the original
System/360 line who had left IBM in 1970 to develop and manu­
facture large IBM-compatible processors. Significant funding for
his company came from Japanese and German computer compa­
nies that intended to use the Amdahl Corporation as a means for
obtaining advanced and for entering the
<C>0}3VH91ifelffi.bH9A�P
 Onrush of Technology 311

U.S. computer market. Gene Amdahl's technical and manage­

ment skills , his knowledge of IBM technology and business plans,
and the source of his funding made his corporation a serious
threat.22
Only four months after FS was terminated, the first Amdahl
computer was installed. Known as the Amdahl 470 V/6, this
System/37O-compatible computer was targeted for the top of the
IBM line. Two agonizing years passed before IBM was able to
complete its shift in development emphasis and announce its first
"post-FS" computer. Not surprisingly it was a high-end computer,
capable of competing with Amdahl's offerings. "The 3033 Pro­
cessor, a member of the System/370 family, together with exten­
sions to MVS and System/370 architecture, will provide large
systems users with significantly increased data processing capabil­
ity," the IBM March 1977 announcement asserted. Built under
the leadership of John E. Bertram, who had proposed the plan
for shifting from FS back to 360-370 architecture, the 3033 lived
up to it" billing and successfully blocked Amdahl's effort to domi­
nate the high end of the mainframe computer business.23

Strategic Concerns

The termination of FS climaxed a decade of uncertainty about

the future of System/360 and its underlying architecture. Fears
that rapid progress in integrated circuits would cause its early
demise were replaced by recognition that it had become an
industr y standard in which IBM had the largest investment. Pro­
tecting that investment was fundamental to the company's future.
The validity of that strategy is readily confirmed. In 1989, a
quarter of a century after System/360 was announced, products
based on 360 architecture and its extensions still accounted for
more than half of IBM's revenue and a much higher fraction of
its profits. This architecture was also the basis of more than 70
percent of the estimated $160 billion worldwide inventory of all
computers priced over $1 million and more than 50 percent of
the $260 billion worldwide inventory of all computers priced over
$100,000. The vast majority of installed computers based on 360
architecture and its extensions had been manufactured by IBM,
although an increasing number were being supplied by Fujitsu
Limited and Hitachi Limited, frequently through cooperative
arrangements with vendors such as the Amdahl Corporation.24
Copyrighted Material
312 Building IBM

A critical problem faced by IBM was that no computers priced

under $100,000 used an architecture based on System/360, and
by 1989 these lower-priced computers already accounted for half
of the worldwide inventory. Even when System/360 was an­
nounced in April 1964, the company's engineers had been unable
to achieve compatibility in systems smaller than the Model 30. For
this reason the Model 20 (announced in November) was not fully
compatible.25
In July 1969 the IBM System/3 had been announced in an
effort to satisfy the needs of yet smaller businesses. There was no
pretense of compatibility with System/360. System/3 even fea­
tured a smaller punched card that was 2% inches high and 31/4
inches wide. Six-bit characters were employed, and round holes
were used to encode up to 96 characters per card. Designed with
the company's new monolithic integrated circuits, System/3
rented for less than $1,000 per month, which was half the typical
price of a System/360 Model 20.26
In October 1969 the IBM General Systems Division was estab­
lished to manufacture System/3 and to develop future low-cost
systems. A significant accomplishment of this division was re­
vealed by the October 1978 announcement of the IBM Sys­
tem/38, designed to replace the successful but aging System/3.
Under development since 1973 in the General Systems Division's
laboratory in Rochester, Minnesota, System/38 incorporated
many advanced features originally planned for FS. These in­
cluded a single-level store, object-oriented addressing, and a high­
level machine interface to the user. Developing these features for
the small system was easier than for FS because of the limited
market and simpler applications for which it was intended.27
System/38 achieved considerable popularity in the early 1980s,
and its successor AS/400 series was even more successful. As the
decade of the 1990s began, the annual revenue from the AS/400
series and related products exceeded the total annual revenue of
the Digital Equipment Corporation, which had the second high­
est revenue in the computer industry. 28
As a low-end product, the AS/400 was relatively expensive,
however, with prices ranging up to more than $500,000. What
IBM needed was a strategy for competing in the market for truly
low-cost systems and computer terminals, which had grown rap­
idly since the late 1960s.
The company's Office Products Division appeared to be well
positioned to play a ffij}jVfigiflbfd ma'ewOJ a strategy. Its products
 Onrush of Technology 313

included typewriters, magnetic-tape- and magnetic-card-driven

typewriters, communicating typewriters, copiers, and dictating
equipment. Previously called the Electric Typewriter Division, it
had become a "completely autonomous operation" in 1955 so it
could better control its costs and operate in the spartan manner
needed in its price-sensitive market.29 As noted in a memorandum
written for the Corporate Technical Committee in 1970, "The
data processing world is evolving toward a state in which terminals
and office products are virtually indistinguishable." By combining
some of the systems design capabilities of other divisions with the
manufacturing, marketing, and service capabilities of the Office
Products Division, there was "an opportunity to enter this new
marketplace while taking advantage of the traditional IBM
strengths. "30
Thus as the 1970s began, IBM appeared to be well structured
to prosper in office equipment and small systems. This apparent
strength was dissipated, however, by ill-defined and conflicting
missions of the Office Products, General Systems, and Systems
Development divisions. Furthermore because office products, ter­
minals, and small systems typically had lower profit margins than
large systems, it was difficult to get corporate management inter­
ested in new product initiatives.31
A relatively unsuccessful effort to market the fifty-pound IBM
5100 portable computer, introduced at a price of $9,000 to
$20,000 in 1975, confirmed for many that there was little future
for IBM in small computers.32 At the same time, however, several
companies (including Apple, Atari, Commodore, and Tandy)
successfully introduced personal computers, and industrywide
sales of small computers began increasing at a rate in excess of
50 percent per year.

The IBM PC

Determined to overcome all internal impediments to the success­

ful introduction of small computers, the Corporate Management
Committee took extraordinary action in 1980. To develop, manu­
facture, and market an IBM PC, it established an independent
business unit (IBU) in the company's laboratory in Boca Raton,
Florida. An attractive feature of an IBU was its ability to provide
"small-company" flexibility and incentives to groups developing
products outside IBM's main line of business. Application of the
IBU concept to a p���rlaY expand the company's
314 Building IBM

traditional business at the low end entailed considerable risk, but

it appeared to be the only way to bring about the many changes
in business practices that were likely to be needed.33
Given considerable freedom to chart its own course, the Boca
Raton IBU adopted many policies and practices that were unusual
in IBM. To achieve broad market acceptance, for example, the
PC was to have an open rather than proprietary architecture. To
reduce development time and costs, the operating system was to
be purchased from an independent software firm, and hardware
components were to be open for competitive bids from inside
and outside the company. In perhaps the most dramatic depar­
ture from tradition, the PCs were to be sold through retail chan­
nels rather than by the company's highly touted (and costly) sales
force.34
Spurred on by a brilliant advertising campaign that featured
Charlie Chaplin's little tramp, sales of the PC soared beyond all
expectations. But lacking proprietary control over the PC's basic
design elements, the company was soon faced by competition
from lower-priced IBM-compatible computers. Manufacturers of
these clones often obtained their components from the same
vendors that supplied IBM-thus making it difficult to convince
buyers that the company's higher-priced products were superior.
Even more significant was the company's inability to provide
customers with service that was substantially better than that of
its competitors. The costs of IBM's traditional service were judged
to be far too high, and no satisfactory low-cost alternative was
found. The limited service provided to PC customers represented
a significant departure from past practice. It will be recalled that
Tom Watson had promoted the idea that "IBM means service."
His father had insisted on providing the best service in the
industry. And Herman Hollerith had refused to install any equip­
ment until he had assured himself that it would handle the cus­
tomer's jobs effectively.

RIse Architecture

The sustained success of System/360 architecture, combined with

the failure of FS, created yet another problem. Product managers
in IBM had become resistant to almost any new architectural
proposals-including the RISC (Reduced Instruction Set Com­
puter) architecture. Based primarily on ideas of IBM Fellow John

Copyrighted Material
 Onrush of Technology 315

Figure 20.3. The IBM Personal Computer

Announced in August 1981, the IBM Personal Computer (PC) used an Intel
(l6-bit, 4.77 megahertz) microprocessor and a disk operating system (DOS)
supplied by Microsoft. Memor y capacities from 16 KB to 256 KB were avail­
able as were one to two disk drives with capacities of 160 KB each (KB =

1,024 bytes). The IBM PCs were sold through IBM Product Centers, Com­
puterland, and Sears, Roebuck and Company. Prices ranged from $1,565 to
approximately $6,300. The system shown above with the standard 8O-charac­
ter-per-second dot-matrix printer, keyboard, system unit (with two disk
drives) , and monitor sold for approximately $4,500.

Copyrighted Material
316 Building IBM

Cocke, RISC architecture gives small computers several times the

performance of those designed with earlier architectures. First
implemented in the experimental 801 Minicomputer, named af­
ter the IBM 801 Building (Watson Research Center) in which it
was built, RISC met considerable resistance in a company dedi­
cated to extending the 360-370 architecture. Although IBM used
RISC in the input-output processors of its 3090 series of large
computers beginning in 1985 and then in its RT PC workstation
announced in 1986, it did not fully embrace the architecture until
the RISC System/6000 family of advanced workstations was an­
nounced in February 1990. By then the power of RISC had been
exploited in the marketplace by Sun Microsystems, Hewlett
Packard, and others-forcing IBM to catch up with its own tech­
nology.
The RISC System/6000, as announced in 1990, exemplified the
dramatic onrush of technology during the twenty-six years since
System/360 was first announced. Even the most powerful model
(POWERserver 540) sat inconspicuously beside a desk rather
than filling a large room. It processed 41 million instructions per
second, making it five to fifty times more powerful (depending
on problem mix) than the most powerful of the early System/360
models. Its electronic logic circuitry had up to 800,000 transistors
per silicon chip versus only 1 transistor per chip as first an­
nounced with System/360. Its maximum memory size of 256
megabytes was just 256 times more than was offered on the largest
of the early System/360 line, and its internal disk storage capacity
of 2.5 gigabytes was 25 times the capacity of the 24-inch diameter,
25-disk module of the IBM 2302 Disk Storage announced with
System/360 in April 1964.35

Copyrighted Material
21
Demands of the Future

The announcement of IBM's RISe System/6000 was made one

century after Herman Hollerith won the contract to tabulate the
census of 1890. During the intervening one hundred years, dra­
matic changes had occurred in the technology, operation, and
public perception of information processing systems. The public
had had little awareness of their existence during the first fifty
years, except for Social Security checks printed and issued on
punched cards beginning in the 1930s. During the second fifty
years, public awareness increased significantly as electronic tech­
nologies made possible previously unimagined applications in the
exploration of outer space, the production of nuclear power,
airline reservation systems, factory automation, medical systems,
and many others.
By the end of the 1960s IBM management had determined that
continued growth of the industry could be achieved only by
making the power of computers more directly available to wider
segments of society. The Future System (FS) project was estab­
lished to accomplish this objective. Emphasis on FS, however,
tended to obscure the need for other decisions. The company's
inability to achieve adequately low costs in terminals and small
systems, for example, had been expected to be offset by FS's
control of the entire system. To ensure compatibility, customers
were likely to have a strong initial preference for attaching IBM
peripheral devices to FS.
With termination of FS and continued reliance on architec­
tural extensions of System/360, customer acceptance of periph­
eral devices supplied by other vendors was expected to continue
to increase. Indeed all the anticipated problems that had trig­
gered the decision to develop FS were unchanged. Product costs
and prices were continuing to decline at an exponential rate, and
Copyrighted Material
318 Building IBM

competition was increasing. It appeared unlikely that IBM could

maintain for long both its dominant market share and its high
profit margins. An upturn in demand for computers had delayed
the day of reckoning, but it was only a matter of time before the
onrush of technology would force dramatic changes in product
strategy.

New Leaders

New decisions and actions were urgently needed, but the "take
charge" leaders who had driven IBM into large-scale electronic
computers and ushered in the System/360 era were no longer
available. Tom Watson had resigned as chief executive officer
following a heart attack in November 1970.1 His typical twelve- to
fourteen-hour days and the strain of System/360 had taken their
toll. Consistent with his standard approach to any major IBM
problem, he had put Learson in charge. This temporary transfer
of authority became final when the board of directors elected
Learson chairman and CEO beginning in June 1971. Learson's
term lasted only until December 1972 because a new IBM policy
required senior corporate executives to resign in the year in
which they turned sixty. 2
Speculation abounds that the policy was established to prevent
certain individuals from assuming higher positions or from stay­
ing in those positions too long. Officially the policy was estab­
lished to provide advancement opportunities for the large
number of competent people who had been hired into the com­
pany during the postwar years. Concerning his own successor as
CEO, Tom Watson had apparently never believed it should be
Learson. Watson's heart attack forced a change in plans, but only
briefly.
Watson said he selected Frank T. Cary to be his successor when
Dick Watson left the company in 1970 to become the U.S. ambas­
sador to France. Cary, who had a master's degree in business
administration from Stanford University, had joined IBM as a
marketing representative in Los Angeles in 1948. Rising rapidly
in the sales and marketing organizations, he became an IBM vice
president and group executive and general manager of the Data
Processing Group in 1966 and a member of the board of directors
in 1968. "He rarely spoke up at meetings," Watson recalled, "and
it wasn't his style to step in and save the day the way Learson and
Copyrighted Material
 Demands of the Future 319

I did. He didn ' t make heroic moves and he didn't make glar ing
mistakes; when he ran into a problem he simply figured out how
to fix it."3
Fixing the immediate problems caused by the termination of
FS was relatively easy. A back up plan for improving System/370
with new hardware and software was put in place. The time lost
in develop i ng the new products was costly, but not devastating.
Finding a solution for the longer-term problems was not so easy.
Conservatism bred in the FS failure and the continuing profi ta­
bility of extensions to System/360 inhibited decisive action during
the 1970s.

Adding to the problems Cary faced was an antitrust suit filed by

the U.S. Department of Justice in Januar y 1969. Concluding that
IBM ' s size gave it excessive powers, the department sought to
break the company into smaller competitive un its
As is often the case in antitrust suits, a key issue was definition
of the market that was alleged to have been monopolized. The
Department ofJustice chose to define the market so as to exclude
leasing companies , plug-compatible manufacturers, compan ies
such as DEC whose computers it j udged to be not general pur­
pose, and companies that provided only some of the products
and services in the government ' s arbitrarily defined market. By
this definition, IBM's share of the revenue was about 70 percent
throughout the 1960s and 1970s. If one summed up the revenues
for computer products and services of the one hundred top
companies in the computer industry, as IBM chose to do in its
defense , however, its share would have dropped from nearly 60
percent in the mid-1950s to less than 40 percent by 1969 when
the suit was filed.4
The antitrust suit droned on for thirt een years. The trial itse lf
did not begin un til May 1975, and it lasted six y ears. The man­
agement of the trial is incomprehensible . The judge, for example,
required that page after page of deposition testimony be read
into the record so he could hear them. Then he routinely ab­
sented himself from court while they were being read. The trial
called upon 974 witnesses and produced 104,400 transcript
pages.5 Final l y in J anuar y 1982 the government dropped the case,
asserting it had been "wi thout merit."

Copyrighted Material
320 Building IBM

The decision was long overdue according to most knowledge­

able observers. Mter reading much of the trial transcript and
interviewing scores of participants, the editor of American Lauryer
described the trial as "a farce of such mindboggling proportions
that any lawyer who now tries to find out about it so he can
explain it to a client or a friend will be risking the same quicksand
that devoured the lawyers involved in the case." Concerning the
primary issue in the case, he said, "The reality of the 1970s and
1980s is that IBM had and has no monopoly unless one strains,
as the government trial lawyers persisted in doing, to gerryman­
der a bizarre market definition."6
In the meantime IBM had spent tens of millions of dollars per
year to defend itself. An even greater cost was the involvement of
key executives who might better have concentrated on new busi­
ness. Furthermore, all major business decisions made during
those thirteen years had to be analyzed for their impact on the
antitrust case. The government's suit also encouraged several
companies to file their own antitrust suits against IBM. Every one
of these was decided in IBM's favor, but the cost was high.'
Recalling that Tom Watson chose to settle the antitrust case of
1952 so he could concentrate on business matters-and that
Hollerith finally gave up fighting the government in 1910 for
similar reasons-one can ponder the wisdom of fighting the case
filed in 1969. Many of the reasons for fighting appear to have
vanished with the FS failure. Unable to create a new system
architecture capable of dominating the market, IBM could only
anticipate increasing fragmentation of its marketplace and a de­
clining benefit from its own size.
Considering the heavy constraints antitrust enforcement
placed on IBM's ability to compete, it might have been wiser to
break the company into several smaller units as the government
proposed. Support for this view is provided by the experience of
AT&T. Yielding to government pressure, it agreed to divest itself
of its local telephone business in the same year that IBM "won"
its antitrust case. Mter that the value of AT&T stock more than
doubled whereas the value of IBM stock declined. "I am con­
vinced," the CEO of AT&T told shareholders in 1994, "that the
breakup forced us to make changes in the 1980s that prepared
us for the competitive challenges we are confronting in the
1990s."8

Copyrighted Material
 Dt!mands 0/ the Future 321

�
c
..
..,
30,000 Q.
..
'" .., c u
... .2 �
..,
..,
11 '"
u ..
(5 10,000
II> U
C �
.., � .., c
0 Q.
e .!: ... �
0
'0 U Q. r--
u
� c
3,000 <I: 0 .., .., C
� .,. � c
II> �
c:
� .� e 0
.2
1,000 II> m
....
>-
e
�

� � ..
Q. .!:
c: c: 0
i
.., .., u :5! � c>
.. E U
0
c:
� .. \D

� �
..
'" ..,. a.. ......
..
�
:::J (I)
c:
.... E II> E
300
.c .. U �
�
t
... 0 f- e
0) u c: ... .. m
:::J e .. -'" u 'c- 0)
.::
.... <II
c: � ..,. u 0 .. '2 ,..
(I)
0)
> 100 e (I) ..,. a; Q.
.. � It:
iii II> ..,
(I) U

l !
It:
0
f-
u 0 (I)
.� u...
II>
� 0) \D

l �
It:
II>
30 . :: ..,
� a; ......
(!) .., E
..
10 w
(!) Vi
>-
<I: (I)
(I)

1920 1930 1940 1950 1960 1970 1980 1990

Year

Figure 21.1. ffiM's Gross Revenue

The worldwide gross revenue of IBM increased 15,OOO--fold from 1915 to
1993 as revealed by this logarithmic graph. Shown in millions of dollars per
year, the revenue is not adjusted for inflation. Some key events in the com­
pany's history are indicated.

Coda

From the vantage point of the early 1990s, it is difficult not to

e m phasi ze the recent difficulties IBM has experienced. But in
reviewing the company's history and the dramatic social and
technological changes it has sur vived, one cannot help but be
impressed more by its successes than by its failures. Its decades
of effective responses to dramatic technological c hanges, begin­
ning in the 1940s, are particularly noteworthy. M any companies
flourished briefly and then floundered while IBM moved forward
with apparent invincibility. Not until the mid-1980s did the com­
pany ' s financial per formance give clear evidence of the difficul­
ties it was experienci n g as the cost of computer technology
continued to spiral downward.9
Key to IBM's long-term success had been the absolute commit­
ment of Herman Hollerith, Thomas ]. Wat<;on, Sr., and Tho­
mas]. Watson,Jr. They understood the importance oftechnology

Copyrighted Material
322 Building IBM

and were continually active in the crucial interfaces between

technical developments and business strategies. They demanded
the best from themselves and their subordinates. They were quick
to punish and quick to reward. They insisted on satisfying the
needs and desires of their customers. And both Watsons culti­
vated relationships with people inside and outside the company
that kept them alert to changes in technology, competition, and
customer requirements.
In words as valid today as they were when spoken in 1932,
Watson, Sr., asserted, ''There is no business in the world which
can hope to move forward if it does not keep abreast of the times,
look into the future and study the probable demands of the
future. "IO

Copyrighted Material
Appendixes

Appendix A: IBM Worldwide Revenues and Year-End Populaticm

Year Revenues in $ Million Year-End Population

Gross Net
Income Earnings U.S.A. Total

1914 4 1,217 1,346

1915 4 1,517 1,672
1916 6 2,323 2,529
1917 8 2,851 3,063
1918 9 2.848 3,127
1919 11 2 2.660 3.139
1920 14 2 2.437 2 7 31
,

1921 9 2,487 3,001

1922 9 2,432 3.043
1923 11 2 2.399 3.161
1924 II 2 2. 47 2 3.384
1925 13 3 2.633 3.698
1926 14 4 2.741 3,953
1927 14 4 3.133 4.866
1928 15 5 3,491 5.102
1929 18 7 4.189 5.999
1930 19 7 4,339 6.346
1931 19 7 4.:n4 6,3:11
1932 17 6 4.94 6 6.311
1933 17 6 6,591 8 202
.

1934 19 7 5.674 7.613

1935 21 7 6,268 8.654
1936 25 8 6.4 74 9,142
1937 31 8 7,589 10.834
1938 34 9 7.602 11.046
1939 38 9 7.610 11.315
1940 45 9 8.602 12.656
1941 60 10 10, 1 70 14,207
1942 86 8 14.703 18.754
1943 1 31 9 17.262 21.251
1944 1 40 10 15,133 21,126
1945 138 11 14.476 18,257
1946 116 19 16,977 22,492
1 9 47 139 24 16,556 22,591

Copyrighted Material
324 Building IBM

Year Revenues in $ Million Year-End Population

Gross Net
Income Earnings U.S.A. Total

1948 156 28 18,292 24,940

1949 183 33 18,938 27,236
1950 266 37 20,702 30,261
1951 335 32 24,255 35,124
1952 412 34 28,614 41,458
1953 497 39 30,810 46,170
1954 570 59 33,732 50,225
1955 696 73 39,033 56,297
1956 892 87 51,192 72,504
1957 1,203 110 60,281 83,588
1958 1,418 152 61,159 86,736
1959 1,61 3 176 65,646 94,912
1960 1,817 205 70,050 104,241
1961 2,202 254 75,954 116,276
1 962 2,591 305 81,493 127,468
1963 2,863 364 87,173 137,612
1964 3,239 431 96,532 149,834
1965 3,573 477 111,087 172,445
1966 4,248 526 129,023 198,186
1967 5,345 651 144,206 221 ,866
1968 6,889 871 154,874 241,974
1969 7,197 934 159,976 258,662
1970 7,504 1,018 156,859 269,291
1971 8,274 1,079 149,022 265,493
1972 9,533 1,279 146,895 262,152
1973 10,993 1 ,575 152,177 274,108
1974 12,675 1,838 160,542 292,350
1975 14,437 1,990 156,938 288,647
1976 16,304 2,398 158,380 291,977
1977 18,133 2,719 170,808 310,155
1978 21,076 3,111 1 81,057 325,517
1979 22,863 3,011 190,448 337,119
1980 26,21 3 3,397 194,423 341,279
1981 29,070 3,610 205,230 354,936
1982 34,364 4,409 214,434 364,796
1983 40,180 5,485 218,671 369,545
1984 46,309 6,582 238,9551 394,9301
1985 50,718 6,555 242,541 405,535
1986 52,160 4,789 237,274 403,508
1987 55,256 5,258 227,949 389,348
1988 59,598 5,741 223,208 387,1 12
1989 62,654 3,722 215,929 383,220
1990 68,931 5,967 205,421 373,289
1991 64,766 (2,861) 190,666 344,396
1992 64,523 (4,965) 161,980 301,542
1993 62,716 (8,101) 125,000 256,207
1. Includes Rolm Corp. (9,902).
Nole: Business Machines
of 8 May 1941 reported
that there were 250 persons in
ss presumablCEAMrj i�l�H,fa� U,oOO persons in 1941; Business
"this bus ine ,"

Opy, 19 e a ena
 Appendixes 325

Machines of 16 April 1942 reported that there were 250 persons in the business
in 1914 and over 12,000 tod ay The number of employees worldwide peaked
" ."

at 407,080 in 1986.
Source: Based on data assembled by the IBM archives and from IBM Annual
Reports.

Appendix B: IBM's Early Patents Filed on '1Qbulators and Time Recorders during
Five-Year Periods

All
Filing CTR-IBM
Date Inventors Bryce Car rol Ford Lake Peirce

1911-15 11 3 0 0 0 7
1916-20 65 30 0 7 3
1921-25 143 42 11 7 21 12
1926-30 301 52 29 16 3 20
1931-35 363 45 13 10 24 12
1936-40 305 44 14 8 20 1
1941-45 405 21 10 3 21 0
1946-50 513 7 9 6 11 0

Totals 2106 244 87 50 107 55

Source: Based on d at a in the IBM Pate n t Digest of 12 February 1952.

Appendix C: Early Electronic Computing Circuit Patents

Filed u.s. Pat. Ti tle Inventor(s) Company

1/20/40 2,580,740 Accounting Apparatus Dicki n s on IBM

3/20/40 2,595,045 Calculating Machine Desch, NCR
Mumma

7/18/40 2,404,739 Calculating Machine Mumma NCR

5/23/41 2,402,989 Accounting A pparatus Dicki ns on IBM
5/23/41 2,580,741 Accounting Apparatus Di cki nson IBM
8/25/42 2,446,945 Electronic Computing Morton, Flory RCA
Device
11/02/42 2,409,689 Electronic Computing Morton, Flory RCA
Device
Source: B. E. Phelps, 1980: "Early Electronic Computer Developments at IBM,"
Annals f th£ History of Computing 2. pp. 253-67.
o

Copyrighted Material
326 Building IBM

Appendix D: IBM U.S. Product Revenues by Year (in $ millions)

Commercial
604 Stored IBM
Other and Program U.S.
Year SAGE Military epe Computers Total

1952 29 9 0 334
1 953 4 49 13 1 41 0
1954 14 42 17 4 461
1955 47 35 21 12 564
1 956 93 28 23 49 734
1 957 1 22 70 26 124 1000
1958 1 00 III 29 209 1 172
1959 97 90 30 306 1 31 0
1 960 57 72 31 399 1 436
1961 10 173 31 556 1694
1962 13 1 20 25 797 1925
1963 0 60 20 1045 2060
1964 0 56 16 1305 2288

Note: Revenues from product programs that were significant to IBM's entry into
commercial electronic computers are depicted in this table. Particularly impor­
tant in the early years was the revenue and associated experience gained from
the 604 and epe products. Revenue began in 1948 and exceeded that from
commercial stored-program computers through 1955. SAGE development be­
gan almost two years after commercial stored-program computer development,
but revenue, beginning with design contracts, came earlier; there was no
revenue from commercial products until after they were placed on rental.
Military products other than SAGE consisted at first of electronic bomb­
ing/navigation systems for long-range (8-52 and 8-70) aircraft, and evolved into
missile guidance systems. All revenues include monies derived from electrome­
chanical as well as electronic devices in the systems. Service revenues are
generally included only in the total column, although for rental machines,
maintenance service was included in rental. The 604 and epe revenue is
somewhat understated, omitting the accounting machine part of epe revenue.
Saurc£.5: Defendant's Exhibits in U.S. v. IBM (1969): DX2609, DX2609A,
DX3815, and DX3822.

Copyrighted Material
 Appendixes 327

Appendix E: Top Curp(IYQte Officers, 1911-1994

Chief Term as
Executive Chairman
Chairman President Officer! or President

George W. Fairchild 1911 July to

1 924 December
Percy H. Brundage 191 1 July
George W. x 1 91 1 August to
Fairchild (acti ng) 1912 April
Frank N. Kondolf x 1 912 April to
191 4 December
None 1915 January to
1915 March
Thomas J. Watson2 x 1915 March to
1949 September

None 1925 January to

1949 September

Thomas J. Watson a 1949 September

to 1956June
John G. Phillips 1 949 September
to 1952January

Thomas J. Watson, b 1952 January to

Jr. 1961 May

None 1 95 6 J un e to
1961 May

Thomas J. Watson, x 196 1 May to

Jr. 1971 June

Albert L. W illiams 1 961 May to

1966 March

T. Vincent Learson 1966 April to

1971June

T. Vincent Learson x 1971 June to

1 972 December

Frank T. Cary 1971 June to

1974 February

Frank T. Cary c 1973 January to

1983 February

John R. Opel d 1974 February to

1 983 February

John R. Opel e 1 983 February to

1 986 May

John F. Akers f 1 983 February to

1 989 May

Copyrighted Material
328 Building IBM

Chief Term as
Executive Chairman
Chairman President Officerl or President

John F. Akers f 1986 June to

1993 March
Jack D. Kuehler 1989 May to
1993 February
None 1993 February-
Louis V. Gerstner, Jr. x 1993 April-
1. A letter in the Chief Executive Officer column indicates service as chief
executive while also serving as chairman or president. Letters other than x
indicate that the term as chief executive differed from the term as chairman
or president, as follows:
a. Term ended 1956 May
b. Term began 1956 May
c. Term ended 1980 December
d. Term began 1981 January
e. Term ended 1985 January
f. Term began 1985 February
2. Thomas J. Watson join ed the Computing-Tabulating-Recording Company in
May 1914 as general manager. In February 1924 the company name was
changed to International Business Machines Corporation.
Source: All but most recent entries are from E. W. Pugh, L. R. Johnson, and
J. H. Palmer, 1991: IBM's 360 and Early 370 Systems (Cambridge, Mass.: MIT
Press), pp. 649-50.

Copyrighted Material
 Appendixes 329

Appendix F: Computer Improvements

Technological advances have produced dramatic changes in speeds, capacities,

and prices of computer systems as illustrated in the tables. Top-of-the-line disk
files and low-priced mainframe computers were used for comparison. By the
] 9805 there was a vast array of lower-priced, lower-performance computers and
disk storage units available, which are not revealed below.
All data are as of the announcement; dollar figures are not adjusted for
inflation. Data points p rio r to 1980 are from E. W. Pugh. L. R. Johnson, and
J. H. Palmer, 1991: IBM's 360 and Early 370 Systems (Cambridge, Mass.: MIT
Press), pp. 645-646.

Price of Computer Processing

Purchase Price
per Instruction
Announcement Executed Price
System Year per Second Ratio

650 1953 $242.29 2423

System/360 Model 30 1964 6.13 61
System/370 Model 135 1971 2.71 27
4341 1979 .35 4
ES/9000 Model 190 1990 .10

Computer Processing Speed

Announcement Speed
System Year Ratio]

650 1953 1
]401 1959 7
System/360 Model 30 1964 43
System/370 Model 135 1971 214
4341 1979 1143
ES/9000 Model 190 1990 15200
1. Processing ratios are based on selected sample tests run by IBM.

Copyrighted Material
330 Building IBM

Computer Main Memory Capacity

Announcement Maximum Capacity

System Year Characters Ratio

1401 1959 4,000 1

System/360 Model 30 1964 66,000 17
System/370 Model 135 1971 246,000 62
4341 1979 4,194,000 1,049
ES/9000 Model 190 1990 536,576,000 134,144

Price of Computer Disk Storage

Purchase Price
Announcement per Million Price
Machine Year Characters Ratio

350 1956 $9760 976

1301 1961 3313 331
2314 1965 1243 124
3330-11 1973 191 19
3370 1979 41 4
3390 1992 101 1
1. Substantial discounts were widely available.

Disk Read-Write Speed

Announcement Characters Speed

Machine Year per Second Ratio

350 1956 10,000 1

1301 1961 90,000 9
2314 1965 312,000 31
3330 1970 806,000 81
3350 1975 1,198,000 120
3370 1979 1,859,000 186
3390 1992 4,200,000 420

Copyrighted Material
Notes

1 Hollerith: Inventor and Entrepreneur

1. The three patents issued to Hollerith on tabulating equipment on 8 January
1889 were U.S. Patent 395,783, filed 23 September 1884: "Apparatus for Com­
piling Statistics"; U.S. Patent 395,782, originally filed 23 September 1884, di­
vided and filed 27 October 1885: "Art of Compiling Statistics"; U.S. Patent
395,781, filed 8 June 1887: "Art of Compiling Statistics."

2. G. D. Austrian, 1982: Herman Hollerith, Fargotten Giant of Infarmation Processing

(New York: Columbia University Press) , pp. 1-4, 43 . His continuing sensitivity
to his poor spelling is revealed by the following sentences in a letter to T. J.
Wilson on 7 August 1 9 1 9: "Please make due allowances for errors of spelling
etc. Life is too short to correct type writing." The Columbia School of Mine s
was the predecessor of the Columbia University School of Engineering.

3. K. S. Reid-Green, February 1989: "The History of Census Tabulation," Scien­

tific American, pp. 98-103.
4. Austrian , 1982: pp. 1-4.

5.M. G. Anderson, 1988: The American Census: A Social Histary (New Haven: Yale
University Press) , especially pp. 9, 83-87.
6. John Shaw Billings first became involved with the U.S. Census in 1870, and
he had responsibility for vital statistics in the censuses of 1880 and 1890. As a
member of the army, he was not on the census payroll. Beginning in 1876 he
also ser ved as medical adviser to the trustees of the Johns Hopkins Fund and
helped plan the Johns Hopkins Hospital. He became a full professor of hygiene
at the University of Pennsylvania in 1895 when he retired from the army and
spent his final years helping to organize the New York Public Library. See L. E.
Truesdell, 1965: The Dl!IJelqpment of Punch Card TaUuwtion in the Bureau of the
Census, 1890-1940 (Washington, D.C.: U.S. Department of Commerce) , pp. 28-
30.

7. H. Hollerith, 7 August 1919: to J. T. Wilson, as reproduced in the IBM

document, 1957: "Historical Development of IBM Products and Patents," pre­
pared by J. Hayward, pp. 18-23. Wilson was a former sales agent with the
Tabulating Machine Company.
8. Hollerith, 7 August 1919.
9. Austrian, 1982: pp. 11-23.
Copyrighted Material
332 Notes to Pages 5-10

1 0. L. E. Truesdell, 1 965 : The Development of Punch Card Tabulation in the Bureau

of the Census, 1890-1940 (Washington, D.C.: U.S. Department of Commerce),
pp. 17-24. Charles W. Seaton, who was chief clerk during the 1 870 census, was
paid $15,000 for his invention by a special act of Congress in 1 872. At the time,
Francis A. Walker was superintendent. Seaton replaced Walker as superinten­
dent of the census in 1 88 1 when Walker became president of the Massachusetts
Institute of Technology.
1 1 . Reid-Green, February 1 989.
1 2. Manual telegraph systems were in commercial use by the late 1 840s. The
Wheatstone system, devised in 1858, was probably the first to employ perforated
tape to operate the transmitting mechanism. Paper tape-driven automatic
telegraph systems were placed in commercial operation during the 1 860s and
were able to handle about 70 words per minute . Hollerith would, of course,
have been aware of these systems as is indicated by Austrian, 1 982: p. 1 3.
1 3. Hollerith, 7 August 1 91 9.
1 4. A. Hyman, 1 982 : Charles Babbage, Pioneer of the Computer (Princeton, NJ.:
Princeton University Press) , pp. 1 64-66, 254-55; M. Gleiser, September 1 979:
''The Loom of Lyons," Datamation, pp. 2 1 6-26.
15. Truesdell, 1965: pp. 38-39. The card used by the Baltimore Department of
Health had 192 hole locations, arranged in 6 rows of 32 locations each. Hole
locations were positioned near the edges of the card (3 rows on each of. the
two long edges) to permit punching with the short jaws of a hand-held punch.
16. T. C. Martin , 11 November 1891: "Counting a Nation by Electricity," The
Electrical Engineer 12, pp. 52 1-30.
1 7. J. S. Billings, 1887: "On Some Forms of Tables of Vital Statistics, with Special
Reference to the Needs of the Health Department of a City," Public Health Papers
and &PorL� 13, American Public Health Association, pp. 203-2 1 as quoted in
,

Truesdell, 1965: p. 32.

18. U.S. Patent 395,783, filed 23 September 1 884: H. Hollerith , "Apparatus for
Compiling Statistics"; U.S. Patent 395,782. originally filed 23 September 1884,
divided and filed 27 October 1885: H. Hollerith, "Art of Compiling Statistics."
1 9. U.S. Patent 395,781 , filed 8 June 1 887: H. Hollerith, "Art of Compiling
Statistics. "
20. Austrian, 1982: pp. 39-40.
21. Martin, 1 1 November 1891 .
22. U.S. Patent 487,737, filed 1 0 March 1 891 : H. Hollerith, "Keyboard-Punch."
Hollerith referred to the punch as a "key-board" punch because the large flat
plate, with holes into which the stylus was pressed. had a "key" printed near
each hole, "indicating the items to be punched into the cards." To further
reduce operator effort for the census work, Hollerith created a gang punch
that permitted a fixed group of holes in the first four columns to be punched
into as many as six cards at a time. These holes represented a given geographical
district and were identical for all data from the same district. See, for example,
Martin. 11 November 1 89 1 . Use of the gang punch for railroad cost accounting
applications is discussed in the American Machinist, 31 July 1 902: pp. 1073-75.

Copyrighted Material
 Notes to Pages 11-15 333

23. H. Hollerith, 1 9 Sep tember 1884: to Albert Meyer, Hollerith Papers, as

quoted in Austrian, 1982: p. 20.
24. Austrian, 1 982: pp. 20-22, 28.
25. Hollerith's three patents titled "Electro-Magnetically-Operated Air-Brake for
Railway-Cars" were filed on 30 April 1885 and issued on 12January 1886: U.S.
Patent 334,020; U.S. Patent 334,02 1; and U.S. Patent 334,022.
26. Hollerith's cousin in St. Louis was Henry Flad, president of the Mallinckrodt
Brake Company, which had been established to promote an improved brake
invented by John F. Mallinckrodt. Flad, who was then in his early sixties, had
distinguished himself as an engineer and was elected president of the American
Society of Civil Engineers in 1886.
27. Austrian, 1982 : pp. 27-35.
28. Austrian, 1 982: pp. 42-49. Hollerith's equipment wa� awarded one of the
five gold medals received by Americans at the exhibit in Paris.
29. Martin, 1 1 November 189 1 . The other two members of the commission were
Prof. Henry Gannett and L. M. E. Cooke. Other systems considered were ones
proposed by W. C. Hunt and C. F. Pidgen. As described by Martin: "Mr. W. C.
Hunt proposed to transfer the details given on the census enumerators' sched­
ules to cards, distinctions being made in part by the color of the ink and in
part by writing on them , the results being reached afterwards by hand sorting
and counting. . . . The next plan, that of Mr. C. F. Pidgin, contemplated the
use of "chips," which should be duly assorted and counted. These chips were
to be slips of paper of various colors embracing data printed in different colors
so as to indicate the readings of the schedules."
30. Austrian, 1 982: pp. 50-54,67.
3 1 . Martin, 1 1 November 1 89 1 .
32. L. Bodio, December 1893: The Inventive Age 4, reprinted in "A Forecast of
1893-and Its Fulfillment," The Tabulator (The Tabulating Machine Co.), August
1 9 1 6, p. 3. Luigi Bodio was director general of Italy's Imperial Bureau of
Statistics.
33. In August 1 894 the Census Bureau had finished its work and returned all
105 tabulators to Hollerith. On top of the huge loss of revenue was the added
cost of storing the equipment for an as yet unknown use. See Austrian, 1 982:
p l 04
. .

34. A. D. Chandler, Jr., 1 977: The Visible Hand: The Managerial Revolution in
American Business (Cambridge, Mass.: Belknap Press). pp. 1 09-2 1 Although
.

Hollerith's main commercial thrust was in railroads, the Prudential Life Insur­
ance Company had become interested in his equipment even before he won
the U.S. Census business. In the spring of 1890 Hollerith installed two machines
at the Prudential office� in Newark, New Jersey. See Austrian, 1 982: p. 83.
35. Austrian, 1 982: pp. 1 1 3, ]25-26.
36. Cards were fed, nines row first, between a conducting roller and wire
brushes, one per column, that made electrical contact with the roller through
the holes. Rotation of each wheel was begun when the corresponding brush
detected a hole in the card, and all wheels stopped rotating when the Os row

Copyrighted Material
334 Notes to Pages 16-22

was reached. Thus each wheel of the accumulator was advanced by the number
of digits represented by the corresponding hole in the card. The carry opera­
tion was handled within the accumulator by a spring-loaded mechanism that
advanced a wheel one position each time a lower-order adjacent wheel rotated
from the 9 to the 0 position. See A. H. Dickinson, 2 October 1 969: interview
by L. M. Saphire.
As early as January 1 887 Hollerith had filed for a patent on an integrating
tabulator. Issued in June 1 890, this patent describes a means for adding num­
bers rather than simply counting by ones. Intended for use with the stationary
pin-box card reader, it consisted of a motor-driven cylinder with nine strips of
conducting material of graduated lengths, parallel to the axis of the cyl inde r
and distributed uniformly about its circumference, plus nine electric-contact
brushes arranged parallel to the axis of the cylinder. When the cylinder rotated
one revolution, each brush made contact with as many of the strips of conduct­
ing material as it passed over-the brush at one end making contact with all
nine strips, and the others contacting progressively fewer strips as the strips
became shorter. When the automatic card feeder was introduced , the concept
of measuring the distance of a hole from the zero location to determine its
numerical value was a simple extension of this earlier method. See U.S. Patent
430,804, filed 4 January 1 887: H. Hollerith, "Electrical Calculating System."
37. Austrian, 1 982: pp. 97-98, 138, 1 53.
38. E. W. Byrn, 19 April 1 902: "The Mechanical Work of the Twelfth Census,"
Scientific American, p. 275.
39. M. Campbell-Kelly, 1 989: ICL: A Business and Technical Histury (Oxford:
Oxford University Press) , pp. 13-16.
40. Campbell-Kelly, 1 989: p. 8; Austrian, 1 982: p. 1 07.
41. Mrs. R. P. Porter, 5 December 1 �84: to Mrs. H. Hollerith, Hollerith Papers,
as quoted in Austrian, 1982: p. 108.
42. Campbell-Kelly, 1989: pp. 16-19.
43. Campbell-Kelly, 1989: pp. 19-30.
44.]. Connolly, 1967: History of Computing in Eurupe (New York: IBM World Trade
Corporation) , pp. 10-1 8. The businessman who successfully established the
German Hollerith Company, Dehomag, was Willy Heidinger, previously an
agent for adding machines. In 1 922 CTR obtained 90 percent of the stock of
Dehomag through a sale by Heidinger.

2 Origins of IBM
1 . G. D. Austrian, 1982 : Herman Hollerith, Fargotten Giant oj InJarmatioll Process i ng
(New York: Columbia University Press), pp. 4-5, 38, 47, 74-78.
2. Austrian , 1 982: pp. 54-59, 1 1 3- 1 4, 1 60-6 1 .
3. K. S. Reid-Green , February 1989: 'The History of Census Tabulation," Scien­
tific American, pp. 98-L03.
4. L. E. Truesdell, 1965: TlU' Development oj Punch Card Tabulation in the Bureau
of the Census, 1890-1940 (Washington, D.C.: U.S. Department of Commerce).
pp. 117-21; Austrian. 1 982: p. 227. S. N. D. North, who served as director of

Copyrighted Material
 Notes to Pages 22-28 335

the census from 1903 to 1909, was preceded by William R. Merriam from 1899
to 1903.
5. Newcomb, Churchill, and Frey, 1 April 1909: to Senator R. La Follette as
quoted in Austrian, 1982: p. 235.
6. Austrian, 1982: pp. 261-302.

7. H. Hollerith, 7 August 1919: to J. T. Wilson, as reproduced in the IBM

document, 1957: "Historical Development of IBM Products and Patents," pre­
pared by]. Hayward, pp. 18-23. Wilson was a former sales agent with the
Tabulating Machine Company.
8. C. R. Flint, 1923: Memories of an Active Life (New York: G. P. Putnam's Sons),
pp. 4-10, 85-114, 288-97.
9.A. D. Chandler, Jr., 1977: TIu! Visible Hand (Cambridge, Mass.: The Harvard
University Press), pp. 315-44.
10.Flint, 1923: especially pp. 244, 28�322. See also News/34-38, April 1985:
"IBM's Forgotten Founding Father," by Teresa Elms, pp. 64-65.
II. IBM Manual of Instruction, circa 1935: "Development of International
Business Machines Corporation," Educational Department, Endicott, N.Y
12. U.S. Patent 314,717, issued 31 March 1885:]. E. Pitrat, 'Weighing and Price
Scale."
13. U.S. Patent 393,205, issued 20 November 1888: W. L. Bundy, '"Time-Re­
corder."
] 4.
The card time recorder was based on U.S. Patent 528,223, issued 30 October
1894: Daniel M. Cooper, "Workman's Time Recorder." The dial time recorder
was based on a patent issued in 1888 to a Scottish physician, Alexander Dey.

15. ffiM Manual of Instruction, circa 1935. It has been noted somewhat humor­
ously that a fourth company involved in the merger, the Bundy Manufacturing
Company, is generally ignored because it contributed only the word "Company"
to the name.
16. E. A. Ford, 18July 1911: to H. Hollerith, as quoted in Austrian, 1982: p. 310.
17. Minutes, 15 July 1911: Fifth meeting of the board of directors of the
Computing-Tabulating-Recording Company.
18. S. Engelbourg, 1954: " International Business Machines: A Business History,"
Columbia University doctoral thesis, p. 44.
19.IBM Manual of Instruction, circa 1935. The first president ofCTR was Percy
H. Brundage.
20. ffiM received 100,000 shares of Hobart class B stock. The International Scale
Division (heavy-duty scales, etc .) was excluded from this sale, but in 1936 the
foreign subsidiaries were also transferred to Hobart. See special stockholder
meetings of 25 June 1935 and 20 July 1936.
21. Agreement, 21 October 1958: International Business Machines Corporation
and Simplex Time Recorder Company. The gross profit of ffiM's Time Equip­
ment Division (i.e., profit before deductions of expenses not distributed by
product) had dropped to $7.7 million in 1958, following a rise from $6.7 million

Copyrighted Material
336 Notes to Pages 28-34

to $9.5 million between 1954 and 1 957; see H. W. Trimble, Jr., 1 0 February
1959: to V. R Hansen, assistant attorney, General Antitrust Division.
22. Two other events in the company's early history should also be noted here.
Both occurred in 1933. IBM was reorganized from a holding company to an
operating company, and it entered the electric typewriter business through the
purchase of the Electromatic Typewriter Corporation of Rochester, N.Y See
Annual report, 1 933: International Business Machines Corporation.
23. The view that IBM was founded when T. J. Watson, Sr., became general
manager of CTR in 1 9 1 4 is broadly accepted. For example, the anticipated
announcement of the appointment of Louis V. Gerstner, Jr., as chief executive
of IBM was reported by the New York Times on 26 March 1 993, p. ID, as follows:
"Mr. Gerstner, 5 1 , who is chairman of the RJR Nabisco Holdings Corporation,
would be the sixth chief executive in the 79-year history of the International
Business Machines Corporation, and the first picked from outside the com­
pany's ranks. "

3 Watson: A Man with a Mission

1 . T. G. Belden and M. R. Belden, 1 962: The Lengthening Shadow (Boston: Little,
Brown) , pp. 93-94. See also T. J. Watson, Jr., and Peter Petre, 1 990: Father, Son
& Co. (New York: Bantam Press) , p. 1 5.
2. T. J. Watson, Jr., and Peter Petre, 1 990: Father, Son & Co. (New York: Bantam
Press) , pp. 1 1-1 3.
3. A. D. Chandler, Jr. , 1 977: The Visible Hand (Cambridge, Mass.: Harvard
University Press) , pp. 302-1 4. Integrated business structures for manufacturers
of machinery had first evolved in the sewing machine industry, beginning in
the 1 850s. By the 1 890s the Singer Sewing Machine Company had become fully
integrated, with its own timberlands, iron mills, and manufacturing plants, as
well as marketing, sales, and service organizations. It dominated its industry at
home and abroad.
4. Belden and Belden, 1 962: pp. 4-1 1 .
5 . Watson, Jr., and Petre, 1 990: p . 49. This is a story Watson, Sr. , chose not to
tell in his official biography; see Belden and Belden, 1 962: pp. 1 2-13.
6. Belden and Belden , 1 962: pp. 14-1 7.
7. Belden and Belden, 1 962: p. 33. For background info rination, see pp. 1 7-5 1 .
8 . R L. Palmer, 1 3 September 1 99 1 : interview by E . W. Pugh; Belden and
Belden, 1 962: p. 1 1 3.
9. Belden and Belden , 1 962: pp. 66-73.
1 0. Watson, Jr., and Petre, 1 990: p. 1 3.
1 1 . Belden and Belden, 1 962: pp. 84-87.
1 2. Watson, Jr., and Petre, 1 990: pp. 1 3-1 4.
1 3. Belden and Belden, 1 962: pp. 1 3, 87. Watson was no doubt also influenced
in his decision to ban alcohol at all business functions by the strong objection
Charles R Flint had to its use.
1 4. Watson, Jr., and Petre, 1 990: pp. 49,1 4 1 -42.
Copyrighted Material
 Notes to Pages 34-38 33 7

15. Belden and Belden, 1962: p. 98.

1 6. Thomas J. Watson was the top executive officer of IBM for forty-one years
(from 1 5 March 1 9 1 5 until 8 May 1956) when he turned over the position of
chief executive officer to his older son, Thomas]. Watson, Jr. The senior Watson
died of a heart attack a little more than one month later, 19 June 1956. During
those years IBM's gross revenue increased over 200-fold from $4 million to $892
million, the number of employees increased over 50-fold from 1 ,346 to 72,504,
and data processing, which accounted for less than 25 percent of the company's
revenue in 1 9 1 4, grew to 70 percent by 1 956. The revenue contributions in
1956 and 1914 are found in B. F. Wiegard, 26 March 1958: to D. Orton,
"Questions from Beldens."
17. Belden and Belden, 1962: pp. 126-36 and photo section. The words to
IBM's song, "Ever Onward," are as follows:
There's a thrill in store for all,
For we're about to toast
The corporation that we represent.
We're here to cheer each pioneer
And also proudly boast
Of that "man of men," our sterling president.
The name of T. J. Watson means a courage none can stem:
And we feel honored to be here to toast the I.B.M.
First Chorus
EVER ONWARD-EVER ONWARD!
That's the spirit that has brought us fame!
We're big, but bigger we will be,
We can't fail for all can see
That to serve humanity has been our aim!
Our products now are known in every zone,
Our reputation sparkles like a gem!
We've fought our way through-and new
Fields we're sure to conquer too
for the EVER ONWARD IBM.

1 8. A. H. Dickinson, 2 October 1969: inter view by L. M. Saphire.

19. H. Hollerith, 1 December 1 9 1 4: transcribed remarks to T. J. Watson as
quoted by Austrian, 1 982: p. 331 .
20. U.S. Patent 1,830,699, filed 1 1 March 1 9 1 4: Hollerith, "Automatic Control
for Tabulating Machines."
2 1 . Belden and Belden, 1 962: p. 1 1 2.
22. G. D. Austrian, 1 982: Herman Hollerith, Forgotten Giant of Informat ion Pror'!!ssing
(New York: Columbia University Press) , p. 346.

4 Building an Engineering Organization

1 . Business Machines, 1 January 1954: "40 Years of Pioneering," pp. 3-11; see also
Watson, Jr., and Petre, 1 990: pp. 1 4-1 5.
2. M. Campbell-Kelly, 1989: ICL: A Business and Technical History (Oxford: Oxford
University Press) , p. 34; Austrian, 1982: p. 271 . Powers emigrated to the United
States in 1 889 at age eighteen.
3. H. Hollerith, 4 January 1899: to S. G. Metcalf, as quoted in Austrian, 1982:
p. 270. Powers also providebp�lnW�e,m, operator keyed in all desired
338 Nous to Pages 38-44

holes before causing the device to punch all holes in a card at once. This
permitted operators to check for errors before punching any holes.
4. W. J. Eckert. May 1 963: "Early Computers." IBM Research News, pp. 7-8, says
Clair D. Lake told him this. See also A. H. Dickinson, 2 October 1969: interview
by L. Saphire.
5. Campbell-Kelly, 1989: p. 35. A photograph of a Powers connection box may
be found in the first group of photographs. between pages 52 and 53, of
Campbell-Kelly's book.
6. S. W. Dunwell, 31 July 1 99 1 : interview by E. W. Pugh.
7. G. F. Daly, 1 953: "Development of IBM Unit Record Equipment, Inventors
and Laboratories," p. 3.
8. Austrian, 1982: p. 331.
9. IBM Sales Training Manual, circa 1 956: "History and Development of Inter­
national Business Machines Corporation," Endicott, N.Y According to this
document: 'The development and research engineering [of CTR] in 1 914
consisted of one engineer and seven draftsmen at the Endicott plant who
worked on the development of new time-recording products, and one engineer
and three draftsmen in Uxbridge, Massachusetts, who were doing all of the
development work on the EAM line of products. The average hourly rate of
pay in the manufacturing plants in 191 4 was 32 cents and the work week was
54 hours. None of the benefits, such as insurance, hospitalization, and retire­
ment were in existence at that time; and air conditioning, fluorescent lights.
and maple floors were either unknown or unheard of for industrial use."
10. Daly, 1 953: pp. 3-4. Clair D. Lake had been appointed chief engineer of
the Tabulating Machine Division in 191 6 when E. A. Ford left the company for
a few years to develop his ideas on diesel engines. Lake retained that title in
Endicott.
1 1. IBM document, May 1922: "Suggested announcement for I.B.M. of Mr.
Bryce's appointment as Chief Engineer of Computing Tabulating Recording
Company." Bryce joined CTR in December 1 917.
1 2. Think, April 1 949: "The Light He Leaves Behind."
1 3. T. G. Belden and M. R. Belden, 1 962: The Lengthening Shadow (Boston:
Little, Brown), pp. 96-97.
14. S. Engelbourg, 1954: "International Business Machines: A Business History,"
Columbia University doctoral thesis, p. 270; Complaint, 21 January 1 952: U.S.
v. IBM, U.S. District Court, Southern District of New York, N.Y, p. 7.
1 5. U.S. Patent 1 ,830,699, filed 1 1 March 1 914: H. Hollerith, "Automatic Con­
trol for Tabulating Machines."
1 6. F. Thomas, circa 1957: "Interference Litigation between Hollerith, Peirce,
Tripp, and Powers, 1917 to 1 924." Thomas was a member of D. W. Cooper's
Law Office: Kerr, Page, and Cooper, the firm that represented the International
Time Recording Company and later IBM. This report was included as Chapter
11 in the IBM document, 1957: "Historical Development of IBM Products and
Patents" prepared by J. Hayward. As a result of the interference proceedings,
the following patents were issued: Tripp (U.S. 1,824,581 dated 9/22/31) , Hol-
Copyrighted Material
 Notes to Pages 44-52 339

lerith (U.S. 1,830,699 dated 1 1 /3/31 ) , and Peirce (U.S. 1 ,930,283 dated
1 0/10/33) .
17. Thomas, circa 195 7 .

1 8. Engelbourg, 1 954: p. 270; Complaint, 21 January 1 952: U.S. v. IBM, U.S.

District Court, Southern District of New York, N.Y, p. 7. The royalty payments
charged to Powers were reduced by half to 12.5 percent of gross rentals and 9
percent of gross receipts from the sale of cards. This reduction followed the
reorganization of the Powers company following receivership in 1922.
19. Belden and Belden, 1 962: pp. 116-17.
20. T.]. Watson, 1 June 1922: to]. E. Rogers, general manager, International
Time Recording Company, and others.
2l. E. W. Pugh, 9 March 1 993: to file. This memorandum analyzes information
prepared by Charles R. Doty for T. J. Watson, Sr. Doty's report describes all
projects in the Endicott Laboratory that had been abandoned without resulting
in a product, from the time when such records were kept until 1945. Aban­
doned projects would primarily be a subset of what today are called advanced
product development efforts. During the first three years for which comparative
data were available, Doty's report lists the cumulative cost of projects terminated
in 1928, 1 929, and 1 930, respectively, to have been (in thousands of dollars)
325, 59, and 1 86 for accounting machines and only 3, 22, and 14 for time
recording equipment.
22. E. A. Ford, 19 February 1924: to T.]. Watson.
23. T. J. Watson, 29 February 1924: to E. A. Ford.
24. E. A. Ford, 1 July 1926: to T.]. Watson. O. E. Braitmayer, 21 June 1 926: to
T. J. Watson explains the reason for the bonus and describes it as canceling
Ford's note for $3,500, which he borrowed when he returned to CTR, plus an
extra $1,500. Ford had left CTR in 191 6 to pursue his ideas on diesel engines.
This development was not successful, causing him to be in debt when he
returned to CTR; see Daly, 1953: p. 4.
25. W. W. McDowell, 29 June 1946: to Dwayne Orton.
26. F. M. Carroll, 25 April 1925: to T. J. Watson.
27. Daly, 1953: pp. 5, 1 3. Watson had also given this assignment to a third
inventor (from the Time Recording Division) who soon dropped out of the
contest when it became clear that Lake and Carroll had more promising
approaches.
28. S. W. Dunwell, 31 July 1 991: interview by E. W. Pugh.
29. G. F. Daly, 1 953: pp. 1 5-1 7; 1 2 August 1 969: interview by L. M. Saphire.
30. C.]. Bashe, L. R. Johnson, J. H. Palmer, and E. W. Pugh, 1 986: IBM's Early
Computers (Cambridge: MIT Press) , pp. 1 0-1 4.
3l. Campbell-Kelly, 1 989: pp. 82-83. This reference is particularly interesting
because it discusses the response of British companies to IBM's 80-column card
in addition to that of Remington Rand.
32. E.]. Rabenda, 20 February 1 992: discussion with E. W. Pugh.
33. Bashe et aI., 1 984: pp. 12-15.
Copyrighted Material
340 Notes to Pages 53-61

5 Responding to the Great Depression

1. Econostat, 1 April 1933: "National Cash Register Cleans Honse."
2. Bashe et aI., 1984: pp. 18-19; IBM sales manual, circa 1934.
3. American Machinist, 16 June 1 937: "From Inventor's Shop to International
Corporation," pp. 481-561; see especially pp. 487, 499. The vice president for
manufacturing in 1933 was Walter F. Titus, who previously had been the sales
manager in Detroit. The manager of engineering (more commonly referred to
as the laboratory manager) was Lawrence S. Harrison, who previously had been
in charge of IBM's customer engineering.
4. IBM News SDD Endicott, September 1 968: "The Future Depends on Engi­
neering," Special edition celebrating 35th anniversary of the North Street
Laboratory, p. 4. Endicott laboratory managers from 1933 to 1968, as listed in
the article, were as follows: L. S. Harrison, 1 933-38; G. H. Armstrong, 1938-42;
W. W. McDowell, 1942-50; E. W. Gardinor, 1950-5 1 ; E. J. Garvey, 1951-53;
J. A. Haddad, 1953-54; F. E. Hamilton, 1954-56; and J. J. Troy, 1 956- .

5. S. W. Dunwell, 31 July 1991: interview by E. W. Pugh.

6. IBM Document, 25 November 1957: responses prepared to questions posed
by the Beldens on 8 November 1957 for use in The Lengthening Shadow.
7. IBM Manual of Instruction, circa 1935: "Development of International Busi­
ness Machines Corporation," pp. 23-24.
8. G. F. Daly, 1953: "Development of IBM Unit Record Equipment, Inventors,
and Laboratories," pp. 8-10. Daly particularly credits C. D. Lake with training
many new engineers, including himself, J. M. Cunningham, F. E. Hamilton, W.
L. Lewis, W. W. McDowell, and R. E. Page.
9. American Machinist, 16 June 1 937: pp. 49 1 -92.
10. Business Machines, August 1935: "New Field Opened to College Women";
Kristina Brandt, 24 December 1991: discussion with E. W. Pugh; 27 January
1992: interview by E. W. Pugh. The vice president quoted first in the text is F.
W. Nichol, the second is O. E. Braitmayer. Information about Ruth M. Leach
may be found in Current Biography 1 948, pp. 373-74. In 1935 Glenn H. Arm­
strong was in charge of the education program, having previously served as
sales manager in Cleveland. In 1938 he replaced L. S. Harrison as laboratory
manager in Endicott. See E. J. Rabenda, 20 February 1992: discussion with E.
W. Pugh.
11. American Machinist, 16 June 1937: pp. 487-88.
12. American Machinist, 16June 1937: pp. 55�58; R. Miller, 27 May 1958: to the
Beldens in response to questions asked on 1 3 September 1957 for use in The
Lengthening Shadow.

13. American Machinist, 1 6June 1937: p. 549.

14. American Machinist, 16June 1937: pp. 549-55. The choice of manufacturing
tooling was based not only on the operations needed to make the various parts
with the specified tolerances but also on the quantities to be produced, the
materials used, the quality of the finish, and whether or not heat treatment was
required.
15. American Machinist, 16dB»y,J;gh1e�,r9r' 539-41.
 Notes to Pages 62-68 341

1 6. DunwelI, 31 July 1991.

1 7. F. J. Wesley, 19 May 1955: to L. H. LaMotte with attachment titled, "Accom­
plishment Summary-25-Year IBM Record of Mr. F. J. Wesley."
18. Business Machines, 20 October 1938: "An Interesting Business," by F. J. Wesley.
See also Wesley, 19 May 1955. Compendia of Pointer reprints were published
periodically. One of the last, "IBM Electric Punched Card Accounting Pointers
to Better Service," dated February 1 947, included in its preface an echo of the
1920s: "Some 45-column diagrams are still shown . . . because they may be easily
converted for 8O-column machines. "
19. Business Machines, 20 October 1938.
20. DunwelI, 3] July 1991.
2 1 . Wesley, 19 May 1 955.
22. David W. Rubidge made the changes described in the Social Security
collator. See S. W. Dunwell, 5 August 1991: interview by E. W. Pugh; D. W.
Rubidge, 1 1 November 1 991: discussion with E. W. Pugh. Rubidge modified the
wiring of the single comparing mechanism, which was capable of handling
numbers with up to 16 digits, into two portions of 9 and 7 digits each. He
assigned the 9-digit portion to making high. low, or equal comparisons on the
9-digit Social Security numbers, as was needed for collating control. The 7-digit
portion was used for making an equals comparison on the numeric portion of
the name field to ensure that the names on two cards corresponded when the
Social Security numbers were equal. If they failed to correspond, the card in
the secondary deck was selected by the collator for manual review.
23. Business Machines, 20 October 1 938.
24. Clarence W. Avery, the uncle of S tephen W. Dunwell, was inducted into the
Automotive Hall of Fame in Midland. Mich in 1990. The plaque honoring
.•

Avery reads in part: "Clarence W. Avery (l882-1949)-Credited with develop­

ing the moving assembly line in the mass production of automobiles for the
Ford Motor Company. . . . This development sparked the beginning of mass
production in the automotive industry."
25. Dunwell, 3 1 July 199 1 . Dunwell was also given the task of finding ways to
reduce the radio interference caused by IBM products.
26. S. W. Dunwell, 1 5 August 1 99 1 : interview by E. W. Pugh.

6 Support for Academic Research

1 . L. J. Comrie, 1 928: "On the Construction of Tables by Interpolation," Royal
Astronomical Society Monthly Notices 88, pp. 50�23. See also H. H. Goldstine,
1972: The Computer from Pascal to von Neumann, p. 1 07.
2. G. W. Snedecor, 1 928: "Uses of Punched Card Equipment in Mathematics,"
American Mathematical Monthly 35, pp. 1 6 1 -69.
3. J. F. Brennan, 197 1 : The IBM Watson Laboratory at Columbia University-A
Histor y (Armonk. N.V.: IBM), pp. 3-4. Wood remained an IBM consultant for
twenty-eight years.
4. D. R Piatt, 19 December 1949: to W. J. Eckert, "Statistical Calculator"; Bashe
et aI., 1 986: p. 23.
Copyrighted Material
342 Notes to Pages 69-73

5. Bulletin of Information on the International Test Scoring Machine, October 1936:

published by the Cooperative Test Service of the American Council on Educa­
tion.
6. R B. Johnson, 19 May 1 977: discussion with C. J. Bashe et al. as summarized
on 10 June 1977. Among considerations that caused the IBM board to decide
against buying Johnson's test-scoring machine concept was the Patent Depart­
ment's conclusion that Lasker, of Remington Rand, had invented pencil-mark
sensing before Johnson.
7. IBM Document, circa 1944: Employment Record of R B. Johnson.
8. T. J. Watson, 16 March 1 936: to R B. Johnson.
9. Bulletin of Information on the International Test Scoring Machine, October 1936.
10. IBM Document, circa 1 944. When company salesmen warned Watson that
the Test Scoring Machine would never make money, Watson reportedly re­
sponded, "Who wants to make money out of education?"; see Brennan, 1 971 :
p. 5.
1 1 . News Release, 6 January 1946: Columbia University Public Information
Office.
1 2. W. J. Eckert, January 1940: Punched Card Methods in Scientific Computation,
published by the T. J. Watson Astronomical Computing Bureau, p. 77. John C.
McPherson in a letter of 6 January 1 982 to E. Nanas says: "Eckert's calculator
was an early and primitive 'computer' using punched cards for storage of tables
and intermediate calculated results which required a person to move the cards
between the punched card accounting machine, summary punch and punched
card multiplier, directed by a manual program controller. It was used to carry
out highly precise calculations of planetary orbits and other differential equa­
tions. "
13. W. J. Eckert, May 1963: "Early Computers," IBM Research News, pp. 7-8.
14. H. Aiken, circa 1 937: "Proposed Automatic Calculating Machine, " reprinted
in IEEE Spectrum, August 1964: pp. 62-69. The origin of this memorandum by
Aiken is described in Spectrum as follows: "A copy of the unpublished memo­
randum in which Aiken first outlined his conception of Mark I was recently
discovered in Professor Aiken's Harvard files by Mrs. Jacquelin Sanborn Sill,
and through her efforts is presented here with the author's permission. The
memorandum itself is undated, but an unknown recipient's handwritten nota­
tion 'Prospectus of Howard Aiken, November 4, 1 937' puts an upper bound on
the date for its preparation." The date of receipt by the "unknown recipient"
and some specific references to IBM equipment indicate the document was
intended for presentation to IBM. Aiken had received a bachelor's degree in
electrical engineering from the University of Wisconsin in 1 923. From then to
1 932 he held various jobs and studied physics at the University of Chicago. He
received the M.S. in physics in 1 937 and the Ph.D. in 1939 from Harvard
University.
15. I. B. Cohen, July 1980: Foreword to "History of Mechanical Computing
Machinery" by George C. Chase, in Annals' of the History of Computing 2, pp. 1 98-
200.
1 6. H. Aiken, 3 November 1937: to J. W. Bryce.
Copyrighted Material
 Notes to Pages 74-78 343

1 7. H. E. Pim, 7 January 1 938: to J. w. Bryce, "Howard Aiken, Harvard

University. "
18. F. E. Hamilton, 2 1 August 1944: untitled report describing development
history of the Harvard Machine, which was transmitted to C. D. Lake by J. J.
Robbins on 28 January 1949.
19. L.J. Comrie, 1946: "Babbage's Dream Comes True," Nature 158, pp. 567-68.
See also S. Rosen, 1969: "Electronic Computers: A Historical Survey," Computing
Surveys 1, pp. 7-36. At about the time Howard Aiken began promoting con­
struction of the large Mark I supercomputer, Konrad Zuse in Berlin, Germany,
had begun work on a much smaller unit with a mechanical memory, arithmetic
and control units made of secondhand telephone relays, and program control
by perforated tape ( 35 mm movie film). His computing device is said to have
been operational in 1941 . Because it was destroyed during the war, however,
there are numerous unanswered questions. With a reported completion date
approximately two years before the Mark I, Zuse's machine is credited as a "first"
by many historians. It is a rather limited first, however, since the device was too
small to do useful work. See, for example, P. E. Ceruzzi, 1 983: Reckoners (West­
port, Conn.: Greenwood Press), pp. 10-40. In contrast, the Calculation Control
Switch designed by Wallace Eckert, and operational by 1936, made it possible
for standard electromechanical punched-card equipment to operate in a se­
quence-controlled mode similar to that of Zuse's machine and the Mark I. The
Calculation Control Switch was used extensively by Eckert for his work in
astronomy, beginning at least four years before Zuse's device is said to have
been completed.
20. I. B. Cohen, 1 988: "Babbage and Aiken," Annals of the History of Computing
10, pp. 1 7 1 -93, especially pp. 181-83.
2 1 . IBM brochure, 1 945: "IBM Automatic Sequence Controlled Calculator." See
also C.J. Bashe, 1 982: "The SSEC in Historical Perspective," Annals of the History
of Computing 4, pp. 296--3 1 2. Francis E. Hamilton, who reported to Lake, was
the engineer responsible for day-to-day direction of the project. Another major
contributor was IBM engineer Benjamin M. Durfee.
22. Cohen, 1988: especially p. 178.
23. U.S. Patent 2,61 6,626, issued 4 November 1 952: Clair D. Lake, Howard H.
Aiken, Francis E. Hamilton, and Benjamin M. Durfee, "Calculator."

7 Research for Patents and Devices

1 . EAM Patent Digest with Use Status, 16 February 1950: an IBM engineering
report obtained from the Endicott laboratory in 1981 with the help of J. H.
Wellburn.
2. IBM Document, circa 1 953: "Bryce Patents by General Subject." This docu­
ment was found in the files of George F. Daly as part of his materials on IBM
patents. Although undated, the last patent on the list is U.S. 2,641 ,976, dated
06-1 6-53, and titled, "Photographic Color Recording Machine." A tabular sum­
mary of data in this document is provided by E. W. Pugh and C. C. Coppola,
22 April 1991: "Bryce's patents."

Copyrighted Material
344 Notes to Pages 78-87

3. Think, April 1949: "The Light He Leaves Behind." See also E. W. Pugh and
C. C. Coppola, 15 and 22 April 1991 : to file; these two memoranda provide
tabular summaries of data in the documents, "IBM Patent Digest of 1952" and
"Bryce Patents by General Subject."
4. IBM document, circa 1 949: "James Wares Br yce. " This undated and unsigned
document was evidently created at the time of the April 1949 article in Think
and contains identical wording plus additional information.
5. E. J. Rabenda, August 1 969: interview by L. M. Saphire.
6. D. W. Rubidge, 11 November 1 99 1 : discussion with E. W. Pugh.
7. Business Machines, 4 April 1 949: "Mr. J. W. Bryce, IBM Pioneer, Inventor, Chief
Consultant, Technical Staff Dean, Dies."
8. Financial Bulktin Service of Dow, Jones & Co., 5 May 1 930. Bryce's assistant,
who replaced him as head of the Patent Department, was W. M. Wilson.
9. A. H. Dickinson, 2 October 1 969: interview by L. M. Saphire.
10. J. w. Bryce, 26 January 1939: to T. J. Watson.
II. B. E. Phelps, 1 980: "Early Electronic Computer Developments at IBM,"
Annals of the History of Computing 2, pp. 253-67.
12. R. L. Palmer, 13 September 1 99 1 : interview by E. W. Pugh. Among the
reasons Palmer accepted the IBM job offer was the caliber of those who
interviewed him, especially Clair D. Lake, who impressed him very much at the
time and in the years that followed.
13. R. L. Pal mer, 25 July 1967: interview by L. M. Saphi re.
14. H. J. Reich, August 1 939: "Trigger Circuits." ElRctronics, pp. 1 4- 1 7; W. H.
Eccles and F. W. Jordan, 1 9 1 9: "A Trigger Relay Utilizing Three-Electrode
Thermionic Vacuum Tubes," Radio Review 1, pp. 1 43-46.
1 5. A. W. Burks, 1947: "Electronic Computing Circuits of the ENIAC," Proceedings
of the Institute of Radio Engineers, pp. 756-57; Engineering Research Associates
Inc. staff, 1950: High-SPeed Computing Devices (New York: McGraw-Hili) , pp. 1 2-
31.
1 6. U.S. Patent 2,584,811, filed 27 December 1 944: B. E. Phelps. "Electronic
Cou n ti ng Circuit"; B. E. Phelps, 9 December 197 1: 'The Beginnings of Elec­
tronic Computation," IBM Technical Report.
17. Phelps, 1980. The number of additions and the positioning of the column­
shift mechanism for multiplication were determined by the number in the
multiplier register.
1 8. G. H. Armstrong, 21 May 1942: to J. V. Atanasoff.
1 9. M. R. Williams, 1985: A History of Computing Technology (Englewood Cliffs,
NJ.: Prentice Hall), pp. 266-71.
20. S. Rosen, 1990: "The Origins of Modern Computing," Computing Retliews 31,
pp. 449-62. See also the "Responses" by H. H. Goldstine, pp. 463-64; A. W.
Burks and A. R. Burks, pp. 464-67; J. v. Atanasoff II, pp. 467-69; N. Stern,
pp. 469-70; B. A. Galler, pp. 470-71; C. R. Mollenhoff, pp. 471-74; A. R. Mack­
intosh, pp. 474-75; E. A. Weiss, pp. 475-7 7 ; S. Rosen, pp. 477-80.

Copyrighted Material
 Notes to Pages 89-90 345

8 World War H Activities

1 . A. L. Norbe rg , 1 990: "High-Technology Cal c ulati on in the Early 20th Cen tury :
Punched Card Mach in ery in B usi ne ss and Govern me n t,» Technology and Culture
31, pp. 753-79, especially 764, 771 .
2. Belden and Belden, 1 962: pp. 1 90-2 1 7.
3. The Munitions Manufac turin g Corporation was incorporated on 31 March
1 941 . On 1 April 1941 F. H. M. Hart was elected president, and on 7 April 194 1
the 21 5 acres o f land i n Poughkeepsie were purc hased for $201,546. Under
Hart's le adersh ip the small food cannery was con verted to a munitions fac tor y,
the pac kin g plant into offices, and later the residence into a cafeteria . By
February 1 942 a new 1 40,00o..square-foot manufac turi ng building (00 1 ) was
completed and over 250 employees had been hired. On 1 3 July 1 942 the
Munition s Manufacturing Corporation was absorbed by IBM and i ts facilities
were desi gn ated IBM Plant No. 4. IBM P lant No. 1 was in Endicott, Plant No.
2 was in Washingto n , D.C., Plant No. 3 was in Rochester, and Plant No. 5 was
in Toronto. See Business Machines, 16 April 1 942, "IBM E mp loyee s at
Poughkeepsie Hear Mr. Watson," and IBM documen t, 4 Oc tobe r 1 948: "Con­
tractor 's Re por t on Ordnance Manufacture in Worl d War II-Suppleme n t
Number One, " pp. 9-10. In September 1 946 IBM pl an ts in the U.S. were
re de sign ated , with the Pough keepsi e plant changing fro m Plant 4 to Plant 2.
The num berin g of company plants lost favor and was discontinued some years
later. In 1 942 IBM purchased 123 acres on Route 9, j ust south and east of the
main p l ant. to be used for e mp loyee recreation . In 1944 IBM purch ase d
Cliffdale, a 2 1 7-acre estate on Boardman Road east of the main plan t where
research and development facilities were e stablished. See IBM booklets, "Page s
from the Past" and "50 years in Pough keepsie, " produced in 1976 and 1991 ,
respectively, for the 35th and 50th anniversaries of the Poughkeepsie facilities.
4. Business Machines, 28 Sep tem ber 1 950: "Ordnance Group is Told of IBM
Production Plans," p. 1 ; IBM document, 18 December 1 950: "Research and
Developmen t Facilities," p. 6.
5. The 1 112 percen t constraint on profits voluntarily set by IBM on its munitions
produc tio n has been widely quoted as 1 p erc en t since at least as early as 1 9 76
when the official IBM corporate publication , "IBM . . . Yesterday and To day, "
so stated it. Inte rn al documen ts of the immediate postwar p eriod, however, give
the number a.� 1 112 percen t. An IBM docum en t, circa 1948, "A Resum e of the
Activities of Mr. Th o mas J. Watson from 1 939 to 1 945, " gives the most compre­
hensive discussion of this policy: "When IBM, in April 1 94 1 , entered into its
fi rst contract for the pro duc ti on of munitions, sligh tly more than a year before
the enactment of the Renegotiation Act, Mr. Watson , with the approval of the
Board of Directors, volun tarily established for IBM a policy of n ot acceptin g
any profits on war con tracts which could in any way be regarded as unreason­
able . . . . In conformity with this policy, through forward price redu c tio ns and
refunds, voluntarily made, IBM has limited its over-all profits on its war co n trac ts
with the government to approximate ly 1 112% net after taxes, which self-limita­
tion has always been satisfactor y to the gove rnme n t. " The 1 1/2 percen t number
is also found in the official biography of T. J. Watson ; se e T. G. Be lden and
M. R. Belden, 1 962: The Lengthening Shadow (Boston: Little, Brown) , p. 210.
Copyrighted Material
346 Notes to Pages 90-92

6. Business Machines, 22 January 1 942: "IBM Engineering Staff Is Devoting Entire

Time to Nation's War Program."
7. IBM document, 30 June 1948: "Contractor's Report on Ordnance Manufac­
ture in World War II-Section on Engineering, Research, and Development, "
submitted by the IBM Corporation to the Chief of Ordnance, United States
Army, War Department, Washington, D.C. According to the document, it was
"Prepared under the general direction and supervision of the late Mr. C. A.
Kirk, Executive Vice President and formerly Vice President in Charge of Manu­
facturing, and Mr. J. C. McPherson, Vice President in Charge of Engineering,
New York, N.Y " Among the ninety-nine separate research and development
projects were the development of mobile machine record units, punched-card
teletype systems, the Radiotype system for radio transmission of typed messages,
an array processor for deciphering intercepted enemy messages, a relay calcu­
lator for ballistics calculations, synthetic training devices (or simulators) for
pilots, a scrcalled data repeater that used multiple cathode ray tubes to display
relative altitudes and other information on hostile and friendly aircraft, ground
and aerial odographs, portions of the analog computer and fire control system
for the B-29 bomber, and a mechanical grenade thrower.
8. F. J. Wesley, 1 9 May 1955: to L. H. LaMotte with attachment titled, "Accom­
plishment Summary-25 Year IBM Record of Mr. F. J. Wesley. "
9. Business Machines, 1 9 February 1 942: "Department of Logistics is Formed to
Coordinate All IBM Activities. " The intimate knowledge of foreign industrial
methods and resources that Watson ascribed to F. W. Nichol were undoubtedly
acquired when Nichol was in charge of the company's business abroad, an
assignment he was given in 1 930 when IBM's products were in use in 77
countries; see Financial Bulktin Service of Dow, Jones & Co., 5 May 1930.
10. Business Machines, 26 March 1 942: 'We Must Think in Terms of We-All for
Victory, President Watson Tells Members of Organization in Address Heard in
All Company's Factories in U.S. "
1 1 . Business Machines, 28 September 1950. The term EAM was increasingly used
to describe all of IBM's punched-card equipment following the change in the
name of the responsible division from Tabulating Machine Division to Electric
Accounting Machine Division in late 1937. A survey of Business Machines during
the late 1930s and early 1940s reveals that the division was called the Tabulating
Machine (TM) Division in September 1 937, the Electric Accounting Machine
(EAM) Division in November 1 937, the Electric Bookkeeping and Accounting
Machine (EBAM) Division in January 1 938, and again the Electric Accounting
Machine (EAM) Division from April 1 939 onward; see L. R. Johnson, 28
February 1 985: to file.
1 2. Business Machines, 26 March 1 942. Watson's prophecy of increased produc­
tion by American industry was fulfilled as is described in a retrospective article
in Think Magazine 's Diary of u. s. Participation in World War II, 1950: 'The Miracle
of Production , " p. 1 85. "Global war demanded a miracle of production . . . .
The number of persons engaged in manufacturing establishments jumped from
10,400,000 at the start of the defense program Uuly 1940] to 1 7,200,000 at the
peak of the war effort [November 1 943] . Of this latter number 1 0,300,000
persons were engaged in munitions work and 6,900,000 in nonmunitions manu-

Copyrighted Material
 Notes 10 Pages 93-96 34 7

facturing. The work week was lengthened so production could increase. Work­
ers' hours rose from an average of 37.4 per week to 45.5, though many worked
longer hours in critical fields. Weekly earnings naturally increased, rising from
an average of $25.52 to an average of $46.00 per week at the end of the war
[August 1945 ) . " The increase in worker hours by 1 0 1 percent, combined with
improved tools and a spirit of cooperation, increased total output by 1 1 7
percent.
1 3. IBM Document, 30 June 1 948. Each Mobile Machine Records Unit Detach­
ment was a self-contained military unit. A typical detachment comprised two
I O-ton trailer vans, one of which was desig,led for administrative operations and
the other to hold most of the EAM equipment. There were also two 2 1/2_ton,
six-wheel drive trucks, two electric generators, and one command car. The
personnel of a mobile records detachment at full strength consisted of twenty­
nine enlisted men and three officers, all of whom had received basic military
training as well as accounting machine instruction. Many mobile units were
commanded, operated, and serviced by former IBM employees who had en­
tered the military service.
1 4. IBM Document, 30 June 1 948.
15. C. R. Doty, 15 December 1 960: "IBM Tele-Processing Systems and Equip­
ment," IBM Technical Report, p. 1 1 .
1 6. T. J. Watson, president of IBM, and Walter S. Gifford, president of AT&T,
agreed that if any patents were involved between IBM and AT&T, rights there­
under would be waived for the duration of the emergency. Eight U.S. patents
were subsequently issued to IBM engineer C. R. Doty on this equipment and
modifications thereof; see Doty, 15 December 1960.
1 7. IBM Document, 30 June 1 948.
1 8. Doty, 15 December 1960, reports the following: "The excellent results
obtained by the Air Corps with these machines led to their use during the war
by various other branches of the Armed Services, such as the Quartermaster
Corps, Signal Corps, the Army Ordnance Department, the Army Transportation
Corps, and Naval Air Depots. Also. the Office of Defense Transportation , the
U.S. Treasury Department, the Weather Division of the Army Air Forces, the
U.S. Weather Bureau and others." The volume of punched-card data transferred
from point to point by telegraph during the last two years of the war was
estimated to exceed 4 million cards per month.
1 9. IBM Document, 30 June 1 948. Brigadier General F. W. Stoner was the chief
of army communications.
20. Walter Lemmon, while at Columbia University, had invented the means by
which radio circuits could be trimmed so that all tuning condensers could be
on a single shaft. Prior to his invention, the better radios had several dials that
had to be aligned individually. Lemmon got a royalty on nearly every radio
produced and was quite well off financially. See S. W. Dunwell, 28 May 199 1 :
"Recollections of the IBM Radiotype. "
2 1 . Cue, 2 9 March 1941 : "People," p. 32.
22. Rochester, N.Y, newspaper, 1 July 1933: "Typewriter Plant Sold to XY Firm. "
IBM purchased the assets o f Electromatic for 2, 105 shares of IBM stock; assumed
Copyrighted Material
348 Notes to Pages 96- 99

the entire preferred stock ($72,000 par value) and guaranteed payment of a 6
percent dividend from July 1 , 1 934; agreed to pay a royalty on the net sales of
typewriters at the rate of 5 percent on such sales for the period including 3 1
December 1 938; and assumed the lease o n the factory building a t Rochester;
see R. S. Hansen, 28 May 1 958: to A. R. Miller, 'Watson Biography." The
resulting International Electric Writing Machine Division operated as an inde­
pendent business unit; it manufactured, sold, and serviced all its products; see
American Machinist, 16 June 1937: "From Inventor's Shop to International
Corporation," pp. 559-64.
23. The possibility of entering the typewriter business had been discussed at
least as early as 1 928 when Bryce considered using electric typewriter elements
as the basis for EAM input-output equipment and in 1 929 when Watson pon­
dered the benefits of typewriters that could aiso perform arithmetic. J. W. Bryce,
25 January 1 928: to T. J. Watson; F. J. Tillman, 3 January 1 929: to T. J. Watson;
J. W. Bryce, 17 January 1 929: to T. J. Watson .
24. IBM Document, 3 0 June 1948. The Radiotype consisted o f several units and
used holes punched in paper tape for temporary message storage. The first unit
was a transcription electric typewriter on which the message was typed by an
operator. On the underside of the typewriter was mounted a permutation unit
that translated the keyed-in letters and numerals into code by closing one or
more of six electric circuits. These circuits operated magnets that caused
punches to perforate combinations of up to six holes in the tape. Transmission
was automatic, with perforated tapes being fed through a transmitter unit at a
speed of 600 ch arac ters a minute. If messages were few, the permutation unit
could send its signals directly to the transmitter, thus bypassing the tape perfo­
rating and reading steps. For receiving messages, similar units were employed.
The Radiotype system was sold by IBM late in 1945 to Globe Wireless, Ltd.,
with headquarters in New York City. Among the considerations that favored
selling the business was that the Radiotype business would place IBM in com­
petition with one of its best customers, AT&T. The potential loss in AT&T
business appeared to outweigh any possible gains from marketing Radiotype.
As it turned out, Radiotype was not a particularly successful product. See R. G.
Canning, circa 1990: "Early Days of Electronics," an unpublished six-page manu­
script. The use of Radiotype coupled directly to encryption and decryption
equipment, is reported by J. C. McPherson, 8 August 1 99 1 : discussion with
E. W. Pugh.
25. S. W. Dunwell, 28 May 1 991 : "Recollections of the IBM Radiotype."
26. James Greene volunteered with Dunwell and served as manager of the
operation much of the time. Dunwell served as the technical director. At the
end of the war, Dunwell was put in charge of the entire punched-card machine
operation, which then employed about 1 ,400 people. See S. W. Dunwell, 1 5
August 1 991 : interview by E. W. Pugh, pp. 4-8.
27. S. W. Dunwell, 1 February 1 992: "A description of one way in which
punched-card machines with attached relay calculators were used for code­
breaking during World War II." In describing this common cryptographic
method, Dunwell noted that the "carry" was omitted when the key digit was
added to the message digit; e.g., 9 plus 7 was represented as 6 rather than 16.

Copyrighted Material
 Notes to Pages 102-1 10 349

28. S. W. Dunwell, 1 5 August 1 991 : interview by E. W. Pugh; Dunwell, 1 February

1992. Among the modified units ordered by Dunwell was a machine that could
read information from a card and punch the result of a computation on that
information in the same card at the speed of 1 00 cards per minute-a capability
not previously available. The Type 600 multiplying punch introduced in 1931
could take two factors from a single card, multiply them, and punch the result
in a blank field of the same card; however, this was at a much slower speed; see
chap . 4.
29. Dunwell, 15 August 1 99 1 .
30. R. L . Palmer, 1 3 September 1 99 1 : interview by E . W. Pugh .
3 1 . David Kahn, 199 1 : Seizing the Enigma (Boston: Houghton Mifflin) , pp. 31-
48, 245.
32. M. R. Williams, 1 985: A History of Computing Technology ( Englewood
Cliffs, N J.: Prentice-Hall) , pp. 287-90.
33. Kahn, 1 99 1 : pp. 49-8 1 , 99, 231 . For more information on the role of the
British Tabulating Machine Company in building bombes and other military
devices, see M. Campbell-Kelly, 1 989 : ICL: A Business and Technical History,
pp. 1 15-23.
M. Decryption of German messages was of course only one of many factors
that helped win the Battle of the Atlantic. The introduction of shorter wave­
length (9. 7 versus 1 ,500 centimeter) radar by the Allies was another. The various
factors that affected the outcome are discussed by Kahn, 1991 ; see in particular,
pages pp. 239-40, 259, 274-75.
35. Palmer, 13 September 1 99 1 .
36. E . Tomash and A . A. Cohen, 1 979: 'The Birth o f an ERA: Engineering
Research Associates, Inc. , 1 946-1955," Annals' of the History of Computing 1,
pp. 83-97.
37 . Palmer, 13 September 1 99 1 .

9 Fubwe Demands
1. mM News SDD Endicott, September 1968: "North Street Laboratory Cele­
brates 35th Anniversary," pp. 1-4.
2. S. W. Leslie, 1 983: Boss Kettering (New York: Columbia University Press) ,
pp. 1-37 ; T. G. Belden and M. R. Belden, 1 962: The Lengthening Shadow ( Boston:
Little, Brown) , p. 1 1 3. Charles F. Kettering was born in Loudonville, Ohio, in
August 1 876. He died in November 1 958. His partner in founding Delco was
Edward A. Deeds.
3. Saturday Evening Post, 12 August 1 933: "Boss Kettering, " by Paul de Kruif, as
quoted in Leslie, 1 983: p. 36.
4. Leslie, 1 983: p. 49.
5. T. J. Watson, 1 June 1 922: to J. R. Peirce. Among the members of the Future
Demands Committee who received identical letters were S. M. Hastings (presi­
dent, Computing Scale Company) , J. R. Peirce (of Case Peirce Accountin g

Copyrighted Material
350 Notes to Pages 1 10-1 1 6

Machine Company) , O. E . Braitmayer (assistant general manager, Tabulating

Machine Company) , and Clement Ehret (assistant to the general manager,
Tabulating Machine Company) .
6. IBM Manual of Instruction, circa 1 935, p. 5; Austrian , 1 982: p. 337.
7. The role of Otto E. Braitmayer is inferred from references to him or letters
from him in the files of T. J. Watson. His title in 1 922 is given in the letter of
1 June 1 922 in which Watson appointed him a member of the Future Demands
Committee.
8. Business Machines, 26 September 1 949: "Mr. Clement Ehret, Head of Market
Research, Dies. " Ehret had joined the company as a sales representative in 1 9 1 6
and became assistant to the general manager o f the Tabulating Machine Divi­
sion in 1 9 1 9. Two years later he became treasurer of the division and soon
thereafter sales manager.
9. C. Ehret, 1 9 February 1927: to T. J. Watson.
10. J. W. Bryce, 1 7 October 1 928: t? T. J. Watson.
1 1 . Sales Managerrumt, 6 April 1 929: "International Business Machines Inaugu­
rates a Future Demands Department," C. Ehret as told to A. R. Hahn, pp. 9-10.
This publication describes itself as "the weekly magazine for marketing execu­
tives. "
1 2. Business Machines, 26 September 1 949.
1 3. IBM document, May 1 955: "Accomplishment Summary," written by F. J.
Wesley with annotations by L. H. LaMotte and M. B. Smith.
14. S. W. Dunwell, 31 July 199 1 : interview by E. W. Pugh.
1 5. Business Machines, January 1 940: "Mr. McPherson Named to Head Future
Demands Department"; J. C. McPherson, 1 1 December 1 990: interview by E.
W. Pugh. Interestingly both J. C. McPherson and F. J. Wesley, who had headed
Commercial Research since 1 937, attended IBM Sales Class 56 in Endicott.
1 6. Special machine requirements had required only occasional engineering
support during the 1920s, but by the time the North Street Laboratory was
established, special equipment design had become one of two primary duties
of the Production Engineering Department. G. F. Daly, 1953: "Development of
IBM Unit Record Equipment, Inventors, and Laboratories," p. 1 2.
1 7. W. J. Eckert, 1948: ''The IBM Pluggable Sequence Relay Calculator," in
Mathematical Tables and Other Aids to Computation 23, published by the National
Research Council, pp. 1 49-6 1 ; H. H. Goldstine, 1 972: The Computer from
Pascal to von Neumann, pp. 1 27-30.
1 8 . The manager of Engineering, Glen H. Armstrong, was brought from the
Endicott laboratory to headquarters in 1 942 to conduct a reduced new products
effort now named Special Research; see Business Machines, 1 6 July 1942: "Mr.
Armstrong Heads New IBM Research Department. "
19. J . C . McPherson, 1 1 December 1990: interview by E . W. Pugh. See also J.
C. McPherson, 1 7 June 199 1 : discllssion with E. W. Pugh. McPherson's view that
Watson was the "chief engineer" takes on greater significance when one recalls
that the titled position of chief engineer remained unfilled after Bryce vacated
it in 1 926 to establish and manage the Patent Department.
Copyrighted Material
 Notes lo Pages 1 1 7-122 351

10 Preparing for Peace

1 . Business Machines, 28 Se p te mber 1 950: "Ordnance Group Is Told of IBM
Production Plans. "
2. IBM document, 1 0 J u ne 1943: summary of "Meeting of Engineers" called by
T. J. Watson , Sr. , a t 590 Madison Avenue, New York City; IBM document, 6 July
1943: summary of meeting in Mr. Watson's office with J. C. McPherson, K. J.
Mackenzie, and G. H. Ar m stron g .
3. IBM document, 1 8 June 1 943: summary of "Engineering Meeting called by
Mr. Watson. "
4. IBM document, 1 8 June 1 943: pp. 6-9.
5. The effort to develop an accounting machine for small businesses based on
the Electromatic typewriter was managed by R L. Hou ston in Endicott from
about 1 943 to 1 947. It was abandoned after an expendi ture of approximately
$33,000. The machine was designed to operate from three i n pu t card decks.
1
The small, 30-column cards ( 2 5h 6" x 3 /2") , were to be punched with a six-hole
code. The machine would duplicate or summary punch a new deck, or punch
a tape that could be mailed to a remote point for automatic reproduction into
standard 8O-column cards. All machine operations were controlled by a "se­
quence card" instead of the conventional control panel. See unda te d report on
'Typewriter Developments," prepared for Watson, Sr. , by C. R Doty, listing
research and development projects terminated at Endicott.
6. IBM document, 1 8 Ju n e 1 943: pp. 12- 1 3.
7. C. ]. Bashe , L. R Johnson, ]. H. Palme r, and E. W. Pugh, 1986: IBM's Early
Computers ( Cambridge , Mass.: MIT Press) , p. 481 .
8. J. C. McPherson, 1 1 December 1 990: interview by E . W. Pugh . The ac tual
monthly rental of the IBM 407 Accounting Machine ranged from $800 for the
least expensive to $920 for the most expensive as listed in a 1 957 sales manual.
9. Bashe Johnson , Palmer, and Pugh, 1 986: p. 482.
,

10. E. W. Pugh, L. R Johnson, and ]. H. Palmer, 1 99 1 : IBM's 360 and Early 3 70

Systems (Cambridge, Mass.: MIT Press) , pp. 574-75; H. S. Beattie and R A.
Rahen kam p 1 98 1 : "IBM Typewriter I nnovati on , IBM Journal of Research and
, "

DevelQ/JTTIfflt 25, pp. 729-39; F. T. May, 1 98 1 : "IBM Word Processing Develop­

ments," IBM Journal of Research and DevelopTTlfflt 25, pp. 74] -53. The Selectric
type element looked much like a golf ball with raised characters on its surface.
It twisted, turned, and hammered-moving from left to right as it typed out
the message, rather than having the entire paper carri age move from right to
left as in conventional typewriters of the time. These innovative features, plus
the ability to change type elements easily for different print fon ts, contributed
to the exceptional popularity of the Sele ctric typewriter an d to the use of the
type element in many different computer terminals developed and manufac­
tured by IBM and others.
1 1 . U.S. IBM Year End Book, 1 955: A. L. W il l iams s copy for internal use only;
'

"Pages from the Pas t 1 976: a booklet produced by IBM to recognize its
, "

thirty-fifth year in Poughkeepsie.

1 2. IBM document, 6 October 1 943: M i nu tes of Engineering M ee ting Held
" in
Mr. Watson's Offic e ."
Copyrighted Material
352 Notes to Pages 122-132

1 3. J. W. Bryce, 3 October 1 944: to H. H. Aiken.

1 4. C. J. Bashe, 1 982: "The SSEC in Historical Perspective," Annals of the History
of Computing 4, pp. 296-3 1 2.
1 5. A. H. Dickinson, 1 8 July 1 967: interview by L. M. Saphire, pp. 41-44.
1 6. W. J. Eckert, 20 July 1 967: interview by L. M. Saphire.
1 7. Carl A. Bergfors was the engineer in Dickinson 's group who worked with
Phelps to design and build the IBM 603 Electronic Multiplier.
1 8. B. E. Phelps, 1980: "Early Electronic Computer Developments at IBM,"
Annals of the History of Computing 2, pp. 253-67.
19. Bashe, 1 982. Hamilton and Seeber benefited from consultations with Bryce,
Dickinson, McPherson, and Eckert in establishing specifications for the super­
computer.
20. IBM brochure, 1948: "IBM Selective Sequence Electronic Calculator"; J. C.
McPherson, F. E. Hamilton, and R. R. Seeber, Jr., 1 982: "A Large-Scale, Gen­
eral-Purpose Electronic Digital Calculator-the SSEC," Annals of the History of
Computing 4, pp. 3 1 3-26; Phelps, 1 980. The SSEC worked with numbers of 1 9
digits; a 20th digit position in storage units was used exclusively to store the
sign (plus or minus) of a number. The digit storage capacities given here follow
the lead of most prior SSEC descriptions by reckoning 20 digits per storage
unit.
2 1 . Phelps, 1980; Bashe, Johnson , Palmer, and Pugh 1 986: pp. 52-59. The
,

description of Codd's major contribution is taken from a brief biography that

preceded his 1981 ACM Turing Award Lecture, published in Communications of
the A CM 25, 1 982, pp. 1 09-1 7.
22. J. F. Brennan, 1 97 1 : The IBM Watson Laboratory at Columbia University: A
History (Armonk, N.Y: IBM) , pp. 22-25.
23. W. J. Eckert, August 1 946: "Facilities of the Watson Scientific Computing
Laboratory," Proceedings of the Research Forum, pp. 75-80.
24. Bashe et aI. , 1 986: p. 528.
25. Brennan , 1 97 1 : pp. 1 3-1 7 .

26. V. Bush, 1 945: Science the Endless Frontier: A Report to the President (Washington:
U.S. Government Printing Office) .

11 Government-Funded Competition
1 . For a description of wartime atomic weapons research using punched-card
machines at the Los Alamos Scientific (now National) Laboratory, see N.
Metropolis and E. C. Nelson, 1 982: "Early Computing at Los Alamos," A nnals
of the History of Computing 4, pp. 348-57.
2. C. J. Bashe, L. R. Johnson, J. H. Palmer, and E. W. Pugh, 1986: IBM's Early
Computers (Cambridge, Mass.: MIT Press) , pp. 59-68.
3. T. G. Belden and M. R. Belden, 1962: The Lengthening Shadow (Boston: Little,
Brown) , pp. 294-95.
4. The Bush differential analyzer is described in M. R. Williams, 1985: A History
of Computing Technology (Englewood Cliffs, NJ.: Prentice-Hall) , pp. 209-1 2. For
Copyrighted Material
 Notes to Pages 132-135 353

more detail on its origins and place in history, see L. Owens, 1 986: "Vannevar
Bush and the Differential Analyzer: The Text and Context of an Early Com­
puter," Technology and Culture 2 7, pp. 63-95.
5. An overview of the role of the federal government in supporting research
and development activities during and shortly after World War II may be found
in A. Hunter Dupree, 1986: Science in the Federal Government: A Histmy of Policies
and Activities ( Baltimore: Johns Hopkins University Press) , pp. 369-75. For an
account of postwar activities in computers and computing, see M. Rees, 1 982:
"The Computing Program in the Office of Naval Research, 1 946-1 953," Annals
of the History of Computing 4, pp. 102-20.
6. J. C. McPherson, 24 September 1 946: to W. J. Eckert.
7. W. J. Eckert, 1 1 October 1 946: to J. C. McPherson, "Development of Elec­
tronic Computing Devices."
8. J. C. McPherson and W. J. Eckert, 1 2 November 1946: to T. J. Watson.
9. N. Stern , 1 98 1 : From ENlAC to UNIVA C (Bedford, Mass.: Digital Press) ,
pp. 7-65. See especially pp. 33-36 for the relationship between J. W. Mauchly
and J. V. Atanasoff.
10. A good account of J. P. Eckert's contribution to the ENIAC is provided by
Herman Goldstine who observed it firsthand; see H. H. Goldstine, 1 972: The
Computer from Pascal to von Neumann (Princeton: Princeton University Press) ,
pp 1 48-56
. .

1 1 . M. V. Wilkes, 1 968: "Computers Then and Now," Journal of /he Association for
Computing Machinery 15, pp. 1-7 (presented earlier as an ACM Turing Lecture
in August 1967) ; Bashe, Johnson, Palmer, and Pugh, 1986: pp. 55-58. The
ENIAC project proposal was for a machine with about 5,000 vacuum tubes and
a cost of $1 50,000. Completed late in 1 945, ENIAC had 1 8,000 tubes and cost
about $750,000. Its 1 8,000 tubes can be compared with 1 2,500 in IBM's SSEC
and with 1 ,500 tubes in the Colossus, a special-purpose computer built in
England to decipher German codes that became operational in December
1943. See, for example, S. Rosen, 1 990: "The Origins of Modern Computing,"
Computing Reviews 31, pp. 450-81 , especially p. 457.
12. F. L. Alt, 1972: "Archaeology of Computers-Reminiscences, 1945-1947,"
Communications of the ACM 15, pp. 693-94. The difficulty of rewiring ENIAC
for each application was alleviated in 1 948 at a sacrifice of machine speed. A
permanent wiring was accomplished that enabled the machine to read (and
execute) its program as an ordered set of up to several hundred digit-pairs from
decimal switch banks originally intended to store 1 2-digit mathematical func­
tion values. Each of some eighty 2-digit numbers caused a particular operation;
the set of operations had been chosen to give ENIAC a suitably broad repertoire
of instructions and was embodied in the permanent wiring. Mter that, the
ENIAC was programmed by writing a sequence of numbers and setting it into
the switches. See also G. W. Reitwiesner, 1 1 April and 1 2 December 1 986: private
communications to J. H. Palmer.
13. W. Aspray, 1 990: John von Neumann and the Origins of Modern Computing
(Cambridge, Mass.: MIT Press) , pp. 5-1 3, 25-34.
1 4. Stern, 1 981 : pp. 58-59.
Copyrighted Material
354 Notes to Pages 135-137

15. Goldstine, 1 972: p. 192.

1 6. John von Neumann, June 1945: "First Draft of a Report on the EDVAC, "
reprinted i n N. Stern, 1 98 1 : From ENIA C to UNIVAC (Bedford, Mass.: Digital
Press) , pp. 1 77-246; E. W. Pugh, 1 984: Memories That Shaped an Industry (Cam­
bridge, Mass.: MIT Press) , p. 3; Goldstine, 1972: pp. 1 84-210.
1 7. A stored-program computer is "a digital computer that, under control of
internally stored instructions, can synthesize, alter, and store instructions as
though they were data and can subsequently execute these new instructions";
see IEEE Standard Dictionary of Electrical and Electronics Terms, ANSI/IEEE Std.
1 00-1 988, 4th ed. , 8 July 1 988, published by the Institute of Electrical and
Electronics Engineers, New York, N.Y Consistent with this modern definition
of a stored-program computer, the SSEC could execute sequences of instruc­
tions (each encoded as 20 decimal digits) read from its paper tape readers and
from its relay memory of 150 individually addressable units. The instruction-al­
teration capability was typically exploited by a program, stored either on tape
or in relay memory, which altered one or two instructions in the relays, after
which those instructions were executed. Detailed examples of instruction
modification are given in U.s. Patent 2,636,672, filed 19 January 1949: F. E.
Hamilton, R. R. Seeber, Jr., R. A. Rowley, and E. S. Hughes, Jr. , "Selective
Sequence Electronic Calculator. " See especially claims 70 and 99 (columns 303
and 309, respectively) and the descriptive material in columns 271-88.
18. The claims of the SSEC patent were particularly broad, covering such
concepts as system control of input-output devices and pipelining, as well as the
basic Moore School-type of stored-program computer. During an interference
between the SSEC patent and the ENIAC patent, then owned by Sperry Rand,
according to a 1 964 IBM Patent Operations report: "Out of 1 10 claims con­
tained in this patent, 1 05 were copied by Sperry-Rand into their ENIAC appli­
cation [for purposes of the interference process] . In the interference which
followed, they were completely unsuccessful and were not granted a single one
of the claims which they had copied from our SSEC patent. " See E. Lester,
February 1 964: "History of IBM Systems Patent Coverage, " written for IBM
Patent Operations; B. E. Phelps, 7 November 1973: to C. ]. Bashe "SSEC."
1 9. Bashe, Johnson, Palmer, and Pugh, 1986: pp. 47-59. The EDSAC, which was
built at Cambridge University in England and operational in May 1949, is
generally considered to be the first operational full-function computer based
on the Moore School design. Delays in completing the EDVAC were unavoid­
able following the 1 946 resignation of Eckert and Mauchly from the Moore
School. Kite Sharpless, who took over as head of the project, left with some
others to form their own company in 1 947. Others left during the ensuing years.
With Richard Snyder as chief engineer, EDVAC was ultimately completed in
1 95 1 and joined the ENIAC in 1952 at the U.S. Army Ordnance Corps Ballistic
Research Laboratory at Aberdeen Proving Ground in Maryland. The EDVAC
contained just over 3,500 vacuum tubes and about 27,000 other electronic
components. See M. R. Williams, 1 985: A History of Computing Technology (Engle­
wood Cliffs, NJ.: Prentice-Hall) , pp. 323-34, 349-53.
20. ]. P. Eckert, Jr. , 1 976: 'Thoughts on the History of Computing, " Computer 9,
pp. 58-65.
Copyrighted Material
 Notes to Pages 137-141 355

2 1 . J. P. Eckert, Jr , and J. W. Mauchly, 30 September 1 945: "Automatic High­

.

Speed Computing A Progress Report on the EDVAC, » Moore School of Elec­

-

trical Engin eering , Un ive rsity of Pennsylvania.

22. Goldstine, 1 972: p. 1 96. The mode of distribution of the documen t and its
impact on knowledgeable readers is revealed by M. V. Wilkes, who recalls that
L. J. Comrie (back from a trip to the United States) loaned him a copy of von
Neumann's report in May 1946. Not havin g a means for copying i t , Wilkes says,
"I sat up late into the nigh t reading the report . . . . I recogn ized this at once
as the real thing, and from that time on never had any doubt as to the way
computer develo pmen t wo uld go See M. V. Wilkes, 1 985: Memoirs of a Computer
."

Pioneer ( Cambridge, Mass.: MIT Press) , pp. 10�.

23. Stern, 1 98 1 : pp. 82-84. A simple description of the Selectron sto rage device
is provide d by Pugh, 1 984 : pp. 1 4-1 5.
24. J. C. McPherson, 23 Septe mbe r 1946: to file, with copies to A H. Dickinson,
W. J. Ec kert, and F. E. Hamilton. During these discussions, McPherson learned
that the EDVAC would be designed to economize its use of vacuum tubes,
having perhaps as few as 500. It was estimated that an EDVAC could be
replicated, after a few had been built, for less than $20,000. Sub seque n t expe­
rience revealed both of these es ti mates to be far too low.
25. Goldstine, 1972: p. 241; Stern, 198 1 : p. 94. The co urse was held for eight
weeks during July and August 1 946.
26. ]. C. McPherson, 1 2 March 1 97 1 : deposition testi mony in HoneyweU, Inc. v.
Sperry Rand Cbrpuration, U.S. District Court, District of Minnesota, pp. 1 1 7-22.
The reason given for the last-minute exclusion of mM representatives is highly
suspect since some who had no government contracts did attend. Among th ese
was M. V. Wilkes as described by him in 1 985: Memoirs of a Computer Pioneer
(Cambri dge, Mass.: MIT Press) , Chap . 1 2.
27. In a different context, N. Stern, 1 98 1 : From ENIAC to UNWAC, p. 41 , notes
that during the war "those c o m panie s cooperating [in electronics) with the
NDRC [National Defense Research Committee] were worried th at mM might
gain information and hence a competitive advan tage . "

28. Stern , 1 98 1 : pp. 66-1 1 5; see espec ially p p 84-85, 88-9 1 , 1 03.
' .

29. Stern, 1 98 1 : p. 92. Based on her 1 977 interviews with J W. M auchly an d J

. .

P. Eckert, Nancy Stem provides the interesting insight that Eckert's wife and
father agreed with Mauchly and h el ped persuade Eckert to attempt the inde­
pen dent commercial venture. Stern also reports that Eckert "expressed some
remorse over h is decision not to go with IBM"; see Stern, 1 981 : p. 265, n. 18.
30. Stern, 1 98 1 : p. 1 45.
3 1 . Stern, 1 98 1 : pp. 1 05-06. The evaluation of the proposal was made by George
Stibitz, a mathematician at the Bell Telephone Laboratories who designed a
number of c omputi ng devices using telephone switching relays, begi n n i n g with
a "partially automatic compu te r in 1 940. This work began at about the time
"

Howard Aiken and mM un de r took their collaboration o n the Mark I. See

Go lds tin e, 1972: pp. 1 1 5-2 0 .
32. Williams, 1985: pp. 360-63.

Copyrighted Material
356 Notes to Pages 141-148

33. The support of EMCC was arranged by the American Totalisator vice
president, Henry Strauss, who simultaneously became chairman of the nine­
member board of EMCC.
34. Stern, 198 1 : pp. 1 1 6-59; see especially pp. 1 19-29, 1 46-5 1 . The first UNI­
VAC to be installed at a Remington Rand facility was the seventh to be produced
and was shipped only two months before the one to GE. The shipment dates
and customers for the first eight UNIVACs are as follows: ( l ) Bureau of Census,
Washington, D.C., 3/5 1 ; (2) Air Force Comptroller, Arlington, Va., 3/52; (3)
Army Map Service, Bethesda, Md., 4/52; (4) Atomic Energy Commission,
N.Y.U., New York, N.Y., 4/53; (5) Atomic Energy Commission, Livermore, Calif.,
1 1 /52; (6) Bureau of Ships (DTMB) , Bethesda, Md., 5/53; (7) Remington Rand
UNIVAC, DPC, New York, N.Y., 10/53; (8) General Electric, Louisville, Ky.,
1 /54. See Sperry Univac's Commemorative Program for the National Computer
Conference 1 98 1 Pioneer Day Banquet, 6 May 1 98 1 .
35. E. Tomash and A . A Cohen, 1979: "The Birth of a n ERA: Engineering
Research Associates, Inc., 1 946-1 955," Annals of the History of Computing 1,
pp. 83-97. Functioning as a separate division of Remington Rand, ERA was the
responsibility of the director of advanced research, Lieut. Gen. (ret. ) Leslie R.
Groves, who had served during World War II as head of the Manhattan Project
( 1 942 through 1946) . The Atlas computer contained 2,700 vacuum tubes and
2,385 crystal diodes. It was run twenty-four hours a day with 10 percent of the
time allocated to preventive maintenance. Reportedly only sixteen hours of
unscheduled maintenance were required in the first five hundred hours of
operation-an outstanding record for the time.
36. From 1 955 to 1 970 the value of general-purpose computers shipped by IBM
is reported to have accounted for a declining percentage of the market, drop­
ping irregularly from about 80 to 60 percent. During the same time, the value
of UNIVAC shipments also dropped (very irregularly) from roughly 1 2 percent
of the market to about 6 percent, when it was overtaken by Burroughs. Honey­
well was the third major contender. Mter GE sold most of its computer business
to Honeywell in 1 970 and RCA sold out to Univac in 1 97 1 , Honeywell and
UNIVAC each held 6 to 1 0 percent of the market for several years. See M.
Phister, 1 979: Data Processing Technology and Economics (Bedford, Mass.: Digital
Press) , pp. 36-46.

12 IBM's Initial Respoose

1 . J. W. Bryce, 1 2 April 1 948: to T. J. Watson.
2. Business Machines, 4 April 1 949: Mr J W. Bryce, IBM Pioneer, Inventor, Chief
" . .

Consultant, Technical Staff Dean, Dies."

3. T. J. Watson, J r , and P. Petre, 1990: Father; Son & Co. (New York: Bantam
.

Press) , pp. 31-38, 42.

4. Watson, Jr., and Petre, 1 990: pp. 42, 76-87.
5. Watson, Jr., and Petre, 1 990: pp. 96-109.
6. Watson, Jr., and Petre, 1 990: pp. 1 30-32.
7. Watson, Jr., and Petre, 1 990: pp. 1 38-46.

Copyrighted Material
 Notes to Pages 148-151 35 7

8. E. W. Pugh, L. R. Johnson, and J. H. Palmer, 1 986: IBM's 360 and Early 370
Systems (Cambridge, Mass.: MIT Press) , Appen . D, pp. 649--5 0. The post of
Chairman had not been filled since George W. Fairchild vacated it in December
1924.
9. Fortune Magazine, November 1960: "Q. What Grows Faster than IBM? A. I.B.M.
Abroad"; IBM Magazine, September 1970: "For 'Dick' Watson, A New World of
Policy and Diplomacy," pp . 3-7; M. Campbell-Kelly, 1 989: ICL: A Business and
Technical History (Oxford, England: Clarendon Press) , pp. 1 39--43 . See also IBM
Kingston News, 23 December 1959, p. 3. In January 1 952 Tom Watson was
promoted to president of IBM, and in June 1 954 Dick Watson was promoted
to president of the rapidly growing World Trade Corporation . The final shift
in authority occurred in May 1 956 when Watson, Jr., replaced his father as CEO.
One month later the senior Watson died of a heart attack. A. K. Watson became
a corporate vice president and group executive and a member of the board of
directors in December 1 959.
1 0. Watson, Jr., and Petre, 1 990: pp. 1 35-37.
II. The eagerness of Watson, Sr.; to use electronics in the product li ne is
documented by internal records beginning in 1 943. For example, the minutes
of the engineering meeting held in Watson's office on 6 October 1 943 quotes
Watson as saying: "We are not doing enough on electronics. No tests have been
made with e lectroni cs on the accounting machine. We have done nothing
except with two punches at Endicott, and they don't work well. " The minutes
then assert, "Mr. Watson wants electronics used on the acc ounting machine and
on the printer. "
12. Watson, Jr. , and Petre, 1990: p. 205. Among the many books and articles
that suggest Watson, Sr., opposed the company s entry into electronics is one
'

by R. Sobel, 1 98 1 : IBM Colossus in Transition (New York: Times Books) , p. 1 21 .

Sobel says Charlie Kirk showed Watson, Jr., the prototype electronic calculator,
and that Kirk and Watson, Jr., subsequently "prodded Watson senior into
accelerating the research and then producing the machines. " In a letter of 6
January 1982 to Edward Nanas, IBM director of information, John C. McPher­
son identified this statement as one of many errors in Sobel's book saying that
,

when Watson, Jr., first saw the electronic multiplier in Bryce's laboratory, "Mr.
Watson was there too and needed no prodding to go ahead with immediate
production. "
1 3. The enthusiasm of T. j. Watson, Sr. , for expan din g the electric typewriter
business and modifying typewriters to serve as low-cost EAM equipment is
discussed in chap. 1 0; see especially n. 5. The move of typewriter production
to Poughkeepsie in 1 944 is documented in "Pages from the Past," 1976, a
booklet published by IBM to observe its thirty-fifth year in Poughkeepsie.
14. R. L. Palmer, 31 September 199 1 : interview by E. W. Pugh.
15. C. j. Bashe, L. R Johnson j. H. Palmer, and E. W. P ugh 1 986:
. , , IBM 's Early
Computers (Cambridge, Mass. : MIT Press) , pp. 60-6 1 .
16. B. E. Phelps, 1 980: "Early Electronic Computer Developments at IBM,"
Annals of the History of Computing 2, pp. 253-67; see especially page 258.

Copyrighted Material
358 Notes to Pages 152-155

1 7. A. L. Samuel, December 1 967: interview by L. M. Saphire. See also Bashe

et aI., 1 986: p. 67.
1 8. Phelps, 1 980.
19. Bashe et al. , 1 986: pp. 62-63.
20. Phelps, 1 980.
2 1 . G. J. Toben, 1 1 June 1 968: interview by L. M. Saphire .

22. G. S. Fenn, 1 948: "Programming and Using the Type 603-405 Combin ation
Machine in the Solution of Differential Equations," in Proceedings of the Scientific
Computation Forum, IBM Watson Scientific Computing Laboratory, H. R J.
Grosch , ed. , pp. 95-98. See also G. J. Toben, December 1 949: "Transition from
Problem to Card Program," Proceedings of the Computation Seminar, Endicott, N.Y.,
C. C. Hurd, ed., pp. 1 28-3 1 .
23. Wallace Eckert had designed and installed a Calculation Control Switch to
sequence three interconnected EAM units through lengthy computations at the
Thomas J. Watson Astronomical Computing Bureau at Colum bia University in ,

1937. Although this devic e was not available as a produc t, it was de scribed in
Eckert's widely read 1940 book, Punched Card Methods in Scientific Computation.
At an IBM-sponsored forum on automatic computation methods in August
1 946, Eckert described the facilities of the new Watson Laboratory and men­
tioned a pl an to go well beyond the Computation Control Switch by con ne cting
a relay calcula tor with the 405 Accounting Mach ine and a summary punch. At
a similar forum one year later, he again discussed the laboratory's proj ect in
which a relay calculator was c on n ec ted "with an accounting machine and a
spec i al control box to operate as a baby sequence calculator with instructions
on pun c hed cards." See ' w. J. Eckert, Augu st 1 946: "Facilities of the Watson
Sc ien tific Compu ting Laboratory," in Proceedings of the Research Forum, En dicott,
N.Y., pp. 75-80; August 1 947: "The IBM Department of Pure Science and the
Watson Scientific C om p utin g Laboratory," in Educational Research Forum Proceed­
ings, Endi co tt N.Y., pp. 31-36.
,

24. S. W. Dunwell, 15 August 1 99 1 : interview by E. W. Pugh .

25. C. C. Hurd, November 1 949: 'The IBM Card-Programmed Electronic Cal­
culator," in Proceedings of the Seminar on Scientific Computation, C. C. Hurd, ed.,
pp. 37-4 1 ; J. W. Sheldon and L. Tatum, February 1 952: "The IBM Card-Pro­
grammed Electronic Calcula tor," &view ofElectronic Digital Computers, Joint AlEE­
IRE Computer Conference, December 1 95 1 , pp. 30-36.
26. The IBM 603 was announced in September 1 946 and shipments began
before the end of the year. Customer shipments of the IBM 604 began in the
fall of 1 948 and of the CPC in Se pte m ber 1 949. The first UNIVAC was accep ted
by the Census Bureau on 31 March 1 95 1 . See Bashe et aI., 1 986: pp. 7 1 , 575.
On 29 Jan uary 1 950 the following list of year-end 1949 CPC customers, and
their primary appli cations was ge ne rated by C. M. Mooney at IBM headquar­
,

ters: U n ive rsi ty of California (Los Alamos) , Albuque rque , N.M. (2 machines for
p ar tial di ffe re n ti al equa tio n s ) ; Argonne National Lab. , Aurora, Ill. ( par tial
differential equations) ; Aberdeen Proving Grounds, Baltimore, Md. (general
ballistics, heat transfer, dopIe r control of guided m issile s, reduction of wind
,

tunnel and Theodolite data) ; Cornell Aeronautical Lab o rator y, Buffalo, N.Y.
(reduc tio n of wind tunnel doJ3frl?JNt&8�y Committee for Aeronautics,
 Notes to Page 155 359

Cleveland, Ohio (reduction of wind tunnel and other pressure gauge data,
nuclear shielding) ; Parsons Corporation, Grand Rapids, Mich. (service bureau
on aeronautical design) ; Fairchild Aircraft, Hagerstown, Md. (vibration and
flutter calculations) ; Penn State College, Johnstown, Pa. (distillation calcula­
tions) ; Carbide & Carbon Chemical (Oak Ridge ) , Knoxville, Tenn. (2 machines
for partial differential equations) ; Rand Corporation, Los Angeles, Calif. (2
machines for nuclear shielding, partial differential equations) ; U.S. Dept. of
Commerce Bureau of Standards (UCLA) , Los Angeles, Calif. (nuclear shielding,
reduction of Theodolite data) ; Telecomputing Corporation, Los Angeles, Calif.
(service bureau including vibration and flutter and aerodynamics for Lock­
heed) ; Northrop Aircraft, Los Angeles, Calif. (nuclear shielding, vibration and
flutter calculations) ; Douglas Aircraft (Santa Monica) , Los Angeles, Calif. (aero­
dynamics and vibration and flutter calculations) ; U.S. Dept. of the Interior,
Bureau of Mines, Pittsburgh, Pa. (development of explosives) ; U.S. Naval Ord­
nance, Riverside, Calif. (2 machines for reduction of Theodolite data) ; Boeing
Aircraft, Seattle. Wash. (vibration and flutter calculations) ; Reed Research. Inc.,
Washington. D.C. (service bureau on data reduction and logistic computa­
tions) ; U.S. Dept. of Commerce. Bureau of Standards, Washington, D.C. (Proj­
ect SCOOp. hydrographic (Loran) , optical rate tracing. etc.) ; IBM Service
Bureau. Chicago. Ill. (service bureau) ; IBM Service Bureau. Los Angeles, Calif.
(service bureau) ; IBM Service Bureau, New York, N.V. (service bureau) ; IBM
of Canada, Toronto, Canada (service bureau) ; Navy Hydrographic Office, Wash­
ington, D.C. ( hydrographic [Loran] tables for maps, stars, etc.) ; Consolidated
Vultee, Fort Worth, Tex. (vibration and flutter calculations) ; Woodmen of the
World. Omaha, Neb. ( [application not given] ) ; Air Force, Air Materiel Com­
mand, Dayton, Ohio (Project SCOOP establishment of consumption factors) .
27. The electronic computation speeds of the CPC were 2,000 additions or
subtractions and 86 multiplications or divisions per second. The equivalent
speeds for the UNIVAC were 1 ,700 additions or subtractions, 400 multiplica­
tions, or 250 divisions per second. In a problem mix involving an equal number
of each arithmetic operation, the UNIVAC would execute 1 ,012 per second and
the CPC, 1 ,043. The size of the numbers each handled was approximately the
same: 1 0 decimal digits for the CPC and 1 1 decimal digits for the UNIVAC. The
real performance advantage of the UNIVAC came from its one hundred times
larger internal memory ( 1 2,000 characters) and from its ability to transfer
information in and out of memory at speeds up to 9,000 decimal digits per
second. The electronic circuits were thus kept busy a higher percentage of the
time than on the CPC, which was constrained by the 100-to-I 5O-card-per-minute
rate at which program instructions could be entered. However, each of these
CPC instructions could result in as many as 60 program steps, consisting of any
one of the four arithmetic operations or a number transfer. Thus, in theory,
the CPC arithmetic circuits could also be kept busy full time-but it was far less
likely. See Engineering Research Associates, Inc., 1 950: High-Speed Computing
Devices (New York: McGraw-Hili) , pp. 1 57-58, 1 65, 203-06. In discussing the
CPC. one authority said: 'The limitations of this machine are due to the
relatively small size of its memory. and to the small number of subroutines which
can be plugged up on it at one time. " See B. V. Bowden , 1 953: "Computers in
America," in Faster than Thought. ed. B. V. Bowden (London: Pitman Publish­
ing) . p. 1 77. A computinlLcenter �ould, however, have numerous plugboards
1;opyngnted Matenal
360 Notes to Pages 157-1 61

prewired with a variety of subroutines. Prewired plugboards could be taken off

the shelf and inserted in the CPC in a matter of minutes.
28. W. J. Eckert, August 1946: "Facilities of the Watson Scientific Computing
Laboratory," Proceedings of the Research Forum, held in Endicott, N.Y, (New York:
International Business Machines Corporation) , pp. 75-80; H. R. J. Grosch,
1 99 1 : "Early Computing Courses," Annals of the History of Computing 13, pp. 30�
07; C. C. Hurd, 1980: "Computer Development at IBM," in A History of Comput­
ing in the Twentieth Century (New York: Academic Press) , pp. 389-4 1 8, especially
p. 400. Describing the course offered by Wallace Eckert as a "fully integrated"
part of the Columbia University graduate program, Herb Grosch says, "Starting
in the fall term of 1 945, he offered a course in machine methods with the
assistance of Marjorie Severy Herrick, Lillian Feinstein Hausman, and Eric
Hankam. " These courses predated those offered by Howard Aiken at Harvard
by several years, according to Grosch.
29. W. J. Eckert, 10 May 1 948: to J. C. McPherson.
30. C. E. Love, 16 June 1948: to all district managers, "Employment of Men for
Scientific Calculating Program"; W. H. Reid, 3 1 May 1 949: to G. P. Lovell,
"Scientific Sales Program. "
3 1 . W. H. Reid, 1 August 1 949: to C. E. Love, "Scientific Computation and
Related Activities. "
32. Business Machines, 28 November 1 949: "Dr. Hurd Named Director of Applied
Science Dept. "
33. Business Machines, 1 3 March 1953: "Applied Science Dept. Is Raised to
Division Status. " This article gives a good description of the activities of the
Applied Science Department. The reported elevation to divisional status was
symbolic of its growing importance but had little organizational significance.
34. Computerworld, 3 November 1 986: "Census led computer age by counting on
Univac I," p. 1 50.
35 . When he joined the MIT Radiation Laboratory in July 1 94 1 , Havens was a
third-year graduate student at Cal Tech. having previously received his bache­
lor's degree in electrical engineering from the University of Washington. Two
others hired from the MIT Radiation Laboratory at the same time were Robert
M. Walker and John J. Lentz.
36. B. L. Havens. 31 January 1 968: interview by L. M. Saphire.
37. R. Hopkins. 1 2 August 1 948: to L. H. LaMotte, "Digital Computing Ma­
chines. " The six organizations "planning to order large computing machines at
the earliest possible moment" were listed as the Naval Ordnance Laboratory.
Washington. D.C.; David Taylor Model Basin, Washington. D.C.; Naval Research
Laboratory. Washington, D.C.; University of Illinois; Project Rand. Los Angeles.
Calif. ; and Naval Test Station. Inyokern. Calif.
38. L. H. LaMotte, 28 August 1948: to J. G. Phillips.
39. J. C. McPherson, 12 October and 3 November 1950: to Commander, U.S.
Naval Ordnance Laboratory; A. L. Williams. 7 May 195 1 : to Department of the
Navy. An undated telegram beginning with the phrase "You are hereby awarded
letter of intent for contract NORD" authorized reimbursement of "anticipatory

Copyrighted Material
 Notes to Pages 161-1 69 361

costs incurred by you on and after 1 May 1 95 1 in an amount not to exceed

$40,000."
40. Bashe et al. , 1 986: pp. 1 8 1 -83.

lJ Wotson, ft, 1Gkes 0uJrge

1 . W. W. McDowell, 1 4 December 1 948: to T. J. Watson, Jr.; 1 7 December 1 948:
to W. L. Lewis, "Hiring of Electronic Engineers. "
2. T. J. Watson, Jr., and P. Petre, 1 990: Father, Son & Co. (New York: Bantam
Press) , pp. 1 99-200.
3. Business Machines, 1 2 November 1 942: "Mr. A. L. Williams Is Promoted to
Comptroller of IBM Corp. ; IBM News Magazine, 25 April 1 969: "Mr. Heart-of-the
"

Matter," pp. 7 1 1 .
-

4. Watson, Jr., and Petre, 1 990: p. 20 1 .

5 . Business Machines, 1 2 November 1 942: "Mr. W. W. McDowell Heads Labora­
tory at Endicott"; IBM document, undated: "Employment Record of William
W. McDowell."
6. W. W. McDowell, I I November 1 970: interview by L. M. Saphire.
7. Watson, Jr., and Petre, 1 990: p. 203.
8. Watson, Jr., and Petre, 1990: pp. 1 56-57, 1 96.
9. J. C. McPherson, December 1 990: interview by E. W. Pugh.
1 0. Watson, Jr., and Petre, 1990: p. 201 .
1 1 . J. W. Birkenstock, 2 5 May 1982: interview by E. W. Pugh.
1 2. C. C. Hurd, 18 October 1 950: to W. L. Lewis, "Applied Science Customers
Participating in the National Defense Effort"; 3 January 1 979: testimony in U.S.
v. IBM, U.S. District Court, Southern District of New York, N.Y, p. 86339.
13. J. W. Birkenstock, 30 November 1 950: to T. J. Watson, Jr.
1 4 . J. W. Birkenstock, 1 5 December 1 950: to T. J. Watson, Jr.
1 5. E. W. Pugh, 1 984: Memmies That Shaped an Industry (Cambridge, Ma.� .: MIT
Press) , pp. 34-38.
1 6. C. J. Bashe, L. R Johnson, j. H. Palmer, and E. W. Pugh, 1 986: IBM's Early
Computers (Cambridge, Mass.: MIT Press) , pp . 1 1 2, 1 19, 1 90-95.
1 7. S. W. Dunwell, 8 November 1 949: "Electronic Tape Program"; N. Rochester
and W. Buchholz, 24 March 1 950: "Preliminary Outline of Tape Processing
Machine."
18. W. W . McDowell, 17 December 1 948: to W. L. Lewis, "Hiring of Electronic
Engineers. "
19. Birkenstock, 1 5 December 1 950; N. Rochester, 18 October 1955: to R K.
Richards, "Electrostatic Memory in the 701 . " The committee was chaired by R
M. Bury of the sales organization. His primary assignment at the time was to
help determine how IBM could best contribute to the Korean War effort. Other
members of the committee were Stephen W. Dunwell (Future Demands) ,
Wallace J. Eckert and Llewellyn H. Thomas (Watson Laboratory) , Cuthbert C.

Copyrighted Material
362 Notes to Pages 1 69-1 72

Hurd and Walter H. Johnson (Applied Science) , and Nathaniel Rochester and
C. M. Mooney (Engineering) .
20. N. Rochester, 3 October 1 967: interview by L. M. Saphire; C. C. Hurd, 3
January 1 979: testimony in U.S. v. IBM, U.S. District Court, Southern District of
New York, pp. 86338--44 .
2 1 . W. W. McDowell, 22 January 195 1 : to E. M. Douglas; H. T. Hansfort, 23
February 195 1 : to J. W. Birkenstock.
22. L. H. LaMotte, 21 May 1 952: to internal distribution, "IBM Electronic Data
Processing Machine (Defense Calculator) "; IBM document, 4 August 1952:
"lBM-EDPM Type 701 Order Status"; H. T. Hansford, 28 July 1 952: to L. H.
LaMotte, "Electronic Data Processing Machine Delivery Schedule"; L. H.
LaMotte, 18 August 1 952: to H. T. Hansford, "Electronic Data Processing
Machine Delivery Schedule."
23. Bashe et aI., 1 986: pp. 1 35-36 .
24. R. L. Palmer, 1 3 September 1 99 1 : interview by E. W. Pugh.
25. J. C. McPherson, 5 October 1 95 1 : to John von Neumann. This letter formally
offered a consulting position with a retainer of $ 1 2,000 per year plus travel
expenses for up to thirty days of consulting; any additional days were to be paid
at the rate of $400 per day. The consulting arrangement was officially termi­
nated by a letter to von Neumann, dated 3 February 1 955, signed by T. J.
Watson, Jr. , and accomp an ied by a check "representing the final payment. " It
is interesting to note that Wallace Eckert and John McPherson had attempted
to hire von Neumann early in 1945, well before his "First Draft of a Report on
the EDVAC" was written. At a fee of $5,000 per year, von Neumann was to have
devoted an average of "fifteen percent of his occupational time to this IBM
work, and under IBM's direction," according to an unsigned agreement that
was dated 1 April 1 945. This early contract was not consummated because of

perceived conflicts of interest for von Neumann. Further consideration of a

consulting arrangement was dropped a year later because von Neumann wa�
undertaking work on an electronic computer at the Institute for Advanced
Study in Princeton; see W. Aspray, 1 990: John von Neumann and the Origins of
Modern Computing (Cambridge, Mass.: MIT Press) , pp. 241-45.
26. Ba�he et aI., 1986: pp. 1 35, 1 38--4 1 ; W. Buchholz, 1953: "The System Design
of the IBM Type 701 Computer," Proceedings of the Institute of Radio Engineers 4 1 ,
pp. 1 262-75.
27. IBM Principles of Operation Manual, 1 953: 'Type 701 and Associated
Equipment" (IBM Form 22-6042-0) ; N. Rochester, 26 May 1 952: Memo 1 9 1 ,
"Availability o f the 70 1 . " Production of the IBM 70 1 was a t a n average rate of
one per month from about April 1 953 through June 1 954 when the eighteenth
machine was completed and shipped to Lockheed Aircraft in Burbank, Califor­
nia. A nineteenth machine was assembled from the remaining "spare parts. "
The nineteen IBM 70 1 s were shipped primarily to companies in the aerospace
and other defense-related industries as follows: IBM World Headquarters, New
York, N.Y. ( 1 2/20/52) ; University of California, Los Alamos, N.M. (03/23/53 ) ;
Lockheed Aircraft Company, Glendale, Calif. (04/24/53) ; National Security
Agency, Washington, D.C. (04/28/53) ; Douglas Aircraft Company, Santa
Monica, Calif. ( 05/20/5.3) ; Genera!, Ele{:tri.c Company, Lockland, Ohio
copyngnreu lVTarena,
 Notes to Pages 1 72-1 77 363

(05/27/53) ; Convair, Fort Worth, Tex. ( 07 / 22 / 53) ; U.S. Navy, I nyoker n , Calif.
(08/27/53) ; United Aircraft, East Har tford , Conn. (09/1 8/53) ; North Am eri­
can Aviation, Santa Monica, Calif. ( 10/09/53) ; Rand Corporatio n , Santa
Monica, Calif. ( 1 0/30/53) : Boein g Corp oration , Seattle, Wash. ( 1 1 /20/53) ;
Universi ty of California, Los Alamos, N.M. ( 1 2/ 1 9/53) : Douglas Aircraft Com­
pany, El Segundo , Calif. (01/08/54) ; Naval Aviation Supply, Ph ilade lphia, Pa.
(02 / 1 9 / 54 ) ; Un ive rsity of California, livermore, Calif. (04/09/54) ; General
Motors Corporation , Detroit, Mich. (04/23/54) ; Loc kh ee d Aircraft Com p any,
Glendale, Calif. (06/30/54) ; U.S. Weather Bureau, Washington , D.C.
(02/28/55) . See Annals of the History of Computing 5, April 1 983: "Customer
Experien ce s ; p. 175.
28. New Yorlc Times, 30 April 1 952: "Watson Reports IBM Expansion," p. 4 1 .
29. IBM Record, April 1 953: "New I B M Electronic Data Processin g Machines
Unveiled at World He ad quarte rs in New York. " By contrast, UNNAC was typi­
cally not sufficiently operational to earn revenue until several months after
shipmen t . The second UNNAC, for example, passed its accep tance test and was
shipped in late February 1 952 to the Pentagon, wh ere the manufacturer's
engin eers installed it and pe rfo rm ed limi te d opera tion and checking before it
was formally transferred four mon ths later to the air force on 25 Jun e 1 952.
See R. Kopp, 1 953: "Experience on the air force UNIVAC ," Proceedings of the
Eastern Joint Computer Conference, pp. 62-67.
30. W. G. Bouricius , December 1 953: "Operating Experience with the Los
Alamos 701 , " Proceedings of the Eastern Joint Computer Conference, pp. 45-47.
31 . T. A. Burke (undated) : "Review-Type 701 El ectron ic Data Processing
Machine Program. "

32. E. Tomash and A. A . Cohen, 1979: "The Birth of an ERA: Engi n ee ring
Research Associates, Inc. 1946-1955," Annals of the History of Computing 1,
pp. 83-97. The origi n al Atlas (Atlas I) was installed in Washington, D.C., in
.
December 1 950; see chap. 1 1 .
33. H. H. Goldstine, 1972: The Computerfrom Pascal to von Neumann ( Prince ton:
Princeton Universi ty Press) , p. 329.
34. Bashe et al. , 1 986: pp. 1 73-78.
35. M. E. Davis, December 1 953: wrhe Use of Electronic Data Processing Systems
in the Life Insurance Business," Proceedings of the EasternJoint Computer Conference,
pp. l 1-18.
36. Business Machines, 15 April 1 954: "1'. V. Le arson Named Director of EDPM. "
37. Business Machines, 3 May 1 954: "EDPM Programming Research Department
is Organ ized. "
38. O. M. Sc ott, 25 June 1954: "702 EDPM Equipment." The fourteen installa­
tions of the IBM 702 were, in order of receipt of their e quipm e n t: IBM
Poughkeepsie , Monsanto Chemical, National Security Agency, Aviation Supply
Office ( USN ) , State of California, General Electric (Hanford) , Commonwealth
Edison, Oklahoma City Air Materiel Area (USAF) , Pratt & Whitney, Bank of
America, Chr ysle r, Prudential, General Electric (Schenectady), and Ford Motor;
see L. R. John son , 22 Ap ri l 1976: to J. H. Palmer, "IBM 702 Original Customers. "

Copyrighted Material
364 Notes to Pages 1 77-182

39. IBM Product Announcements, 1 October 1 954: "IBM Type 705 Electronic
Data Processing Machine"; "IBM Type 701-704 EDPM."
40. Bashe et aI. , 1 986: pp. 446-49.
4 1 . Business Machines, 1 December 1 954: "T. V. Learson Promoted by Board to
the Post of Vice-President in Charge of Sales"; 1 9 Jan uary 1 955: "Dr. Cuthbert
C. Hurd Is Promoted; Becomes Director of EDPM. "
42. T. J. Wat'lOn, Jr., 1 3 June 1 983: interview by J. B. Rochester in Computerworld,
pp. 1 0-17.
43. Business Machines, 1 June 1 956: "T. J. Watson, Jr., Named IBM Chief Execu­
tive," pp. 3-4.
44. T. J. Watson, 1 0 February 1 948: to J . G. Phillips.
45. Bashe et aI., 1 986: pp. 73-78. At this time a digit was represented in the
Magnetic Drum Calculator in the same binary-coded decimal (4-bit) code used
in the 603, SSEC, and 604. Engineering considerations later led to the choice
of a 7-bit code for processing circuits and a 5-bit code for the magnetic drum.
The 7-bit code facilitated economical self-checking and arithmetic circuits. But
the cost of magnetic recording heads and circuits worked against its use on the
drum, where main memory was implemented in forty bands of five tracks each.
46. C. E. Love, 1 1 August 1 949: to T. J. Watson, Jr., "Magnetic Drum Develop­
ment. "
47. ]. M. Coombs, 1 947: "Storage of Numbers on Magnetic Tape," Proceedings of
the National Electronics Conference, pp. 201-09. A photograph of an early ERA
drum may be found in E. W. Pugh, 1 984: Memories That Shaped an Industry
(Cambridge. Mass.: MIT Press) , p. 2 1 .
48. Watson, Jr., and Petre, 1 990: pp . 1 99-200; Love, 1 1 August 1 949.
49. IBM documen t, 8 March 1 950: "Agreement, IBM and ERA," signed by T. J.
Watson , Jr., for IBM and ]. E. Parker fo r ERA; Bashe et al. , 1 986: pp. 80-8 1 .
50. IBM documen t (undated) : "Conference o n Magnetic Drum Calculators of
March 1 3 & 1 4, 1 950. " Twenty members of IBM's engineering staff attended
this in ternal IBM conference, including F. E. Hamilton, W. W. McDowell, ]. C.
McPherson, R. L. Palmer, and B. E. Phelps. C. C. Hurd represented the Applied
Science Department, and S. W. Dunwell represented the Future Demands
Department and was instrumental in establishing the stringent cost criteria.
5 1 . Bashe et aI., 1 982: pp. 95-98.
52. S. W. Dunwell, 30 September 1 952: "IBM Calculator Program "; ]. C. McPher­
son, 20 November 1 952: "Competitive Drum Computers. " The seven small
stored-program computers, which sold for prices ranging from $45,000 to
$ 1 00,000, were offered by Consolidated Engineering Corp., Hogan Laborat<r
ries, Marchant Calculators, Inc. , Monroe Calculating Machine Co., National
Cash Register Co. , Remington Rand, and Underwood Corporation.
53. Bashe et aI. , 1 986: pp. 98-1 0 1 ; IBM Development Reports, October 1 952
and November 1 952: "Magnetic Drum Calculator. "
54. ]. W. Birkenstock and W. W. McDowell, 24 August 1 953: "Magnetic Drum
Calculator"; A. J. Perlis, December 1 954: "Characteristics of Currently Available
Small Digital Computers," Proceedings of the Eastern Joint Computer Conference,
Copyrighted Material
 Notes to Pages 182-184 365

pp. 1 1 -1 6; Bashe et aI., 1 986: pp. 1 7 1 -72, 339. When John Hancock in Boston
received the first IBM 650 computer in December 1 954, more than 450 had
been ordered-almost twice the total number projected just before an­
nouncement. One of the best discussions of the utility of the 650 is provided
by E. H. Dohrmann, 1 956: "Data Processing Machine-Business and Industry
'Workhorse, ' " in Punched Card Annual, pp. 20-23.
55. Bashe et aI. , 1 986: p. 1 72.

56. The level of success and durability achieved by the IBM 650 is revealed by
the "Final Report-IBM Computer Product Line," prepared for the Radio
Corporation of America, Advanced Military Systems Division, Princeton, NJ.,
by Auerbach Electronics Corporation on 1 2 February 1 960. According to the
report: 'The IBM 650 is an old machine, but is still selling well. Over the years,
a host of attachments and accessories have been developed for this machine to
increase the range of application. Many of the early installations were for
engineering computation . In later years, the machine was primarily used for
business data processing. Before the 1 401 was introduced, the 650 was a popular
"first" machine even for large companies, with the possibility of a step-up to a
larger computer later. " Tabular data in the report indicate that fifteen hundred
IBM 650s had been installed by the end of 1 959, compared to only seven of
the newly introduced, transistorized RCA 501 computer. For raw arithmetic
speed, the IBM 650 was 0.7 times as fast as the RCA 50 1 , but its lower minimum
cost give it almost a threefold advantage in arithmetic cost-performance over
the 501 . The RCA 50 1 was of course functionally superior because of its larger
memory and input-output capability (Auerbach Associates Records, CBI 30,
courtesy of the Charles Babbage Institute Archives, University of Minnesota) .

14 Programming Compurers
1 . The first issue of the IBM Applied Science Department Technical Newsletter
appeared in June 1 950 and contained general purpose control-panel wiring
arrangements. Distributed to customers and others until 1 957, this informal
newsletter played a role for scientific applications similar to that played by
Pointers for business applications during the 1930s and 1 940s. See C. J. Bashe,
L. R. Johnson, J. H. Palmer, and E. W. Pugh, 1 986: IBM 's Early Computers
(Cambridge, Mass.: MIT Press) , pp. 85-86, 6 1 8 N. 42.
2. A. Hyman, 1 982: Charles BaMage, Pioneer of the Computer (Princeton: Princeton
University Press) , especially pp. 1 95-98. See also J. Baum, 1 986: The Calculating
Passion ofAda Byron ( H amden, Conn.: Shoe String Press) . Ada Lovelace was the
daughter of the poet George Gordon, sixth Lord Byron.
3. A new method for programming ENIAC was first tested in September 1 948.
Proposed by von Neumann, the technique slowed the machine operations by
several fold, but it made programming so much easier that the old method was
never used again . It provided ENIAC with a set of fifty-one orders (instructions)
that could be called on for execution. Within another year, the instruction set
had been expanded from fifty-one to ninety-two. See H. H. Goldstine, 1972:
'The Computer from Pascal to von Neumann" (Princeton: Princeton University
Press ) , pp. 233-34.

Copyrighted Material
366 Notes to Pages 184-187

4. ]. P. Eckert and ]. W. Mauchly, 30 September 1 945: "Automatic High Speed

Computing-A Progress Report on EDVAC," Moore School of Electrical Engi­
neering, University of Pennsylvania, p. 39. This is a report based on work done
under contract to the U.S. Army Ordnance Department.
5. A. M. Turing, 1945: "Proposals for Development in the Mathematics Division
of an Automatic Computing Engine (ACE) ," reprinted with a foreword by D.
W. Davies, April 1 972: Report Com. Sci. 57, National Physical Laboratory,
especially p. 28.
6. M. V. Wilkes, D. ]. Wheeler, and S. Gill, 1 95 1 : The Preparation of Programs for
an Electronic Digital Computer (Cambridge, Mass.: Addison-Wesley Press) ; M. V.
Wilkes, June 1949: summary of his remarks in "Discussion of Plans, Projects and
General Ideas," Riport of a Conference on High speed Automatic Calculating-Ma­
chines, issued by the Cambridge University Mathematical Laboratory, January
1 950, pp. 1 23-33, especially p. 1 33.
7. F. E. Johnston, 1 983: "The One-Card Binary Loader for the IBM 701 , " Annals
of the History of Computing 5, pp. 1 33-35.
8. N. Rochester, 3 August 1 952: "Symbolic Programming," an IBM technical
report; March 1 953: "Symbolic Programming," Transactions of the Institute of
Radio Engineers, Professional Group on Electronic Computers, EC-2, pp. 1 0-15, with
corrections published in the June 1 953 issue of the same journal on p. 27; IBM
document, undated: "IBM Applied Science EDPM Assembly Program-IBM
S02." The assembly program S02 was written by W. F. McClelland of C. C.
Hurd's Applied Science Department.
9. ]. Morison, August 1 953: "701 Symposium Notes," Douglas Aircraft Company,
Inc., report.
1 0. N. Rochester, 10 September 1 953: "A Meeting of 701 Customers in Los
Angeles on August 1 7-18, 1 953," an IBM report.
1 1 . F. Jones, June 1 956: "SHARE-A Study in the Reduction of Redundant
Programming Effort through the Promotion of Inter-Installation Communica­
tion, " Symposium on Advanced Programming Methods for Digital Computl'Ts (Wash­
ington: Office of Naval Research) , pp. 67-70. As described in this article,
"SHARE is a voluntary, informal organization of the users of the IBM Type 704
electronic data-processing machines. It is devoted to ( 1 ) the standardization of
machine language and certain machine practices, (2) the elimination of redun­
dant effort expended in connection with the use of the computer, (3) the
promotion of inter-installation communication , and (4) the development of a
meaningful stream of information between the user and the manufacturer."
1 2 . 1 1 03 Central Exchange, February 1 956: Newsletter Number 8. An informal
predecessor organization, called the 1 1 03 Central Exchange, reported in its
February 1 956 newsletter that "the early 1 1 03A purchasers desired to go beyond
the Central Exchange with its voluntary exchange of completed routines and
actively cooperate in the early stages of program planning and assignment of
programming manpower" (USE Inc. Records, CBI 20, courtesy of the Charles
Babbage Institute Archives, University of Minnesota) .
1 3. H. S. Bright, August 1 960: "Computer User Groups," reprinted in B. H.
Bruemmer, 1 990: "Early Computer User Groups," A nnals of the History of Com­
puting 12, pp. 56-6 1 .
Copyrighted Material
 Nows to Pages 1 88-1 92 367

1 4. P. Armer, November 1 956: "SHARE," Proceedings, Second Annual Electronic

Business Systems Conference, pp. 1 2-1 7.
15. C. j. Bashe, L. R johnson, J. H. Palmer, and E. W. Pugh, 1 986: IBM's Early
Computers (Cambridge, Mass.: MIT Press) , pp. 358-61 .
1 6. D. R. Mason, 1 983: 'The 701 in the IBM Technical Computing Bureau , "

Annals of the History of Computing 5, pp. 1 76-77.

1 7. Morison, August 1953.
1 8. Mason, 1 983.
19. During April 1955, for example, IBM allocated 89 hours of machine time
(25 percent of all time ) on its own 702 for customers to test and de bug
programs prior to installation of their computers. The first customer to receive
an IBM 702 was the Monsanto Chemical Company. It logged 40 hours of
machine time at IBM before its own computer was installed and operational.
See H. B. Ainsley, Jr. , 18 May 1 955: to T. E. Clemmons, "702 Installation Progress
Report." Installation experience with the IBM 702 is typified by the memoran­
dum, J. J. Kenney, 23 November 1 955: to W. J. Mair, "EDPM," which reported
as follows: 'The Type 702 shipped from Poughkeepsie on October 24 for GE

Schenectady was installed and running under testing conditions one week after
the shipping date. Mter installation this machine was operated under normal
and marginal testing conditions until November 1 7, with excellent perfor­
mance. The machine was released to the customer on November 1 7 . "
20. R M. Petersen, january 1 957: "Automatic Coding a t G.E. ," Proceedings of the
Symposium on Automatic Coding, held at the Franklin Institute in Philadelphia.
2 1 . G. Schussel, June 1 965: "IBM vs. RemRand-Part 2." Datamation, pp. 58-66.
The computer for General Electric was the eighth UNIVAC to be produced.
Unlike IBM that retained the first p roduced 701 and 702 for customer support
activities, Remington Rand waited until two months before General Electric's
machine was installed before it placed a UNIVAC at its New York City facilities.
The IBM 702 installation at Ford Motor Company was successfully handling a
payroll of approximately 70,000 hourly employees in March 1 956, as reported
by J. Boccomino, August 1 957: "The Importance of Program Maintenance,"
Systems and Procedures, pp. 9-1 2.
22. J. W. Backus, 1981 : "The History of FORTRAN I, II, and III," in History of
Programming Languages, e d . R L. Wexelblat (New York: Academic Press) , pp. 25-
74; December 1 967: interview by L. M. Saphire;J. W. Sheldon, 1 2 October 1 95 1 :
to C . C . Hurd.
23. D. B. MacMillan an d R H. Stark, February 1 951 : "Floating Decimal Calcu­
lations on the IBM Card-Programmed Electronic Calculator, " IBM Applied
Science Department Technical Newsletter No. 2, pp. 1 6-26. The floating-poin t
number format for the CPC consisted of ten d igits with the decimal point
,

assum�d to be in front of the first digit. The last two d igi ts indicated the power
of ten by which the eight-digit fraction was to be multiplied.
24. The SSEC was e quipped with eight electronic registers so that various data
needed for a computation could be transferred simultaneously from memory.
This helped overcome the disparity in speed between the electronic arithmetic
circuits and the slower electromechanical memory.
Copyrighted Material
368 Notes to Pages 1 92-195

25. J. W. Backus and H . L. Herrick, May 1954: "IBM 701 Speedcoding and Other
Automatic-Programming Systems, " Symposium on Automatic Programming for Digi­
tal Computers (Washington: Office of Naval Research) , pp. 106-13; IBM Manual,
1 954: "IBM Speedcoding System for the Type 701 Electronic Data Processing
Machines' (IBM Form 24-6059-0) ; C. ]. Bashe, L. R johnson , j. H. Palmer, and
E. W. Pugh , 1 986: IBM's Early Computers (Cambridge, Mass. : MIT Press) ,
pp. 333-38. The programmer had to write each Speedcoding instruction as a
string of decimal digits because Speedcoding had no associated assembly pro­
gram to convert mnemonic operation codes or symbolic addresses to real ones.

26. Backus, 198 1 : pp. 26-27.

27. 1- C. McPherson , 15 February 1954: "Speed Coding. " This brief internal
memorandum asserts , "It is estimated that the speed ratio between the 701
working normally and in speed coding is approximately 15 to I."

28. j. E. Sammet, 1992: "Farewell t o Grace Hopper-End o f a n Era ! " Communi­

cations of the Association for Computing Machinery 35, pp. 1 28-31 . Grace M. Hop­
per died on 1 january 1 992.

29. G. M. Hopper, May 1952: "The Education of a Computer," ProceMings of the

Association for Computing Machinery, pp. 243-49; May 1953: "Compiling Rou­
tines , " Computers and Automation 2 , pp. 1-5.
30. D . E. Knuth and L. T. Pardo, 1977: "The Early Development of Programming
Languages, " in Encyctqpedia of Computer Science and Technology, Vol. 7, ed. J. Belzer,
A. G. Holzman, and A. Kent (New York: Marcel Dekker) .

3 1 . Bashe et al., 1 986: pp. 338-42; Backus, 1 98 1 ; H. L. Herrick, 15 April 1 968:

interview by L. M. Saphire.

32. Backus and Herrick, May 1954: p. 1 1 2.

33. Backus, 1981 : pp. 29-33.

34. IBM Programmer's Reference Manual, 1 5 October 1 956: "The FORTRAN
Automatic Coding System for the IBM 704 EDPM"; 1- W. Backus, R J. Beeber,
S. Best, R. Goldberg, L. M. Haibt, H. L. Herrick, R A. Nelson, D. Sayre , P. B.
Sheridan, H. Stern , I. Ziller, R A. Hughes, and R Nutt, February 1 957: 'The
FORTRAN Automatic Coding System," Proceedings of the Western Joint Computer
Omference, pp. 1 88-98. Three authors on loan from other organizations to the
FORTRAN project were S. Best of MIT, R A. Hughes of the University of
California Radiation Laboratory, and R Nutt of the United Aircraft Corpora­
tion . See also Bashe et al. , 1986: pp. 356-58.

35. A. J. Perlis, 6 December 1975: direct testimony in U. S. v. IBM, U.S. District

Court, Southern District of New York, pp. 1 878, 1974. Alan Perlis, a world-re­
nowned professor of computer science at Yale University, presented a strong
endorsement of the pioneering significance of FORTRAN in spite of the fact
that he testified as a witness for the government. In his testimony he said i n
part, "FORTRAN wa s developed b y IBM in the middle fifties a n d a s mentioned
before, h ad an enormous i mpact on a large collection of users and industries
and in a certain sense shaped the way the computer would be used to support
scientific and engineering research in the United States . . . . Prior to '71 I would
say that almost all , that is, 85 to 99 percent, 95 percent of the students were
taught that programming and learning FORTRAN were synonymous, and one

Copyrighted Material
 Notes Lo Pages 1 96-202 369

has only to look at the set of texts available for student use during that period
to observe this effect. U

36. Backus, 1981: p. 30.

37. J. E. Sammet, 1 98 1 : 'The Early History of COBOL," in History ofProgramming
Languages, ed. R. L. Wexelblat (New York: Academic Press) , pp. 199-278.
38. Business Week, 20 January 1962: "Young Team Wins a COBOL Race, " pp. 60-
62.
39. M. P hister, Jr 1979: Data Processing Technology and Economics, 2d ed. (Bedford,
.,

Mass.: Digital Press) , pp. 82-83.

40. Think, May/June 1 982: "Many Are Chosen," pp. 23-25.

15 An Air Defense System

1 . K. C. Redmond and T. M. Smith, 1 980: Project Whirlwind: The History of a
Pioneer Computer (Bedford, Mass.: D i gi tal Press) , pp. 1 68-78.
2. T. J. Watson, Jr., and P. Petre, 1 990: Father; Son & Co. (New York: Bantam
Press) , pp. 232-33.
3. Redmond and Smith, 1 980: pp. 1-4, 1 4, 22-25. The Servomechanisms Labo­
ratory was established under the direction of Professor Gordon S. Brown, with
the help of A. C. Hall, J. O. Silvey, and J. W. Forrester in December 1940.
Forrester received his master's degree in electrical engineering in 1 945.
4. P. O. Crawford, 1 942: "Automatic Control by Arithmetical Operations," thesis
submitted in partial fulfillment of the requirements for the degree of master
of science at the Massachusetts Institute of Technology, p. 1 . To complete his
thesis project quickly, Crawford limited his own analysis to "the elements and
operation of a calculating system for performing one of the operations in the
control of anti-aircraft gunfire, which is, namely, the prediction of the future
position of the target. "
5. The closest approach to practical implementation of electronic arithmetic
circuits had been achieved by Atanasoff and Berry at Iowa State College and
by Palmer and Phelps at the IBM laboratory in Endicott shortly before these
researchers moved on to their World War II assignments in 1942. Little publicity
was given to the work at Iowa State, and the proprietary IBM results were not

disclosed outside the company (see chap. 7 ) .

6 . Redmond and Smith, 1 980: pp. 41-44.
7. E. W. Pugh, 1 984: Memories That Shaped an Industry (Cambridge, Mass.: MIT
Press) , p. 64; Redmond and Smith, 1980: p. 27.
8. Redmond and Smith, 1 980: pp. 58-59. As early as December 1 947, Perry
Crawford had recommended to the Special Devices Division that work on the
cockpit be discontinued because it was "not essential to the program. "
9. Redmond and Smith , 1 980: pp. 109-1 1 , 1 20, 151-67. The estimated costs to
completion in 1 949 of the four computers were as follows: Whirlwind, $3 million
to $5 million; lAS, $0.65 million; UNIVAC, $0.45 million; and E DVAC , $0.47
million.
1 0. Redmond and Smith, 1 980: pp. 1 28, 1 71-73.

Copyrighted Material
370 Notes to Pages 203-208

1 1 . Crawford, 1 942.
12. M. M. Astrahan and J. F. Jacobs, 1983: "History of the Design of the SAGE
Computer-The AN/FSQ-7," Annals of the History of Computing 5, pp. 340-49,
especiaIly p. 341 .
1 3. The first director of Project Lincoln was F. Wheeler Loomis, on leave from
the University of Illinois.
1 4. M. Rees, 1982: "The Computing Program of the Office of Naval Research,
1946-1953," A nnals of the History of Computing 4, pp. 1 02-20, especially p. 1 14;
Redmond and Smith, 1 980: pp. 1 73, 1 9 1 -92, 1 99.
1 5. Redmond and Smith, 1 980: pp. 1 86, 1 95.
1 6. C. R. Wieser, 1 983: 'The Cape Cod System," Annals of the History of Computing
5, pp. 362-69. The early aircraft-tracking demonstrations that justified creating
the Cape Cod System were known as Project Charles.
1 7. J. w. Forrester, 29 April 1 947: "Data Storage in Three Dimensions," Servo­
mechanisms Laboratory Memorandum, No. M-70 (courtesy of MITRE Corpo­
ration Archives) .
1 8. Pugh, 1 984: pp. 66-69.
1 9. Pugh, 1 984: pp. 81-89.
20. J. w. Forrester, June 1 3-30, 1 949: MIT Computation �otebook, �o. 47,
pp. 1 5-27.
21. The graduate student was William N. Papian. Born in 191 6, Papian worked
as an electrician's apprentice and radio mechanic before going to the RCA
Institute to become a radio technician. Mter serving as a radio intelligence
officer during World War II, he entered MIT as a freshman at age thirty. See
W. N. Papian, 1 8 June 1 982: interview by E. W. Pugh.
22. Pugh, 1 984: pp. 76-81 .
23. D . R. Brown, l l June 1952: "Group Leaders' MeetingJune 9, 1 952" (courtesy
of MITRE Corporation Archives) . The person who met with J. C. McPherson
in New York City was �orman H. Taylor.
24. C. C. Hurd, 1 98 1 : "Early IBM Computers: Edited Testimony," Annals of the
History of Computing 3, pp. 1 63-82, especiaIly p. 1 74.
25. J. C. McPherson, 17 June 1952: to J. R. Shipman; Watson and Petre, 1990:
pp. 230-31 . As reported by McPherson, 'They [ the MIT representatives] visited
our various laboratories at Poughkeepsie. Endicott. and the Watson Laboratory,
and were given demonstrations of the Defense Calculator, the Tape Processing
Machine, the Eastman Printer, the Wire Printer (with but a single feed) ,
Hamilton 's Drum Calculator. and Xerographic continuous printing. They spent
some time at the Vestal Laboratory discussing the BDASA Project, and at the
Watson Laboratory we reviewed the NaRC Calculator plans, the Havens delay
unit design (old) , and were shown the cathode ray tube test apparatus, the
operation of the tape reader and controls, and electronic components test unit. "
26. J. F. Jacobs. 1 986: The SA GE Air Defense System-A Personal History (Bedford,
Mass.: MITRE Corporation ) . pp. 43-44. According to Jacobs the scores of the
three top contenders, based on the weighting scheme agreed to by the evalua-

Copyrighted Material
 Notes to Pages 209-21 I 37I

tors before the visits, were as follows: IBM, 1816; Rem ington Rand, 1 374; and
Raytheon, 1 067.
27 J . W. Forrester, 12 May 1953: to A. G. Hill, "Selection of a Company to Work
.

with the Lincoln Laboratory on the Transition System. " Accord ing to Forrester,
this memorandum expands on information provided in an earlier memoran­
dum of 5 November 1 952. The four representatives from MIT who evaluated
and ranked IBM, Remington Rand, and Raytheon were Jay W. Forrester, Robert
R. Everett, Norman H. Taylor, and C. Robert Wiese r. The Bell Telephone
Laboratories and RCA were also among the top five contenders, but they "chose
to withdraw from being considered after the preliminary m ee ti ng on the basis
,

that their existing commitments did not allow them to undertake the work. "
Several secondary sources cite Sylvania as a lead in g contender, but Forrester's
May 1 953 memorandum indicates it was not among the top five.
28. The two individuals se n t by Forrester to review IBM's work on magnetic
core memories were William N. Papian and Kenneth H. Olsen. Both were
graduate students working on different aspects of magnetic core memories as
thesis projects. I n 1 953 a number of MIT staff members were assigned to IBM's
Poughkeepsie laboratory to help with the final testing of the prototype com­
puters; one of them was Kenneth Olsen, who in 1 957 founded the Digital
Equipment Corporation (DEC) .
29. M. K. Haynes, 28 August 1 950: "Magnetic Cores as Elements of Digital
Computing Systems," a thesis submitted in parti al fulfillment of the requ ire ­

ments for the degree of doctor of philosophy in electrical engineering in the

graduate college of the University of Illinois.
30. M. K. Haynes, 26 March 1 952: to C. L. Snyder, vice president, General
Ceramics and Steatite Corporation; Pugh, 1 984: pp. 45-57, 1 05. The experi­
enced engineer who worked with Mike Haynes on the early ferrite core memory
was Edward Rabenda.

3 1 . Pugh, 1 984: p. 95.

32. J. W. Forrester, 15 July 1 953: to AN /FSQ-7 Engineers, "Planning, Scheduling
and Administering the AN/FSQ-7 Program. "
33. R. R . Everett, 1 983: Ed itor s Note," Annals of the History of Computing 5,
" '

p 340 .
.

34. A. P. Kromer, 26 January 1953: to all conferees, "Minutes of Joint MIT-IBM

Conference Held at Hartford, Connecticut, January 20 1953" (document,
,

courtesy of the MITRE Corporation Archives, Bedford. Mass. ) . That the meet­
ings were held in a back room of the IBM branch office in Hartford is related
by R. P. Crago, in H. S. Tropp et aI. , 1983: "A Perspective on SAGE: Discussion,"
Annals of the History of Compu ting 5, pp. 375-98, see p. 384.
35. M. M. Astrahan, 12 June 1968: interview by L. M. Saphire .

36. J. M. Coombs, 20 October 1 967: interview by L. M. Saphire. John Coombs

worked on navy electronics projects at the National Cash Register Company
during World War II and joined ERA when it was formed in 1946. He joi ne d
IBM in 1 952 and initially organized a school for teaching newly hired engineers
about computer technologies and methods. He was given the job of adminis­
trator in charge of IBM's e ngin eering work on SAGE. Later Robert P. Crago
Copyrighted Material
3 72 Notes to Pages 21 1-216

became the engineering leader with both admin istrative and technical respon­
sibilities.
37. A. P. Kromer and R P. Mayer, 29 june 1 953: to AN/FSQ-7 Planning Group,
"Project Grind Meeting ofjune 25, 1 953 (Second D ay) (document courtesy of "

the MITRE Corporation Arc hives , Bedford, Mass. ) . See also M. M. Astrahan, 1 2
june 1 968: interview by L . M . Saphire, especially tape I , p . 41.
38. J. W. Forrester, in H. S . Tropp et al. , 1 983: "A Perspective on SAGE:
Discussion, " Annals of the History of Compu ting 5, pp. 375-98 especially pp. 382-,

83. Henry S. Tropp se rved as mode rator of a discussion among Herbert D.

Benington, Robert Bright, Robert P. Crago Robert R Evere tt jay W. Forrester,
, ,

Joh n V. H arrington John F. Jacobs Albert R Sh iely Norman H. Taylor and C.

, , , ,

Robert Wiese r. The d i scussion was held at the MITRE Corporation on 26

October 1 982. The date of the decision to use a duplex system is provided by
M. M. Astrahan and J. F. Jacobs, 1 983: "History of the Design of the SAGE
Co m puter-The AN/FSQ-7," Annals of the History of Computing 5, pp. 340-49;
see p. 348.
39. Pugh, 1 984: pp. 1 07-1 5.
40. M. M. Astrahan and J. F. Jacobs, 1 983: H istor y of the Design of the SAGE
"

Computer-The AN/FSQ-7 A nnals of the History of Computing 5, pp. 340-49.

,"

41 . IBM press release, 30 June 1956: "IBM-Built Computer Is Heart of Electronic

Air Defense Syste m , date-l i ned Kingston, �.y. Characterization of the SAGE
"

comp u ter as an "electronic brain" would probably not have been per m itted by
Watson, Sr. , but he had died earlier that month.
42. It has often been suggested that the name SAGE was chosen to match that
of Nathan iel Sage the director of MIT's Division of Industrial Cooperation,
,

who had shepherded Proje ct Whirlwind and its predec essor project through
the government contract processes. See, for e xample , J. W. For reste r in H. S. ,

Tropp et al 1 983: p. 394.

.,

43. IBM press rel e ase , 30 June 1 956. The designation AN/FSQ stood for
Army-Navy/Fixed Special Equi pment.
44. IBM Kingston News, 9 October 1 957: "New Array Increases Memory 8-Fold";
IBM brochure circa 1 958: "The SAGE Compute r "; IBM Advanced Systems Sales
,

Manual, 15 October 1 962: "Ground Based Data Processors"; R R Everett, C.

A. Zraket, an d H. D. Ben i ngton 1983: "SAGE-A Data-Processing System for
,

Air Defense, Annals of the History of Computing 5, pp. 330-39, adapted from a
"

paper pre sente d at th e 1957 Eastern Jo in t Compute r Conference.

45. Astrahan and Jacobs, 1 983 : pp. 347-49.
46. J. F. Jacobs, 1983: "SAGE Ove r vi ew," Annals of the History of Computing 5,
pp. 323-29.
47. The study of group behavior in man-machine system s was proposed by John
L. Kennedy, a psychologist and consultant to Rand.
48. C. B aum , 198 1 : The System Builders: The Story of SDC (Santa Monica, Calif : .

Syste m Development Corporation) , pp. 23-27. Wesley Melahn, who was in

charge of
the first five Rand programmers to work on SAGE, later became the
second p re siden t of the System D eve l opmen t Corporation. The first presi de n t
was Melvin O. Kappl e r .

Copyrighted Material
 NoLes to Pages 21 7-224 373

49. R. R. Everett, editor's note on paper by H. D. Ben i n gto n , 1 983: "Production

of Large Computer Pro grams ," Annals of Lhe History of Computing 5, pp. 350-61 .
50. H . D. Benington, 1 983: "Production of Large Computer Programs," Annals
of Lhe History of Computing 5, pp. 350-6 1 . Th is article contains a long foreword
by Be n ing ton , followed by a paper adapted from one presented in Washing to n ,
D.C., in J un e 1 956 at a symposi um on advanced pro gramm i ng methods for
digital computers, sponsored by the Navy Mathematical Computi n g Advisory
Panel an d the Office of Naval Research. Be n i ng ton describes the paper as "a
description of the organization an d te c h n ique s we use d at MIT's Lincoln
La bora tor y in the m i d 1 95 0s to produce programs for the SAGE ai r-de fen se
-

system." The cost of programming was estimated by Be n i n g ton at about $ 55

per instruction. The importance of the 8.5-fold increase in SAGE computer
m emor y size to the operating program is highl igh ted by Benington in H. S.
Tropp e t al., 1 983: p. 377.
5 1 . G. Burck and the editors of Fortune, 1 965: The Computer Age and Its Potential
for Management (New York: Harp e r & Row) chap. 2, ''' On Line' in 'Real Ti me
, '."

5 2 . Astrahan and Jacobs, 1 983: p p . 347-49.

53. Defendant's exhibit 2609A in U. S. v. IBM ( 1 969) , u.S. District C our t ,

Southern District of New York, pp. 34-1 50; R. P. Crago, 19 De ce mber 1 978 :
testim ony in U. S. v. IBM, U.s. D is tr ic t Court, Southern District of New York,
pp. 85963--64 . IBM's revenue from SAGE in the years 1 952 through 1 959 was
$478 million. During the same time IBM's worldwide revenue from electronic
stored-program computer systems was $1 ,32 1 millio n and its total worldwide
revenues were $ 7 ,301 million .

1 6 Chasing New TechMlogies

1 . W. W. McDowell, 4 January 1 950: to J. C. McPherson.
2. The engineering person n e l in Poughkee psi e reached 359, 750, an d 1 ,040 by
year end 1950, 1951 , and 1 952, resp ectively. See Yearbook, 1 952: IBM Engineer­
ing Laboratory, Poughkeepsie, N .¥.
3. Th e four individuals who joined Reynold Johnson in establishing the new
laboratory in San Jose were as follows: James Hood, from th e administrative
staff of th e Endicott labo ra tor y was chosen to help establish administrative
,

procedures; Harold F. Martin, a recent graduate of the California Institute of

Technology, became Jo h nson s administrative assistant; Lou i s D. Stevens, an
'

experienced engineer from the Po ughke eps i e laboratory, served as Johnson's

technical ass i stan t; and William H . Reid, fro m World He ad q uar ters , w as as­
signed by W. W. McDowell to assist in other organ izatio n al matters. See R. B.
Johnson , 1 5 December 1 957: "Remarks on IBM San Jose Stor y. "

4. R. B. Johnson, 1 5 December 1 957: "Remarks on IBM San Jose Story. "

5. A. J. Critchlow, September 1 952, as quoted by D. W. Kean, 1977: IBM San
Jose: A Q!tarter Century of Innovation (San Jose: IBM) , p. 2 1 .
6 . The special represe n tative from the San Francisco branch office was Gle n E.
Perkins, who had joined IBM as a systems service man in New York City in 1 925;
see Business Mac h i n e s 1 963 Christmas issue: pp. 1 4-1 7.
,

7 . Joh n son 1 5
, December 1 957. .
Copyrighted Matenai
3 74 Notes to Pages 224-229

8. G. A. Hotham was the first to propose storing information magnetically in OJ

stack of disks continuously spinning on an axle. He documented this proposal
in his engineering notebook on 14 January 1 953. Because of the cooperative
mode of operation encouraged in the laboratory by R. B. Johnson. it is difficult
to identify specific contributions of individuals. A chronology of the project
reveals that development of the random access file was assigned to W. A.
Goddard in April 1 953. Goddard assigned D. D. Johnson to develop the air
bearing head in early May and J. W. Haanstra to develop the magnetics for the
disk and recording head in late May. In August J. J. Lynott was assigned the tasl<
of developing a servo-drive system for moving the read-write head. and a
method devised by him and T. Noyes was adopted in January 1 954. In October
1 953 J. J. Hagopian developed a spin-coating method for achieving the desired
magnetic layer on the aluminum disks. In November 1 953 L. D. Stevens was
promoted to development engineer and put in charge of the random access
disk project, while R. B. Johnson increasingly devoted himself to newer ad­
vanced technology concepts. See L. D. Stevens, December 1 95 7: "Chronological
Summary-Evolution of the RAMAC."
9. J. Rabinow. August 1 952: "The Notched-Disk Memory." Ekctrical Engineering .

pp. 745-49.
10. IBM Press Release. 6 May 1955.
Noyes and W. E. Dickinson, February 1 956: Proceedings of thR Western joinl
1 1 . T.
Computer Conference, pp. 42-44.
1 2 . In 1 991 IBM be gan manufacturin g 16-megabit memory chi ps that were used
in computers beginning early in 1 992. Each chip had about half the storage
capacity of RAMAC. See the Wall Street journal, 1 4 June 1 99 1 : "Two U. S. Firms
Outpace Japan Rivals-IBM Has Lead in Producing Latest Chip."
1 3. IBM sales manual of electronic data processing machines, 4 September 1 956;
C. J. Bashe, L. R. Johnson. J. H. Palmer, and E. W. Pugh. 1 986: IBM 's Early
Computers (Cambridge, Mass.: MIT Press) , p. 300.
1 4. B. E. Phelps, 15 May 1956: to H. P. Luhn, "RAMAC Addressing Scheme."
1 5. Digital Computer Newsktter 8, October 1956: p. 402; R. P. Daly. February 1 956:
"Integrated Data Processing with the Univac File Computer." Proceedings of tlu
Western joint Computer Conference, pp. 95-98.
1 6. E. W. Pugh. L. R. Johnson , and ]. H. Palmer, 1 99 1 : IBM s 360 and Early 3 7C
'

Systems ( Cambridge. Mass. : MIT Press) . pp. 2 46-59. 261 -70. 495-52 1 , 774n39.
1 7."The IBM 350 RAMAC Disk File," 27 February 1 984: brochure produced
by The American Society of Mechanical Engineers to commemorate their
selection of RAMAC as an International Historic Landmark.
1 8. New York TirTU'S, 1 July 1 948: 'The News of Radio, " as quoted in C. Weiner.
January 1 973: "How the Transistor Emerged." !FEE Spectrum. pp. 24-�3.
1 9 .J. Bardeen and W. H. Brattain. July 1 948: 'The Transistor. A Semi-Conductor
.

Triode, " Physical Review, pp. 230-31 .

20. A. H. Dickinson. 1 8 May 1 950: 'Transistors in IBM."
2 1 . C. J. Bashe. L. R. Joh nson . J. H. Palmer. and E. W. Pugh . 1 986: IBM:I Earl)
Computers, p. 377. Ralph Palmer assigned the task of developing IBM's expertise

Copyrighted Material
 Notes lo Pages 229-233 375

in �Iid-:state te c h nologie s to Arthur L. Samue l , who had joined IBM from the
�mve rslty of I ll i nois i n Sep te mbe r 1949. Sam uel was best known at the time for
hi s w
ork o n el ectro
nic vacuum tube design at the Bell Telephone Laboratorie s ,
w�er� �e had spe nt more than fifteen years prior to joining the University of
Ilh n Oi s m 1 9 46.

22. Leon B rillouin was hired by Columbia University from Harvard with the
h elp of T. J. Wa tson , Sr. , who arranged for him to have an office at the Watson
Laboratory an d to be paid as an IBM consultant. See H. Fleisher, 1 N ove mber
1982: discus sion wi th E. W. Pugh.
23. J. C. Logue, April 1 968: interview by L. M. Saphire. Joseph C. Logue
organized the Course that produced the transis torized 604 and educated many
e ngineers in the new solid-state technologies. Logue had graduated from Cor­
nell in 1 944 with a bachelor 's degree in electrical engineering. He joined IBM
in 1951 after teaching at C orn e ll , earning his master's degree, and working for
two years at the Brookhaven N ation al Laboratory.
24. New York Times, 8 October 1954: "IBM to Expand Laboratory, Shows Tran­
sistor Calculator,� p. 1 ; IBM press release, 8 October 1 954.
25. Bashe et al., 1 986: p. 386.
26. W. W. McDowell, 2 October 1 957: to L. H. LaMotte, J. A. Haddad, H. R.
Keith , H. W. Miller, and E. R Piore , So lid State Comp on en ts "
" .

27. S. W. Dunwell, March 1 968: interview by L. M. Sap h ire .

28. C. C. H urd, 2 1 January 1 955: "High Speed Data Processing Mac h i n es ; 3 1 "

January, "Project Whitney"; Bashe et aI., 1 986: pp. 423-25.

29. C . C. Hurd and W. W. M cDowel l 20 April 1 955: to D. A. Br uc e purchasing
, ,

agent, "Proposal, Computer"; D. A. Br uce 10 May 1955: to C. C. Hurd, "liver­

,

more Computer"; Bashe et aI., 1 986: p. 1 27 Bruce advised Hurd on 10 May

.

1955 that the decision was based upon "lowest overall cost" in wh ich the
"predominant" factor was delivery schedule.
30. R L. Pal m er , 1 3 September 1 955: to fil e , "Proj e ct S tretch . "

31 . For a detailed accoun t of Project Stretch, see Bashe et aI. , 1 986: pp. 41 6-58.
The name "Stretch" had been adopted in Aug ust 1 955 to signify the in te n t of
Palmer and McDowell to stretch well beyond the state of the art in specifying
components for IBM's next supercomputer. See C. C. Hurd 31 Au gust 1955: ,

to L. H. LaMotte .
32. C. C. Hu rd and B. F. Wiegard, 29 February 1 956: to W. H. Bru mme tt Jr. ,

This six-page letter accompanied IBM's 53-page proposal to the Atomic Energy
Com m i ssion s Contracts and Procuremen t Branch in Albuquerque, N.M.
'

33. Faw, 29 February 1 958: to file, "Project Stre tch

H. A. . n

34. H. A. Faw, 2 0 January 1956: memorandum of meeting, Proj e c t Stretch­ "

AEC"; C. C. Hurd, 30 March 1 956: to A. L. Williams, "NSA"; U n ited States

Atomic Energy Co m m i ssion 20 November 1 956: Contract No. AT(29-2)-476.
,

35. P. K. Spa tz, 1 May 1 959: "The STRETCH Development Program," IBM staff
report, p. 26.
36. E. W. Pugh, 1984: MemtWies Thai. Shaped an Induslry (Cambridge, Mass.: MIT
Press) , pp. 1 27, 1 66-75. A particularly important innovation was the load-shar-

Copyrighted Material
376 Notes to Pages 234-237

ing matrix switch that greatly reduced the number of memory drive circuits
and the power required from each circuit.
37. Bashe et aI., 1 986: pp. 390-98. Cost-effective designs that facilitated manu­
facture and service were emphasized in all Stretch components with the expec­
tation that they would be used in commercial computers as well. Rather than
using ''worst<ase'' circuit design methods pioneered in SAGE, the Stretch
engineers devised computer-implemented statistical designs that achieved good
reliability with a significantly better speed/cost ratio.
38. Bashe et aI., 1986: pp. 4O�1 4. Component lead wires were automatically
inserted in predrilled holes in SMS printed-circuit cards, under guidance of an
electronic controller directed by instructions read from punched cards. Reliable
electrical interconnections were achieved by an automated process that sol­
dered all component leads simultaneously to preprinted copper lines on the
back of each card. Back-panel wiring that interconnected circuits between cards
was automatically performed by a paper-tape<ontrolled wire-wrapping ma­

chine. It was one of the earliest examples of digital design information being
generated in a form directly applicable to the manufacturing process.
39. Bashe et aI. , 1 986: pp. 446-49. The IBM 7080, announced in January 1 960
as a higher-performance replacement for the IBM 705 III, was the second
member of the 7000 series. It used the same components as the 7090.
40. IBM Poughkeepsie Laboratory Newsletter, 12 May 1 96 1 : "Stretch at Los Alamos";
E. Bloch, 15 May 1 96 1 : to file, "Summary of Stretch Acceptance Test in Los
Alamos."
41. E. W. Pugh, 20 July 1 965: "Scientific Compute Factor versus Processor
Installation Date," prepared as background for the "IBM Corporate Memory
Strategy" of 1 965.
42. The first computer to challenge the performance of Stretch was Atlas, which
was developed in a cooperative effort between Manchester University and
Ferranti Limited in Great Britain. Completed in 1 962, it is reported to be
"probably the most powerful machine available in the early 1 960s"; see S. H.
Lavington . 1 978: "The Manchester Mark I and Atlas: A Historical Perspective, "
Communications of the A CM 2 1 , pp. 4-1 2.
43. The simulator was devised by H . G. Kolsky, a former lASL mathematician
who joined IBM in 1957, and John Cocke, a Stretch planner; see Bashe et aI.,
1 986: pp. 442-43. John Cocke achieved considerable fame in the 1980s as the
inventor of RISC architecture.
44. C. R. DeCarlo, 2 February 1 961: to W. B. McWhirter. "STRETCH. "
4 5 . Data71Ultion, June 1 961 : "The Shrinking of Stretch."
46. S. W. Dunwell, 8 April 1 964: to T. ]. Watson, Jr.
47. T. ]. Watson, Jr. , 20 April 1 964: to S. W. Dunwell.
48. T. ]. Watson, Jr., 24 March 1 966: remarks at the IBM Annual Awards Dinner.
The IBM Fellow program was initiated in 1 963 to reward creativity of engineers,
programmers. and scien tists by allowing them freedom to define and carry out
their own research programs.
49. Ford, Bacon , and Davis, Inc. , 22 April 1 954: "Report on Organization.
Engineering Department, �yri§� »mlMM/Machines Corporation."
 Notes to Pages 237-240 377

50. Unknown to the younger Watson at the time, his father had even arranged
for private lunchtime meetings with McDowell almost every day for several
months after McDowell was promoted to director of engineering in 1 950. See
Watson, Jr., and Petre, 1990: p. 203.
5 1 . Business Machines, 1 October 1954: "R L. Palmer Is Named Director of
Engineering in Departmental Expansion Program."
52. G. L. Tucker, 3 1 October 1977: interview by M. J. Schiller.
53. IBM Pough keepsie Engineering Newsletter, 1 7 January 1 956: "IBM Engi­
neering Reorganizes-Research to Be Independent under Palmer; Beattie
Heads Product Development Here."
54. IBM Poughkeepsie Engineering Newsletter, 1 7 January 1 956.
55. Business Machines, 22 October 1956: "For IBM: A New Director of Research";
28 December 1 956: "A New Plan for Corporate Organization Is Announced."
56. M. C. Andrews, 2 February 1 983: interview by E. W. Pugh. The "three wise
men" selected by E. R Piore to advise him were M. Clayton Andrews, John W.
Gibson, and Gardiner L. Tucker.
57. IBM Research Magazine, November/December 198 1 : "25 Years of Innovation. "
This entire forty-page issue is devoted to th e history of research in IBM, with
emphasis on the twenty-five years since an independent corporate research
organization was formed.
58. In 1956 Arthur L. Samuel, of the IBM Poughkeepsie laboratory, pro­
grammed an IBM 704 to play checkers, using a method in which the machine
could "learn " from its own experience. It is believed to be "the world's first
self-learning program, and as such a very early demonstration of a fundamental
concept of artificial intelligence. " See G. Wiederhold and ]. McCarthy, May
1992: "Arthur Samuel: Pioneer in Machine Learning, " IBM Journal of Research
and Develcrpment 36, pp. 329-3 1 . Significant IBM research achievements in solid­
state lasers began in 1 960 when Peter P. Sorokin and Mirek]. Stevenson created
the second and third operational solid-state lasers. The first was created by T.
H. Maiman of the Hughes Aircraft Company in June 1 960. Considerable pub­
licity was accorded the IBM Mark I language translator when its translation of
Nikita Khrushchev's speech of May 1960 to the Supreme Soviet was submitted,
just ten days after it was delivered, to the Congressional Committee on Science
and Technology. The speech dealt with the fligh t of the U.S. U-2 spy plane over
the Soviet Union. The Mark I and Mark II translators were developed by a
research group headed by Gilbert King. The field of integer linear program­
ming was initiated by Ralph E. Gomory at Princeton University. Mter joining
IBM in 1 959, he continued to lead this new field until he became IBM director
of research in August 1 970. See Bashe et aI., 1984: pp. 559-65.
59. E. W. Pugh, 1 984: Memories That Shaped an Industry (Cambridge, Mass.: MIT
Press) , pp. 2 1 4-18. The company's first magnetic film memory was shipped on
a System/360 Model 95 to NASA in New York City in February 1 968. One month
later the second and last Model 95 was shipped with its film memory to the
NASA Goddard Space Flight Center. Its 1 20-nanosecond read-write cycle made
it the fastest multimegabit memory in operation for several years. See Pugh et
aI., 1991 : pp. 469-75.
Copyrighted Material
378 Notes to Pages 242-250

60. Research was elevated to the status of an IBM division in November 1 963.
Directors of IBM Research have been Emanuel R. Piore. 1 956--1 960; Gilbert W.
King. 1961-1962; Gardiner L. Tucker. 1 963-1 967; Arthur G. Anderson. 1 967-
1 970; Ralph E. Gomory. 1 970-1986; John A Armstrong. 1986--1 989; and James
C. McGroddy. 1989-.

17 Legacy
1 . Giving the eulogy at the Presbyterian church on Park Avenue. New York City.
where Watson had been an elder. the Reverend Paul Austin Wolfe provided
insight into Watson 's success, saying: "He had a peculiar singleness of mind,
and he saw things simply. It was because of that simplicity of mind that he could
make decisions. " The minister also observed. "Even though he made machines,
he never forgot that they were made by people and for people. " See Fortune,
September 1956: 10m Jr.'s I.B.M .. .. by Robert Sheehan. pp. 9-1 2. 1 1 3-1 7.
2. Herman Hollerith incorporated his business as the Tabulating Machine
Company in December 1 896? three months after he signed a contract to supply
equipment to the New York Central Railroad.
3. UNIVAC was an acronym for UNIVersal Automatic Computer. The Defense
Calculator. which was introduced to the public on 7 April 1 953 as the "IBM
Electronic Data Processing Machines." was already better known simply as the
"70 1 . " See IBM Record. April 1 953: "New IBM Electronic Data Processing Ma­
chines Unveiled at World Headquarters in New York. "
4. A brief discussion of the relations between IBM and Remington Rand. and
their agreement of 193 1 . may be found in Complaint. 21 January 1952: U.S. v.
IBM. U.S. District Court. Southern District of New York, N.Y. . pp. 7-9. Permit­
ting users to purchase cards from other suppliers. without paying a rental
penalty. had little effect on IBM's business. in part because the rotary press
invented by Fred Carroll made the company the lowest-cost producer of cards.
During the next two decades, the company's share of the sales of punched cards
continued to match its increasing share of installed equipment.
5. J. W. Cortada. 1993: Before the Computer (Princeton: Princeton University
Press) , pp. 68-70. 1 1 8-25.
6. Supporting information for much of the material in the first three sections
of this chapter may be found in previous chapters. The strengthening of IBM's
engineering and manufacturing operations during the depression of the 1930s
and World War II. for example. may be found in chapters 5 and 8. respectively.
7. A brief summary of the settlement of the antitrust case against General
Electric may be found in G. Wise. 1 985: WiUis R. Whitney, General Electric, and
the Origins oj U. S. Industrial Research (New York: Columbia University Press) .
p. 1 38.
8. T. G. Belden and M. R. Belden. 1 962: The Lengthen ing Shadow (Boston:
Little. Brown) . pp. 96--9 7.
9. Complaint. 21 January 1 952: U.S. v. IBM. U.S. District Court. Southern District
of New York, N.Y.. p. 9. The complaint noted that Remington Rand machines
were available for purchase or rent. whereas IBM machines could only be

Copyrighted Material
 Notes to Pa{(ts 251-256 379

rented. The estimated annual rental paid by the government in 1 951 was $25
million, up from $15 million two years earlier.
10. Complaint, 2 1 January 1 952: p. 1 5.
I I . A. D. Neale and D. G. Goyder, 19S0: The Antitrust Laws of the U.S.A. : A Study
of Competition Enforced by Law, 3 d ed. (Cambridge: Cambridge University Press) ,
pp. 1-33, 90-1 1 7; especially pp. 20, l OS. The last quotation in the paragraph is
from Judge Hand in United States v. Aluminum Company of America (decided in
1945) as quoted in Neale and Goyder, p. l OS.
12. ''The General Market for Accounting Machines," 2 1 September 1953: draft
stipulation of fact prepared by IBM for US. v. IBM, District Court of the United
States, Southern District of New York, pp. S-25. The principal companies with
which IBM said it competed were R. C. Allen Business Machines, Inc., Bur­
roughs Corporation , Clary Multiplier Corporation, Elliott Addressing Machine
Co., Felt and Tarrant, Inc., Friden Calculating Machine Co., Marchant Calcu­
lating Machine Co. , Monroe Calculating Machine Co. , McBee Co., National
Cash Register Co. , Remington Rand, I nc., Royal Typewriter Co., L. C. Smith &
Corona Typewriters, Inc., and the Underwood Corporation .
1 3. Lehman Brothers, 1 9 May 1 952: "Burroughs Adding Machine Company:
Study of Company and I ts Position in the Business Machine Industry, " Table 1 4
as summarized in Cortada, 1 993: p . 256, Table 1 7. 1 l .
14. Newsweek, 4 February 1952: "Monopolies: IBM on the Spot. "
15. Complaint, 21 January 1 952: pp. 1 7-1 S.
1 6. Complaint, 2 1 January 1 952: pp. 6, 1 4.
1 7. Complaint, 21 January 1 952: pp. 1 9-2 1 .
I S . A good overview of U.S. an ti tr ust law i s provided by Neale and Goyder, 1980.
19. T. J. Watson, Jr., was appointed president of IBM on 15 January 1952, and
the Justice Department filed its antitrust suit on 21 January 1952.
20. T. J. Watson, Jr., and Peter Petre, 1 990: Father Son & Co. (New York: Bantam
Press) , pp. 2 1 6-20, 268-70; Belden and Belden, 1 962: pp. 291 -3 1 3 .
2 1 . Final Judgment, 2 5 January 1 956: US. v. IBM, U.S. District Court, Southern
District of New York, N.¥., pp. 1 , 6-29.
22. Belden and Belden, 1962: p. 309.
23. Final Judgment, 25 January 1 956: pp. 1 6-1 8.
24. Final Judgment, 25 January 1 956: pp. 6-7.
25. Final Judgment, 25 January 1 956: pp. 8-1 5.
26. Final Judgement, 25 January 1 956: pp. 1 2-14.
27. Final Judgment, 25 January 1 956: pp. 1 8-24.
28. New Yom Times, 26 January 1 956: "IBM Trust Suit Ended by Decree-Com­
pany Agrees to Sell Its Electronic Computers and License All Patents. " Stanley
N. Barnes was the chief of the Justice Department's Antitrust Division.
29. Business Machines, 1 June 1956: "T. J. Watson, Jr., Named IBM Chief Execu­
tive," pp. 3-4.

Copyrighted Material
380 Notes to Pages 25 7-259

30. The Electronic Data Processing Industry, 1 956: a study prepared by Arthur
D. Little, Inc., Cambridge, Mass., with the cooperation of White, Weld & Co.,
members of the New York Stock Exchange. This document reported that all 1 9
planned installations o f the IBM 7 0 1 were completed, and that 1 1 o f the
planned 14 installations for the IBM 702 were completed. There were 44 orders
for the recently announced IBM 704 and 1 14 orders for the recently announced
IBM 705. The report did not give the number of Sperry Rand systems installed
or on order.
3 1 . The Electronic Data Processing Industry, 1956. A typical IBM 650 system
rented for about $3,500 per month. An equivalent sales price of fifty times
monthly rental would have been about $1 75,000.
32. In 1960 the gross revenue of the World Trade Corporation reached $360
million, or 20 percent of IBM's total revenue. Most of World Trade 's revenue
came from Europe (67 percent) . The remainder came from Canada ( 1 4 per­
cent) , South and Central America ( 1 1 percent) , and the Asia-Pacific area
including Japan (7 percent) . See Fortune, November 1960: "Q. What Grows
Faster than I.B.M.? A. I.B.M. Abroad," by R. Sheehan.
33. J. Connolly, 1967 : History of Computing in Europe ( New York: IBM World Trade
Corporation) , pp. 1 0-18.
34. M. Campbell-Kelly, 1 989: ICL: A Business and Technical History (Oxford,
England: Oxford University Press) , pp. 1 28, 1 4 1 -43.
35. Belden and Belden, 1 962: p. 298; Campbell-Kelly, 1 989: p. 1 40.
36. Two letters provide insight to the nature and intensity of the dispute
between IBM and BTM over their relationship. In January 1 936 Watson, Sr.,
wrote to Sir G. E. Chadwyck-Healey of BTM saying in part: "Mr. Braitmayer
informs me that your competitor has more machines in use than your company.
When we compare your business and the potentialities of your territory with
the business done in other countries, it prompts me to respectfully suggest that
you and Mr. Phillpotts, who, I understand, are the largest stockholders, come
to the United States, and we will go into the matter." Chadwyck-Healey re­
sponded quickly, saying in part: "I have to repeat that we cannot afford to
continue to pay you Royalties and, in addition, income tax on their amount.
As I have said to you before, English law makes any agreement to do so null
and void, and there is, indeed, a penalty enforceable here on any person who
refuses to allow the deduction to be made. " Excerpts from these letters from
the BTM Board Papers are here quoted from Campbell-Kelly, 1989: pp. 93, 94.
3 7 . Many believed that Watson's fondest dream had been realized when a
low-priced accounting system called the 3000 series was developed in the
laboratories of the German company for the worldwide market. Representatives
from twelve countries had participated in setting its specifications. It was the
first major product from the World Trade laboratories. Renting for only $340
per month and using smaller punched cards than IBM's conventional equip­
ment, it was expected to compete with small-card systems already offered by
Powers-Samas, BTM, and Machines Bull. Over 800 sales were recorded in the
first six months after its introduction in the spring of 1 960. See Fortune, Novem­
ber 1 960: "Q. What Grows Faster Than I.B.M.? A. I.B.M. Abroad," by R. Sheehan.
Powers-Samas had introduced low-priced machines using small cards of two
Copyrig/ited Material
 Notes to Pages 259-265 381

different sizes (4.75 x 2 inches and 2.75 x 2 inches) in 1 932, and BTM had
attempted to match this with a half-length IBM card in 1 936; see Campbell­
Kelly: pp. 85, 99. Despite its initial success, the 3000 series became an embar­
rassment in the marketplace because it was in trod uc e d half a year after the
more versatile and cost-effective IBM 1 40 I .

38. Cam pbell-Ke l ly, 1 989: pp. 43-47, 73-76.

39. W. Desborough 7 June 1 949: "Re mingto n Rand Tabul atin g Machine Divi­
,

sion: General Impressions, " Powers-Sam as Board Papers, as quote d in Camp­

bell-Kelly, 1 989: p. 1 40.
40. Campbell-Kelly, 1 989: pp. 1 39-41 .
4 1 . L. Heide , 1 99 1 : Production: The Development of
"From Inven tion to
Punched-Card Machines by F. R. Bu l l and K. A. Kn u tsen 1 9 1 8-1 930," Annals oj
,

the Histcrry oj Computing, pp. 261-72.

42. Connolly, 1967: pp. E-12, E-1 5 .
4 3 . Fcrrtune, November 1 960.
44. Campbell-Kelly, 1 989: p. 1 49. Arthur K. (Dick) Watson succeeded H. K.
Chauncy as president of the World Trade Corporation in June 1954.
45. The head of BTM's research, Doc Keen , was said to be " open ly hostile
towards the new technology"; see Campbell-Kelly, 1989: p. 1 57.
46. After BT!'vl and Powers-Samas merged in 1 958 to form International Com­
puters and Tabulators Limited, a decade of mergers followed before a stable
state was reached in 1 968 with the creation of In ternational Com pute rs Limited
(ICL) . Strongly encouraged by the British government, ICL was the coun try's
response to the dramatic success achieved by I BM Syste m/ 3 60 announced in .

1964. See Campbell-Kelly, 1 989: pp. 1 44-264.

18 Gambling on System/360
1 . IBM Business Machines, 28 December 1 956: "Special Issue on Williamsburg
Conference. "

2. The staff functions reporting to AI Williams inclu de d finance,

legal, person­
nel, communications, and co m me rc i al
development as well as marketing, ,

manufacturing, and research and engineering support ser vices. Creating the
autonomous Service Bureau Corporati on (mandated by the antitrust settlement
of 1956) was probably Learson's most challenging assignment because of the
profound impact it had on the way IBM worked with customers and because
of the problems associated with transferring employees to an independent
busi n ess unit..

3. R. L. Palmer was promoted to produc t development manager of the Data

Processing Division .

4. E. W. Pugh , L. R. Johnson, and J. H. Palmer, 1 99 1 : IBM 's 360 and Early 370
Systems (Cambridge Mass. : MIT Press) . Appendix E. pp. 65 1-69, provides sev­
.

eral organization charts for the IBM Corporation from 1 956 through 1 976. The
newly formed Advanced Systems Development Division (ASDD) was headed by
Jerrier A. Haddad. He had been engineering man ager of the IBM 701 devel­
opment project. More recently he had headed the Special Engineering Products
Copyrighted Material
382 Notes to Pages 265-270

Division , most of whose activities were included in ASDD. The Data Processing
Division, wi th its manufacturing and development responsibilities removed in
1 959, reported to Orland M. Scott, who also had resp onsibility for the Federal
Systems Division. The Time Equipment Division was sold to the Simplex Time
Recorder Company in October 1 958. Its profit in 1958 was about $8 million or
roughly 6 percent of IBM's total profit.

5. T. J. Watson, Jr., and P. Petre, 1990: Father, Son & Co. (New York: Bantam
Press) , p. 200; chap. 1 3 of this book.

6. C. J. Bashe, L. R. Johnson, J. H. Palmer, and E. W. Pugh, 1 986: IBM's Early

Computers (Cambridge, Mass.: MIT Press) , pp. 465-74. Charles E. Branscomb
had primary responsibility for developing the IBM 1 40 1 computer, the design
of which had resulted from a competition between IBM engineers in France,
Germany, and the United States beginning in mid- 1 955. The objective had been
to develop an accounting machine capable of competing in the world market
with low-cost machines offered by Machines Bull of France and others. Francis
O. Underwood, of the Endicott engineering group, first showed that conven­
tional accounting machine functions of relatively small businesses could be
implemen ted with a stored-program computer more effectively than by any
other electronic means.

7. F. M. Fisher, J. W. McKie, and R. B. Mancke, 1 983: IBM and the U. S. Data

Processing Industry: An Economic History (New York: Praeger Publishers ) , pp �2- . .

53, 1 0 1 . Final Report of SPREAD Task Group, 28 December 1961 , reprinted in

Annals of the History of Computing 5 1 983: pp. 6-26; see p. 24. The numbers of
installed computers listed in these two references are not in full agreement.
Those from the SPREAD report are used here. Development of the 1 403 chain
printer, used on the 1 401 computer, is discussed in Bashe et aI. , 1 986: pp. 489-
93.
8 . Fisher et aI. , 1 983: especially p. 60.

9. M. Campbell-Kelly, 1989: ICL: A Business and Technical History ( Oxford:

Un iversity of Oxford Press) , pp. 201-03.

1 0. Bashe et aI. , 1 986: p. 1 19. The first six computers to be announced with
transistor circuits were the 7070, 7090, 1 40 1 , 1 620, 7080. and Stretch. No two
of these were compatible with each other.

1 1 . Bashe et aI.. 1986: pp. 474-80.

1 2. B. O. Evans, 23 March 196 1 : to W. B. McWhirter, uDS Product Line."

1 3. Pugh et aI., 1 99 1 : pp. 1 1 4-22. This book provides detailed accounts of the
development of SLT and IBM System/ 360 and of the creation of the IBM
Components Division .

1 4. F. P. Brooks, Jr 1 983: "Discussion o f the SPREAD

.• Report, June 23, 1982,"
Annals of the History of Computing 5, p p . 27-44.
1 5. The 8000-series design team, which included Gerrit A. Blaauw, was aug­
mented in August 196 1 by several others from Research. including Gene M.
Amdahl, John Cocke , and Elaine M . Boehm. See Brooks, Jr., 1 983: p. 30.

1 6. Final Report of SPREAD Task Group, 28 December 196 1 : J. w. Haanstra,

chairman, B. O. Evans, vice chairman , with J. D. Aron , F. P. Brooks, Jr., J. W.
Fairclou gh , W. P. Heisin I!cJl'Y»g»1ew ��riM' Johnson , M. J. Kelly, D. V.
 Notes to Pages 271-275 383

Newton, B. G. Oldfield, S. A. Rosen, and J. Svigals. Reprinted in A nn als of the

Hi.�tory of Computing 5, 1 983: pp. 6-26.
1 7 . Pugh et aI., 1991 : pp. 1 22-24.
1 8. Final Report of SPREAD Task Group, 28 December 1 961 : p. 7, I I .
1 9. Final Report of SPREAD Task Group, 28 December 1 961 : p . 1 1 .
20. M. V. Wilkes, July 195 1 : ''The Best Way to Design an Automatic Calculating
Machine," Manchester University Computer-Inaugural Conference, pp. 1 6-18.
21. John W. Fairclough was the Hursley laboratory representative to the
SPREAD Task Group. So promising was his proposal that the report com­
manded, "Microprogram controls using a read-only memory shall be employed
unless the cost performance of a conventional control system is less than 2/3
that of a microprogram control system." See Final Report of SPREAD Task
Group, 28 December 1961 : p. 1 2 .
22. The study o f instruction repertoires from across the industry to identify
in�tructions used solely for business or scientific applications was led by J. W.
Haanstra. See Pugh et aI. , 1 991 : pp. 127-28.
23. Chief antagonists in this debate were Gene M. Amdahl, who favored a 6-bit
character and Gerrit A. Blaauw, who favored the 8-bit byte. See Pugh et aI.,
199 1 : pp. 1 48-49, 634.
24. Final Report of SPREAD Task Group, 28 December 1 96 1 : p. 12. Another
imfortant architectural decision was to provide the capability to address up to
22 or 1 6,7 77 ,2 1 6 characters in memory. A so-called base register addressing
scheme was adopted that reduced the bits required in each address from 24 to
1 6. This reduction was particularly helpful to designers of the smaller machines.
See Pugh et aI., 1 99 1 : pp. 1 45-48.
25. Final Re po r t of SPREAD Task Group, 28 December 1961 : p. 9.
26. Pugh et aI., 1 991 : pp. 157-58.
2 7 . IBM document, undated: "Analysis of H 200 Activity as ofJanuary 27 , 1964. "
The document lists various types of orders for the H-200 in IBM's eastern,
midwestern, western, and federal sales regions. Altogether 196 orders for H-200
systems were listed, involving 1 26 IBM accounts.
28. Pugh et al., 1 99 1 : pp. 1 57-64. Announcement of the upgraded 1 40 1 com­
puter was scheduled for 15 February 1 964; see G. F. Daly, lO January 1 964: to
D. E. Slattery.
29. B. O. Evans, 8 January 1964: to P. W. Knaplund, "IBM Future Products. "
30. S. G. Tucker, 1 965: "Emulation of Large Systems," Communications ofthe A CM
8, pp. 753-6 1 . The first emulator was devised in 1 963 by Stuart G. Tucker and
Larry M. Moss in the IBM Poughkeepsie laboratory. It permitted programs
written for the IBM 70 70 to be run on one of the larger System/360 computers.
3 1 . By showing that the smallest of the new computers could emulate the
operation of the 1 401 economically, some of Haanstra's own engineers had
defeated his effort to have Endicott retain control over its product line. See E.
W. Pugh, 1 984: Memories That Shaped an Industry (Cambridge, Mass.: MIT
Press) , pp. 203-04. See also Pugh et aI., 199 1 : pp. 1 60-62.

Copyrighted Material
384 Notes to Pages 275-282

32. T. ]. Watson, Jr., 7 April 1 964: "A Letter from the Chairman, " published in
IBM News, 7 April 1964.
33. IBM News, 7 April 1 964: "IBM Announces System/ 3oo."
34. Pugh et aI., 1 99 1 : pp. 83, 1 56. The 200-fold performance range, projected
by the SPREAD report in December 1961 , was not achieved until the Model 91
(announced in November 1 964) was shipped in October 1967 with three to ten
times the performance of the Model 75. The System/360 Model 20 is not
included in the performance range because it was not fully compatible with
other System/360 processors. It had, for example, only 37 instructions rather
than 1 43. See A. Padegs, 198 1 : "System/360 and Beyond, " IBMJournal o/Research
and Develojnnent 25, pp. 377-90; Pugh et aI., 1 99 1 : pp. 420, 643.
35. IBM press release, Technical Information, 7 April 1 964: "Forty-four Devices
Announced for IBM System/360"; IBM press release, 7 April 1964: "for release
from Poughkeepsie at 1 2 Noon. "
36. IBM News, 7 April 1 964; ]. T. Ahlin, 7 May 1 964: to C . E . Branscomb e t al.,
"System/360 Orders"; L. L. Horn, 1 4 May 1 964: to J. T. Ahlin et aI. , "System/360
Requirements. "

19 Commitment and Delivery

1 . E. W. Pugh, L. R. Johnson , and J. H. Palmer, 1 99 1 : IBM's 360 and Early 370
Systems (Cambridge, Mass. : MIT Press) , pp. 1 62-64. The referenced book pro­
vides substantially more information about most topics found in this chapter.
2. Pugh et aI., 1 99 1 : pp. 62-63.
3. P. N. Whittaker, 23 January 1958: "IBM-Texas Instruments Agreement Sum­
mary," with attachment. A producer of electronic equipment during World War
II, Texas Instruments had begun developing and producing transistors soon
after they were invented. Among IBM's reasons for selecting TI was its reputa­
tion as a technical leader.
4. Pugh et aI., 1 991 : pp. 46, 62-64.
5. Minutes of the Eleventh Meeting of the R & D Board, 10 March 1960:
presentation by B. N. Slade; F. M. Fisher, ]. W. McKie, and R. B. Mancke, 1983:
IBM and the U.S. Data Proce.uing Industry-An Economic History (New York:
Praeger Publishers) , pp. 7 1 -75.
6. Pugh et aI. , 1 99 1 : pp. 56-58.
7. E. W. Pugh, 29 October 1 957: "Report of First Meeting. " In attendance were
M. ]. Kel ly H. T. Marcy, R. L. Palmer, E. R. Piore, and E. W. Pugh. See also
,

Business Machines, April 1 959: "IBM Engages Top Consultant," p. 5.

8. R. H. Bullen, 8 December 1 964: to A. K. Watson, "History of the Components
Division. "

9. E. Bloch, 18 April 196 1 : to B. O. Evans, "Advanced Technology Study." The

task group was established on 3 April 1 961 and reported to B. O. Evans.
1 0. E. W. Pugh. 1 984: Memories That Shaped an Industry (Cambridge, Mass.: MIT
Press) , pp. 57, 1 72.

Copyrighted Material
 Notes to Pages 283-288 385

1 1 . Bloch, 1 8 April 1 961 . The eight engineers who served on Bloch 's task force
were selected from the General Products and Data Systems divisions and Group
Staff.
1 2 . E. Bloch, 1 August 1 961 : "Solid Logic Technology Program-Objectives and
Directions. ..
1 3. The printed lines and resistors had a unique pattern for each circuit type,
with different sizes and numbers of resistors and different locations for mount­
ing silicon chips. Each circuit type comprised only four basic electric elements:
transistors and diodes fabricated on silicon chips, and resistors and conducting
lines printed on the ceramic substrate. The ceramic substrates were 0.455 inch
square and 0.06 inch thick, and the silicon transistor chips were 0.025 inch
square. A newly devised slurry saw cut more than one thousand fully processed
chips out of each 1 .25-inch-diameter silicon wafer. See Pugh et aI., 1 99 1 :
pp. 74-8 1 .
1 4. E . L . Fritz, 1 3 March 1 986: interview by E . W. Pugh; W. J . Schuelke, April
1 967: "Modular Approach to System Design," Automation. The new design
philosophy also provided for easier repair of individual production units with­
out shutting down the entire line.
1 5. IBM News, 25 May 1 964: "MRL Pioneers SLT Production" and "New Design
Concepts Will Offer Maximum Flexibility in New Manufacturing Facility"; Pugh
et aI., 199 1 : pp. 83-92. The decision to manufacture SLT cards and boards
internally had been made to protect proprietary information, to accommodate
engineering design changes more rapidly, and to ensure quality control.
1 6. Robert L. Judge, a manufacturing engineer, had already demonstrated that
cores with 1 9 mil inside diameter and 32 mil outside diameter could be wired
by creating work-hardened bullet-shaped ends of the wires to act as their own
needles as they were driven through the holes in the cores in an array. Cores
of this size were used in the memories being developed for System/360 Models
62 and 70. On his own, Judge extended this capability to the 30 percent smaller
cores with 1 3 mil (0.33 mm.) inside diameter and 21 mil (0.53 mm. ) outside
diameter. These smaller cores helped make it possible to operate the memories
at an 0.75 microsecond read-write cycle. The inventor of IBM's load-sharing
switch was Gregory Constantine, Jr. See Pugh, 1 984: pp. 231-36.
17. Pugh, 1 984: pp. 230-37. The Model 60 had been designed to use a 2-mi­
crosecond cycle memory and the Models 62 and 70 were designed for a
I-microsecond memory that was in serious technical difficulty before responsi­
bility for memory development was transferred to Erich Bloch in the Compo­
nents Division. A crucial change in the Model 65 from the Models 60 and 62
was the use of a control store rather than wired-in logic for control as used in

the Model 75. The Model 65 was thus able to emulate the older IBM 7070
computer, which was important to customers who wanted the greater perfor­
mance of System/360 but had a large investment in software written for the
7070, The self-motivated effort of the IBM engineer, Fernando (Fred) Neves,
who made possible the use of a control store on the Model 65 , is described in
Pugh, 1 984: pp. 225-29.
18. Pugh et aI., 199 1 : pp. 247-6 1 . The IBM 1 30 1 Disk Storage provided 50 tracks
per inch and 520 bits per inch on a track, compared to 20 and 1 00, respectively,
Copyrighted Material
386 Notes to Pages 288-292

on RAMAC. Average access time had been reduced to less than 0.2 secon d from
0.6 second. A single 1 30 1 storage unit could have one or two modules of 25
disks each. Each 25-disk module had capacity for up to 28 million 6-bi t charac­
ters-nearly six times the 5-million-character capacity of the 50-disk RAMAC
storage unit.
19. Pugh et aI., 1 991 : pp. 261-68. Disk packs were given the product designation
of IBM 1 3 1 6. When used on the IBM 1 3 1 1 Disk Storage, a track density of 50
per inch was achieved. A track density of 1 00 per inch was achieved on the IBM
231 1 Disk Storage.
20. Pugh et aI., 1 991 : pp. 228-31 .
2 1 . J . T. Ahlin, 7 May 1 964: to C. E . Branscomb e t aI., "System/360 Orders"; 28
May 1 964: to L. L. H or n "System/360 Requirements"; P. W. Knaplund, 1 1
,

September 1 964: to C. E. Frizzel and G. F. Kennard, "System/360 Order Posi­

tion. "
22. IBM News, 1 6 July 1 97 1 : "CD Observes Its Tenth Anniversary This Week. "
Development facilities built on the same site were larger than the manufactur­
ing facilities.
23. IBM news release, 26 October 1 965. Noting that 300 systems had already
been installed, the release said in part: "International Business Machines Cor­
poration announced today that its customers are being advised that during 1966
most System/ 360's will be delivered 60 to 1 20 days later than originally sched­
uled. The delays will result from problems in building up the rate of production
of the System/360 as rapidly as necessary to meet the unprecedented customer
demands for the new equipment. "
24. Pugh et aI. , 1 991 : pp. 99-1 03. The person who took over responsibility for
SLT production from Bernard N. Slade was Edward J. Garvey, one of IBM's
most respected manufacturing managers. Previously he had been in charge of
manufacturing in Endicott.
25. E. L. Fritz , 1 3 March 1 986: interview by E. W. Pugh. Elliot Fritz believes that
the slurry saw designed for cutting silicon wafers into chips is the first tooling
invention to be patented by IBM; see U.S. Patent 3,241 ,265, filed on 27 June
1 963 by H. Wing.
26. Pugh, 1 984: pp. 245, 249-5 1 ; W. Newman, 29 April 1 982: interview by E. W.
Pugh. Construction of the Boulder plant began in March 1 965.
27. IBM news release, 27 May 1 964: "Executive Promotions Announced. " Just
as A. K. Watson had half a year as head of the corporate staff before being
promoted in May 1 964 to senior vice president (a newly created position) , so
Yin Learson was put in charge of sales half a year before he was promoted to
senior vice president. See Pugh et aI., 1 99 1 : pp. 1 65, 1 69-73.
28. T. J. Watson, Jr., and P. Petre, 1990: Father, Son & Co. (New York: Bantam
Press) . pp. 342-60, especially p. 355.
29. H. E. Cooley, June 1 987: interview by E. W. Pugh. The "four horsemen "
were John W. Gibson , John W. Haanstra, Clarence E. Frizzell, and Henry E.
(Hank) Cooley. Lincoln L. Horn, an IBM vice presiden t, served as Learson's
first assistant in getting System/360 manufactured and shipped.

Copyrighted Material
 Notes to Pages 292-295 387

30. IBM news release, 25 January 1 966: "Executive Elections and New Corporate
Office Announced"; IBMagazine, September 1 970: "For ' Dick' Watson, a New
World of Policy and Diplomacy."

3 1 . General purpose software support offered "free of charge" by IBM and other
systems manufacturers by the early 1 960s typically included utilities that per­
formed routine tasks such as copying records from magnetic tape onto printed
pages; sort routines that rearranged records from one sequence to another;
assemblers that translated a machine-specific language with mnemonic user aids
into actual machine language; and compilers that translated high-level lan­
guages like FORTRAN and COBOL into machine language for specific com­
puters. Also provided were sequences of instructions for operating input-output
devices. By the time System/360 was being defined, most large computing
systems were supported by an operating system that combined all of these support
programs with a control program that controlled system operation by initiating
general purpose programs as needed to process a set of application programs.
32. Pugh et aI., 1 99 1 : pp. 298- 1 2 .
33. F. P. Brooks, Jr. , 23 January 1 963: to G . A. Grover, "Software Design . "
34. L . C . Wood, 2 3 January 1 964: to C. H . Reynolds, "NPL Programming
Review. " New Processor Line (NPL) was the designation of System/360 during
its devel opment Carl H. Reynolds managed the programming activities of the
.

Data Systems Division.

35. Report of the Strategy Committee of the NPL Programming Task Force, 5
March 1964: to C. H. Reynolds. Members of the committee were R. P. Case, C.
L. Foster, M. deY. Roberts, W. P. Simonet, and E. F. Wheeler. The use of a single
operating system (with modular functional levels) had not previously been
embraced because it added yet another level of technical c hal len ge Introduc­
.

tion by IBM in 1963 of operating systems for two small disk-based computers
with memory capacities of only 1 6K characters gave reason to believe that
reduced-function versions of the proposed operating system could be used on
computers with as little as 1 6K bytes of main memory.
36. IBM Product Announcement, 7 April 1 964: System/360, Letter 264-26; IBM
Programm in g Announcement, 7 April 1 964: IBM System/360-Programming
Support Package, Letter P64-56. See also Pugh et aI. , 1 99 1 : p. 320.
37. It was during this time that Fred Brooks introduced the term "architecture"
to the software development process; see Pugh et aI. , 1 99 1 : p. 3 1 8.
38. F. P. Brooks, Jr. , 6 December 1 993: 'The Design of Design, " a talk presented
at the IBM T. J. Watson Research Center in which he spoke briefly about his
early years at IBM. See also Pugh et aI. , 1 99 1 : p. 343.

39. F. P. Brooks, Jr., 1 975: The Mythical Man-Month: Essays on Software Engineering
(Reading, Mass.: Addison-Wesley) , especially p. 25. In the preface, Brooks says,
'This book is a belated answer to Tom Watson's probing questions as to why
programming is hard to manage."

40. IBM Programming Announcement, 31 December 1 964: IBM System/360

Programming Support Letter P64-201 . During November Earl F. Wheeler
,

agreed to accept the commitment to develop the three new operating systems
on the schedule announced in December. The plan for these developments
Copyrighted Material
388 Notes to Pages 296-300

had been devised by James H. Frame, whom Wheeler had replaced as top
programming manager in the General Products Division in October.
4 1 . Pugh et al., 199 1 : pp. 32 1-3 1 .
42. M . Phister, Jr., 1 979: Data Processing Technology and Economics, 2 d ed. (Bedford,
Mass.: Digital Press) , pp. 3�5.
43. Datamation, December 1 964: "RCA's New Spectra 70 Series," pp. 34-36.
44. IBM's Proposed Findings of Fact, January 1 982: pp. 4103, 4 1 25, 4145,
4 151H>6, 4 19�202. This 5,979-page document was prepared by IBM for U.S.
v. IBM, filed 1 7 January 1 969: United States District Court for the Southern

District of New York. Intent on becoming second only to IBM in the computer
industry, RCA hired L. E. Donegan, Jr., and J. Rooney from IBM in 1 969 to
become vice presidents of sales and marketing, respectively, of the RCA com­
puter division .
45. K. D . Fishman, 1 98 1 : The Computer Establishment (New York: Harper & Row) ,
pp. 1 64-65. General Electric retained an 18.5 percent interest in the just­
formed Honeywell Information Systems, and it continued to offer time-sharing
services with its own equipment.
46. Phister, Jr., 1 979: p. 39. The Control Data Corporation was founded in 1 957
by William C. Norris, who had been one of the founders of Engineering
Research Associates (ERA) in 1 946.
47. B. Elzen and D. MacKenzie, 1 994: 'The Social Limits of Speed: The Devel­
opment and Use of Supercomputers," IEEE Annals of the History of Computing,
pp. 41H>1 ; IBM's Proposed Findings of Fact, January 1 982: pp. 4281-88. Con­
trol Data was unable to produce the CDC 6800, scheduled for delivery in May
1967, because of technical problems in achieving the promised high perfor­
mance in a machine compatible with the 6600. The CDC 7600 was announced
in 1 968 as a replacement for the never-delivered 6800. First delivered in January
1 969, the 7600 was not compatible with the 6600.
48. Watson, Jr., and Petre, 1 990: pp. 383-87; IBM's Proposed Findings of Fact,
January 1 982: pp. 4233-35, 427 1 .
49. Fishman, 1 98 1 : pp. 209-1 6; IBM's Proposed Findings of Fact, January 1 982:
p. 45 1 3 .
50. Pugh e t aI. , 1 99 1 : pp. 233-34, 269-70. By the end o f 1 970, IBM had installed
1 10,000 tape drives on System/360 computers. A particularly attractive target
for plug-compatible manufacturers was the IBM 2 3 1 4 Direct Access Storage
Facility. It provided up to eight disk drives (plus a spare) and a storage capacity
of 29.2 million bytes per disk pack. Announced in April 1 965, its cumulative
revenue in the United States was $1 .2 billion by 1972. This was more than the
cumulative revenue of either of the two largest revenue-producing System/360
processing units, the Models 30 and 40.
5 1 . Pugh et aI. , 1 99 1 : pp. 289-95, 522-30. Prominent in the formation of the
Storage Technology Corporation (STC) was Jesse I. Aweida who had led the
development of the most advanced of the IBM 2400 tape drive series on which
STC's first plug-compatible product was based.
52. Electronics Week, 14 January 1 985: "Award for Achievement," pp. 40-44.

Copyrighted Material
 Nores to Pages 300-305 389

53. F. M. Fisher, J. W. McKie, and R. B. Mancke, 1983: IBM and the U. S. Data
Processing Industry (New York: Praeger Publishers) , pp. 307-16. The Greyhound
Corporation, Boothe Computer Corporation, and Itel were among the primary
computer leasing firms. IBM identified 250 leasing firms in 1970, up from 92
in 1966. The popularity of System!360 and certain economic factors facilitated
their spectacular growth. Leasing companies purchased a total of $2.5 billion
of IBM equipment in 1969, up from $200 million in 1965.

20 Onrush of Technology
1 . J. W. Haanstra, September 1 964: "Monolithics and IBM," IBM Technical
Report.
2. T. J. Watson, Jr., and P. Petre, 1990: Father, Son & Co. (New York: Bantam
Press) , p. 354; A K. Watson, 9 October 1964: to J. W. Gibson and E. R Piore.
Design of the "New Generation Systems," which were to be built entirely with
NGT (variously called Next or New Generation Technology) was underway by
early 1 965; see P. Fagg and E . S. Hughes, 27 January 1 965: Minutes of meeting
on the "New Generation Systems."
3. IBM News, September 1 965: Special Edition, "Systems Development Division,
Emphasis on Technical Excellence and Leadership"; IBM News, 16 July 197 1 :
"Components Division Tenth Anniversary This Week."
4. This quotation is based on the author's own recollections of Haanstra's
statements upon several occasions. To survive in this environment, technical
managers had to agree to achieve objectives they believed were impossible. The
result was distrust of one another and reduced ability to work together to
achieve common goals.
5. See chap. 19, "Stress at the Top. " John Haanstra was one of the "four
horsemen" selected by Learson to help solve the System! 360 manufacturing
problems. Getting expert technical help to solve problems in component manu­
facturing had become more difficult because creation of the Systems Develop­
ment Division (SDD) had organizationally separated the component
development and component manufacturing functions. Removed as president
of SDD, Haanstra was assigned to the Federal Systems Division under Bob Evans.
Haanstra remained there until August 1967 when he joined the General Electric
Company to become general manager of the GE Information Systems Equip­
ment Division eight months later. He died in the crash of a small plane he was
piloting in August 1 969.
6. E. W. Pugh, L. R Johnson, and J. H. Palmer, 1 99 1 : IBM's 360 and Early 3 70
Systems (Cambridge, Mass.: MIT Press) , pp. 432-40. IBM's proprietary solder­
ball technology for attaching semiconductor chips to printed lines on ceramic
modules was known as controlled<ollapse chip connections, or simply C-4. The
integrated circuit technology on which it was first introduced was MST, which
stood for monolithic systems technology.
7. Pugh et a!., 1 99 1 : pp. 406-21 .
8 . IBM Product Announcement, 30 June 1970: "IBM System!370-Powerful
New Computing System , " letter 270-52; Pugh et al., 1 991 : pp. 109, 482-84,
495-96. The System!370 models that replaced the earlier lower-performance
Copyrighted Material
390 Notes to Pages 305-309

System/360 models were offered at a price increase of about 25 percent. This

was about equal to the inflation during the intervening years.
9. A. Padegs, 1 98 1 : "System/360 and Beyond," IBM Journal of Research and
DevellJjJment 25, pp. 377-90, especially pp. 379, 380. IBM's ability to modify the
standard, as it did in moving from 360 to 370 architecture, was especially
objectionable to companies whose business plans had been devised to profit
from the market for System/360 products. From a broader perspective, how­
ever, privately created standards (such as IBM's) induce more rapid technologi­
cal progress than do traditional industry standards that are typically achieved
through years of intense negotiations. The creator of a private standard has an
incentive to modify it continually for his own benefit, but if these modifications
do not generally benefit customers as well, the standard will lose its appeal and
its value. See, for example, C. R. Morris and C. H. Ferguson, March-April 1 993:
"How Architecture Wins Technology Wars," Harvard Business Review, pp. 86-96.
1 0. E. W. Pugh, D. L. Critchlow, R. A. Henle, and L. A. Russell, 1981 : "Solid
State Memory Development in IBM," IBMJournal of Research and Develapment 25,
pp. 585-602.
1 1 . R. P. Case, 29 April 1 970: "Relocation: Its Meaning and Value," IBM SDD
technical report, especially p. 3. Dynamic relocation was particularly beneficial
in a computer used to service remote terminals that might initiate new pro­
grams or request infrequently used data. To keep such information in main
memory was impractical, but to require users to move it into low-cost storage
between uses was not convenient. The solution was to provide automatic paging
of information between disk and main memory.
1 2. Data Processing, October-December 1 961 : "ATLAS-Europe 's Largest Data
Processing System," pp. 255-60.
1 3. Pugh et aI. , 199 1 : pp. 355-67, 481-85. Bob O. Evans was "ostracized" to the
position of president of the IBM Federal Systems Division from January 1 965
to October 1 969. He then replaced Charles E. Branscomb as president of SDD
in October 1 969.
1 4. E. W. Pugh, 1 971 : "Research Storage Study, " a task-force report prepared
for the director of Research; E. W. Pugh, 1 985: "Technology Assessment,"
Proceedings of the IEEE 73, pp. 1 756-63. Inherent in this analysis was the realistic
assumption that the ratio of revenue received to manufacturing costs would
change very little compared to the projected change in manufacturing costs.
The optimistic cost-reduction projections for semiconductor memories cor­
rectly assumed a switch from bipolar devices to FETs.
15. Pugh et aI., 1 99 1 : pp. 538-42. Richard P. Case was selected to work with
John R. Opel to organize and manage the FS task force. Beginning his IBM
career as an electrical engineer in the Endicott laboratory in 1 956, Case had
advanced to engineering manager for a mid-range 1 400 series computer before
he moved into programming in 1 962 to participate in the development of
software support for System/360. He had recently been appointed director of
advanced systems for SDD when he was asked to assist Opel with the FS task
force.
1 6. G. Radin, 26 March 1 990: in terview by E. W. Pugh. The term "single-level
store" was used to distin2:Pish th�_ brQad c;on.c�pts embodied in FS storage
fJOpyngntea Marenal
 Notes to Pages 309-312 391

management from the simple "one-level store� implemented by the British

designers of ATLAS. George Radin was the lead architect for FS.
1 7. E. W. Pugh et aI., 8 April 1 970: "Storage Hierarchy White Paper," a task-force
report prepared for the Corporate Technical Committee; 9 December 1 970: to
the Corporate Technical Committee, "Storage Hierarchy White Paper-Eight
Months Later. "
18. R. P. Case, 1 9 June 1973: "FS Project-Schedule History."
1 9. B. O. Evans, 1 6 November 1 974: to file. "A Potential Systems Direction. "
20. B. O . Evans, 1 9 February 1 975: to distribution. "Product Plans"; R . P. Case,
27 January 1 994: private communication . IBM spent approximately $1 billion
per year on research and development during the three years 1 972 through
1 974. Expenditures on FS increased from about $ 1 00 million in 1972 to more
than $400 million in 1 974. Probably somewhat less than half of the FS expen­
ditures were a complete loss.
2 1 . F. P. Brooks, Jr. , 2G-22 June 1 973: hand-written notes made during a two-day
review of the FS project; R. P. Case, 1 8 August 1 973: summary of the comments
made by F. P. Brooks, Jr. , during the FS project review in June.
22. G. M. Amdahl, February 1 979: "The Early Chapters of the PCM Story,"
Datamation, pp. 1 1 3-16. The Heizer Corporation, a venture capital firm, made
an initial $2 million investment in the Amdahl Corporation. Subsequent fund­
ing of $6 million each was provided by Fujitsu, Ltd., of Japan and Nixdorf
Computers of Germany.
23. IBM Product Announcement, 25 March 1977: "Large Systems An-
.. noun cement Overview," Letter 277-20; "3033 Processor Complex Offers
Growth, Improved Price/Performance and New Function , " Letter 277-21 . John
E. Bertram, who had earlier headed the Advanced Computer Systems (ACS)
project in Los Gatos. was president of the relatively small Advanced Systems
Development Division when he made the proposal to revert from FS to 360-370
architecture. He was promoted to the newly created position of vice president
for development and manufacturing of the Systems Products Division to lead
the development of the post-FS systems. In recognition of his contributions,
IBM established the internal John E. Bertram Award for Sustained Technical
Excellence. The first recipient of the award was John Cocke in 1990. See IBM
Research Magazine, Fall 1 990: 'lohn Cocke: Gentleman, Scholar and Super
Computer Scientist," pp. 7-9.
24. Pugh et ai. , 1 99 1 : pp. 64 1-42. It is common to refer to a "System/390
architecture" to emphasize its new features. Nevertheless System/390 architec­
ture is an evolutionary improvement on that of System/370, just as System/370
architecture was an evolutionary improvement on that of System/360.
25. Besides using only a subset of the System/360 instruction set, the Model 20
had a reduced register set and a different treatment of I/O. See Padegs. 1 98 1 ,
especially p. 379.
26. IBM Product Announcement, 30 July 1969: "System/3, Featuring New Size
Punched Card and Low Price Disk Capability, Introduced for Small Business
Use"; IBM News, Rochester, Minn., 10 September 1 969: "System/3 Uses Mono­
lithic, SLD-IBM's Most Advanced Circuitry. " Lawrence A. Wilson transferred
Copyrighted Material
392 Notes to Pages 312-318

from IBM's Endicott laboratory to San Jose to lead the development of low-cost
products-an effort that produced the System/3; see Pugh et aI., 199 1 : pp. 443-
51.
27. IBM Product Announcement, 24 October 1978: "IBM System/38 An­
nounced"; D. N. Reynolds and G. G. Henry, August 1 979: "The IBM System/38,"
Datamation, pp. 14 1-43; G. G. Henry, 20 April 1 990: interview by E. W. Pugh .
28. Datamation, 1 5 July 1 991 : "Can IBM Recast the AS/400?" pp. 28-3 1 . The
estimated annual revenue from the AS/400 and related products according to
this reference was $14 billion. The annual revenue of the Digital Equipment
Corporation was $ 1 2.9 billion in 1990 and $ 1 3.9 billion in 1 99 1 .
29. The Electric Typewriter Division had become a "completely autonomous
operation" in 1 955. See Business Machines, 24 October 1955: "The Division That
Won a Promotion, " pp. 3-5.
30. A Myers, 6 May 1 970: to Communications Products File, "Terminal and
Office Products."
3 1 . Myers, 6 May 1 970.
32. IBM News Release, 9 September 1 975: "Portable Computer Announced by
IBM"; the IBM 5 1 00 was described as being "slightly larger than an IBM
typewriter. "
33. William C. Lowe, laboratory director of IBM's Entry Level Systems Unit in
Boca Raton, Florida, directed the activities that (with Frank Gary's strong
support) established the IBU in which the personal computer was developed.
The development was led by Philip D. ( Don) Estridge, who had joined IBM in
1 958 with a bachelor's degree in electrical engineering. Estridge worked as an
engineer on Project SAGE and as a systems programmer on the Apollo moon
mission at IBM before he undenook development responsibility for the per­
sonal computer.
34. IBM Product Announcement, 1 2 August 1 98 1 : "The IBM Personal Com­
puter"; IBM News Release, 12 August 1 991 : "IBM Celebrates 1 0th AnniverSary
of the IBM Personal Computer, a Decade of Industry Leadership"; J. Chposky
and T. Leonsis, 1988: Blue Magic (New York: Facts on File Publications).
35. IBM Research Magazine, Spring 1 986: 'The 801 Minicomputer," pp. 1-5;
Spring 1 990: "The Second Generation of RISC," pp. 2-4. The term RISC
(Reduced Instruction Set Computer) was coined at the University of California
at Berkeley. The comparative data on System/360 may be found in Pugh et al.,
1 99 1 .

2 1 Demamls of the Future

1 . Thomas J. Watson, Jr., died on the morning of 31 December 1 993 at age
seventy-nine. The cause of death was attributed to complications from a stroke
he suffered earlier in the month .
2. T. J. Watson, Jr. , an d P. Petre, 1 990: Father; Son & Co. (New York: Bantam
Press) ; pp. 392-96. During his brief term as CEO, Learson settled the antitrust
case filed by the Control Data Corporation ( CDC) in December 1968. A crucial
part of the settlement allowed CDC to purchase IBM's Service Bureau Corpo­
ration for considerably l � b �
o
�e811)fatJ�a1stimated market value. The
 Notes to Pages 31 9-320 393

purchase involved the transfer of 1 ,700 people. To ensure that these people
would receive "their existing IBM major benefit plans," the company contrib­
uted $26 million to support these benefits "not for any limited period, but
throughout their careers in SBC. " See Frank Cary, 1 973: to IBMers, "Letter from
the Chairman ," Think 39, inside cover.
3. Watson, Jr., and Petre, 1 990: p. 395.
4. F. M. Fisher, ]. ]. McGowan, and ]. E. Greenwood, 1 983: Folded, Spindled, and
Mutilated (Cambridge, Mass.: MIT Press) , pp. 1-17, 1 1 0-2 1 . As in the antitrust
case settled in 1 956, a key issue was market definition . The reference cited here
provides an excellent discussion of these issues, although it may be considered
biased because the authors served as economic consultants to I BM during the
antitrust case.
5. IBM's Proposed Findings of Fact, January 1 982: pp. 2-3. This 5,979-page
document in four large volumes was prepared by IBM as part of its defense
effort in U.S. v. IBM, filed 1 7 January 1 969: United States District Court for the
Southern District of New York.
6. S. Brill, April 1 982: "What to Tell Your Friends about IBM," American Lawyer.
The magazine's publisher said, "The Justice Department pushed on foolishly
with a case it never defined, never could manage, and therefore couldn't hope
to win. " The trial judge in the U. S. v. IBM suit that ended in 198 1 was David N.
Edelstein, who had previously presided over the earlier antitrust case that ended
in a consent decree in 1 956.
7. IBM's Proposed Findings of Fact, January 1 982: p. 8. In the first private
antitrust case initiated after the Justice Department filed its case, the Greyhound
Computer Leasing Corporation (in November 1 969) charged IBM with monop­
olizing a "market consisting only of leased computers." In a directed verdict in
mid-1972, the judge found in favor of IBM, saying IBM had gained its position
through "superior skills, foresight, and industry. " The second private case,
brought in January 1972 by the Telex Corporation and Telex Computer Prod­
ucts, Inc., charged IBM with monopolizing the electronic data processing
market. In an amendment to the case, Telex defined a narrower market
consisting only of certain types of peripheral equipment designed by IBM,
copied by others (including Telex) , and attached only to IBM processors. The
federal district court in Tulsa, Oklahoma, found in favor of Telex in September
1973 and awarded it $352.5 million (later corrected to $259 million ) in trebled
damages and injunctive relief. Simultaneously the court awarded IBM $21 .9
million damages for theft by Telex of trade secrets on which the copied
equipment was based. The antitrust decision against IBM was reversed by the
Court of Appeals for the Tenth Circuit in January 1 975. The parties agreed to
terminate further legal action, without payment of money by either party, when
IBM agreed not to pursue damages for theft of trade secrets. Adverse ramifica­
tions of the initial verdict in favor of Telex included encouragement of more
private antitrust suits by Calcomp, Hudson General, Marshall, Memorex, and
Transamerica-all filed before the end of 1 973. All cases were settled in IBM's
favor. See D & C Publishers, 1 974: Telex Plaintiff Versus IBM De/endent (Campbell,
Calif. : D & C Publishers) ; Fisher, Gowan, and Greenwood, 1 983: pp. 1 3-1 7; R.
J. Siegel, January-February 1 974: "Where We Stand on the Lawsuits," Think,
pp. 1 7-2 1 .
Copyrighted Material
394 Notes to Pages 320-322

8. R. E. Allen, 9 February 1 994: letter to shareowners, AT&T 1 993 Annual

Report, pp. 2-3. The outcomes of the AT&T and IBM antitrust cases were
announced on the same day by Assistant Attorney General William Baxter.
9. Gordon E. Moore in 1 993 suggested that the end of the exponential increase
in the number of devices on a silicon chip, predicted by his own "Moore's Law,»
was finally at hand. From the mid-1960s to the end of the 1 980s, the number
of circuits per chip had doubled every one to two years. See Electronic News, 5
April 1 993: "Statute of Limitations on Moore's Law?"
1 0. IBM News SDD Endicott, September 1 968: "North Street Laboratory Cele­
brates 35th Anniversary," pp. 1-4.

Copyrighted Material
Index

Advanced Systems Development Divi­ Astronomical Computing Bureau, 72

sion, 265, 269 Atanasoff, John V., 86-87, 133
Aiken, Howard H., 73 76 , 122- 12 3
- , Atlas computers, 141-142, 173,306-
124, 129, 268,342n14 307. 356n35, 376n42
Air defense system, 199-219, 235 Atomic Energy Commission (AEC),
Akers, John F., 328 231, 233
Amdahl, Gene M., 310-311, 382n15, Automatic computation, early, 71-
383n23 73, 342n 12
Amdahl Corporation, 310-311, Avery, Clarence W., 341n24
391n22 Aweida, Jesse I., 388n5 1
American Airlines, 219, 293
American Telephone and Telegraph Babbage, Charles, 7, 74, 184
Company (AT&T), 94, 138.238. Backus,John W., 126, 190-196
255. 320. See also Be ll Telephone Ballistic Research Laboratory, 115,
Laboratories 133, 137
American Totalisator Company, 141 Bardeen, John. 228
Antitrust suits Beattie. H. S. (Bud), 120-121
Control Data Corporation, 298 Bell Telephone Laboratories. 200,
General Electric Company. 249 216. 222, 228,307. See also Ameri-
IBM (1932), 247 ca n Telephone and Telegraph

IBM ( 1952 1956) 25 0-256 25 9

- , , Company
IBM (1969-1982),310,319-320, Bendix Aviation Corporation. 187,
392n2,393n7 257
National Cash Register Company. Berry, Clifford E., 86-87
33-34, 43 Bertram, John E., 311, 391n23
Remington Rand, Inc., 247, 260 Billings, John S., 3, 4, 6, 7, 8, 331n6
Applied Science Department, 156- BINAC, 141
158, 167, 169, 365nl Birkenstock, James w.. 166, 16 7 -

Architecture of computers, 135,305, 168, 279

309-312, 314-316, 387n37, 390n9, BIZMAC, 257
391n24 Blaauw, Gerrit A., 382n15, 383n23
Army Ballistic Research Laboratory. Bloch, Erich, 282-284
See Ballistic Research Laboratory BMEWS (Ballistic Missile Early Warn-
Army Signal Corps. See Signal Corps ing System), 235
Arnold Engineering Company, 206 Bombes, 104-105
AS/400. See IBM AS/400 computer Boulder, IBM facilities at, 291
series Braitmayer, Otto E., 110-111
AS C C (Automatic Sequence Con- Branch offices, 59, 189, 224
trolled Calculator), 72-76. See also Branscomb, Charles E., 382n6,
Mark I computer Copyrighted MalWlll1J 3
396 Index

Brattain, Walter H., 228 Columbia University

Brillouin,Leon,229,375n22 computer pioneering,71, 122-123,
British Tabulating Machine Com- 138, 157
pany (BTM), 18,l O4,258,380n36 educational research,67-69,70
Brooks,Frederick P., Jr.,268-270, Radiotype, 95
293-295, 3lO Statistical Bureau, 68
Bryce,James w., 4 1 Watson Laboratory,127-129
as adviser, 73-76, 77 Commercial Research Department,
EAM , 45,68, 96 63, 91, 93, 96. See also Methods re­
early years, 40-42,45 search
electronics, 81,82 Compatibility of systems,267-275,
patents, 77-81,86,250,325 279,293-294, 296-297,305, 317,
Bull, Frederik R.,260 388n47
BUNCH,297, 299 Competition
Bundy, Harlow E., 26 early computers, 131-143,155-156,
Bundy, Willard L., 26 173-174, 182,230-232,256-257,
Bundy Manufacturing Company,26, 266-267, 273
335n15 internal,47-48, 115-116,265, 267,
Bureau of the Census, 21-23,37, 289
139, 141. See also Census,U.S. international, 258-261,267
Burlington,IBM facilities at,290 System/360 era,296-300,307, 313,
Burroughs Corporation,187, 196, 316,319
257,296,297 tabulators,37-39,42-44, 250-253
Bush, Vannevar,129,132 Compilers, 193-197
Business practices,246-250, 252- Components Division,279-282,302
256 Computers
Business recessions,44-45,53,55- architecture,135, 305, 309-3lO,
56 311-312,314-315,387n37,390n9,
Byte,defined,272 391n24
definitions of,142-143, 170, 175,
Cache, 303-304 245-246
Calculation control switch, 71-72, early electronic,83-87,lO5-106,
343n19, 358n23 124-127,133-135, 152-156, 369n5
Cambridge University,261,354n19 improvements,329-330
Cape Cod System,204 mechanical, 7,71-76,342n12,
Card punches,9, 10, 38, 84, 332n22 343n19
Carroll,Fred M., 40, 46, 47, 55,78, stored-program,135-142,155-156,
325,378n4 185, 221,354n17, 354n18
Carroll press,47, 60,378n4 Computing Scale Company of Amer-
Cary, Frank T., 318-319,392n33 ica, 2-6,28
Case, Richard P., 390n15 Computing Scales Division,44
Census,U.S., 2-3,12-14,21-23. See Computing-Tabulating-Recording
also Bureau of the Census Company (CTR)
Chairmen, presidents,and CEOs of engineering, 37-52,338n9
IBM, 327-328 formation of, 23-27
Chase,George C.,73 headquarters, 45
Clocks,IBM, 42 patents, 25,42-44, 77-78,249-250,
COBOL, 196-197 325
Cocke, John,314-316, 376n43, Comrie, Leslie].,67
382n15,391n23 Connection box,38
Codd,Edgar F., 126, 190 Control Data Corporation (CDC),
Cold war. See National security 237, 296,297-298, 392n2
Colossus, 353n11 Control store, 271-272
Columbia machine, 68 John M.,371n36
Copyrighted MRPlPIa)'
 Index 397

Corporate headquarters, 45,188 Eckert , j. Pr espe r, Jr., 133-141, 156,

Corporate Management Co mmi tt ee, 184
264, 268, 313 Eckert, Wallace J.
Co rpo ra te Technical Committee, at Columbia University, 65, 71-72,
308, 313. See also Research and De­ 343n19
velopment Board computer pioneering, 65, 71-72,
CPC. Sell IBM CPC 123, 124-127,154, 343nI9
Crago, Robert P., 37 1n36 electronic comp uter s, 124-127,
Crawford, Perr y 0., Jr., 200-203, 132 -133, 154, 159-161
218,369n4 at Watson Laboratory, 12 3,127-
Cra y, Se ymour R., 297-298 129, 132-133, 157
Crib,99 and World War II, 123
Critchlow,Arthur j., 224 Eckert-Mauchly Computer Corpo ra­
Cross footing, 52, 84 tion (EMCC),132-141 , 173, 194
Cr yptogra ph y, 98-106 EDPM (Electronic Data Pro ces si ng
CSAW (Co mmun icatio ns Supple men ­ Machines),175-177, 245, 264
tary Activity-Washington), 103, EDSAC (Electronic Delay Sto rag e
105-106 Automatic Calculator), 185,
Customer support, 172-173, 188- 354n19
190, 24&-247, 314, 367n19, Education programs, 57-60,114 ,
367n2] 129, 172-173, 229
EDVAC (Ele ctroni c Discrete Variable
Data General Corporation, 299 Automatic Comput er ) , 135-137,
Da tamatic Corpo ration , 257 200, 202
Data Pro cessing Division, 264 Ehret, Clement, 111, 112, 350n8
Data Syste ms Division,264 Electric T ype w rite r Di vision , 263 ,
Defense Calculator, 167-174, 185- 264, 31 3
186. See also IBM 70 1 computer Electrodata Corporation, 257
De Florez,Luis,200, 201 Electromatic type wr iters, 96, 119,
Dehomag,18, 258, 334n44 121, 347n22, 351n5
Deltamax,206 Electronic computers. See Computers
De pa rt ment of Defen se,196, 202, El ectron ic Cont ro l Compa ny, 139
203 Ele ctro nics pioneering, 81-87, 119,
Depa rtment of Justice, 250-256, 319 121-127, 133-136, 149-155, 325
Design a uto ma ti on, 286 Em ploy ees, 117, 165, 237, 323-325
Deutsche Hollerith Maschinen Ge­ Emulator, 273-275, 383n30
se ll schaft . See Dehomag Endicott, IBM facilities at
Dickinson,A. Hal sey.80-81, 1l8, computers, 178-182, 265-267, 272,
123,228-229,325 273-275, 294,295-296
Digita l E qui pmen t Corporation EAM, 40,55-62,82, 92-95, 98-99,
(DEC), 298-299,312 163
DOS (Disk Operating System),295- electronics, 82-86,286
296 research, 78, 82-86, 222-223
Dun well, Steph en W'o 6 4-65, 9&- time-recording equip ment, 40
103, 154, 168,231, 23 3, 2 37, Endicott:Johnson Shoe C ompa ny,
341n24 55
Dynamic relocation, 305-307, 390n11 Eng ine eri ng. See also H iri ng
at CTR , 37-52
EAM (electric a cco un ti ng ma- employees, 90, 237, 239
chines), 91-93, 98-103, 105, 163, meeting s, 118-119
245, 346n11. See also Tabulating Engineering Research Asso cia te s
machines (ERA), 106, 141-142. 173-174. 179
East Fishkill, IBM fac ili ties at, 289- computers, 141-142, 173-174, 187
290 English Electric Company, 261
Copyrighted Material
398 Index

Engstrom, Howard T., 105, 106 General Products Division, 264-265,

ENIAC (Electronic Numerical Inte­ 270, 295-296,302, 387n40
grator and Computer),1 34 General Systems Division, 312,313
design, 83, 87, 133-135,1 84, Gerstner,Louis V. , 328
353nll, 353n12 Goldstine,Herman H., 137
impact, 156,200 Gomory,Ralph E.,377n58, 378n60
operation,1 3 1, 134-135, 184, Goss, Harry W., 42
353n12, 365n3 Great Depression,53-57,247
Enigma,1 04 Groves, Leslie R.,1 40,208
Essonnes, France, IBM facilities at, Guaranty Trust Company, 37
290 GUIDE, 1 87
Estridge, Philip D. , 392n33
Evans,Bob 0., 268-270, 282, 294, Haanstra, John w., 269-270, 272-
390n13 274, 279,294,295,302-304,389n5
Everett, Robert R., 211 , 217 Haddad,Jerrier A., 269,381 n4
"Ever Onward" (song), 35,337n17 Hamilton, Francis E., 124-1 25, 178-
182,343n21
Fairchild. George W., 26, 27 Hart, Frederick H. M., 90
Fairchild Semiconductor Company, Hart Manufacturing Company,39,90
281, 282, 302 Harvard University, 73-76, 122-123,
Fairclough, John w., 383n21 133, 138, 184, 269
Fellow program,376n48 Havens, Byron L.,1 59- 16 1
Ferranti Limited, 261 , 306 Haynes,Monro K., 209-210
Ferrite cores. See Magnetic-core Hayward, John, 118
memory Herrick, Harlan L., 194-195
Flint,Charles R.,23-27 High level languages. 190-197
Flip-flop circuit,83 Hiring
Floating-point notation, 1 90,272 academic contacts, 76, 1 29
Floppy disks, 228 post-World War II, 1 1 8-119, 122,
Ford, Eugene A., 27, 39-40,45-46, 123, 163, 223, 238-239
55,78,325, 339n24 pre-World War II,39-42, 64-65,
Ford Motor Company, 190, 341n24 70,95, 112-114
Forrestal,James V 106
.• Hitachi Limited, 31 1
Forrester, Jay w., 200-21 2, 298, Hobart Manufacturing Company,28
369n3 Hollerith, Herman,1 -23, 27
FORTRAN, 126,1 94-197. 368n35 CTR relationship. 27, 38-39
Frame,James H.. 387n40 domestic business, 11-1 6,21-23,
FS (Future System), 307-31 1 ,31 2, 378n2
31 7,319,320,391n20 early years, 1-4
Fujitsu Limited,311 family involvement, 1 1, 1 9-21, 39,
Fuller, Frederick L., 78 333n26
Future Demands organizations, 33, foreign business, 12,1�1 8
45.109-116, 1 54.166.169,178 inventive process,4-8, 15
management style. 13,16,1 9,20-
Gang punch, 332n22 21 ,23, 27, 35-36, 246, 247,314,
Garvey. Edward J.. 386n24 321-322
General Ceramics and Steatite Cor­ patents, 1. 4,8-10, 11-12,42-43,
poration, 206, 209-210 332n22. 333n36
General Electric Company (GE) Hollerith card, 244. See also
antitrust suit,249 Punched cards
computers. 138. 1 89-190. 196. 197, Hollerith machines. 13-16. See also
307, 389n5 EAM; Tabulating machines
SAGE project. 21 9 Honeywell Corporation,196.273,
Steinmetz, 80, 82 289, 296,297
Copyrighted Material
 Index 399

Hong Kong Core House, 291 IBM 1401 computer, 265-268, 382n6
Hopper,Grace M., 193-194 IBM 1410 computer, 268
Hotham, Geoffrey A., 374n8 IBM 1440 computer, 288
Hurd, Cuthbert C., 157-158, 166, IBM 1470 computer, 274
167-168. 231-232 IBM 2302 disk storage, 288
Hursley,England,IBM laboratory IBM 231 1 disk storage drive. 288
at,271 IBM 2314 direct access storage facil-
ity, 388n50
lAS computer. See In s titute for Ad- IBM 3033 processor, 311
vanced Study IBM 3090 series, 316
IBM 40 tape-to-card punch. 94-95. IBM 3330 disk storage, 305
102-103 IBM 5100 portable computer,313
IBM 57 card-ta-tape punch. 94-95 IBM 7000 computer series, 177,268
IBM 285 numeric printing tabulator, IBM 7030 computer,235-236,
53 376n42
IBM 305 RAMAC, 226-227 IBM 7090 computer, 177, 235
IBM 350 disk storage unit, 226-227 IBM 8000 computer series, 268-269,
IBM 405 alphabetic accounting ma- 282, 382n15
chine, 52, 153, 154 IBM alphabetic tabulating models,
IBM 407 accounting machine, 121 50-51
IBM 600 multiplying punch,5 1-52, IBM AS/400 computer series, 312
349n28 IBM card. See Punched cards
IBM 601 electric multiplier, 51-52, IBM CPC (Card-Programmed Elec-
72, 81, 153 tronic Calculator), 152-155, 183,
IBM 603 electronic multiplier, 124, 190-192, 194,216, 221
131, 149, 151, 153 IBM PC (Personal Computer), 313-
IBM 604 electronic calculating 314, 3 15
punch, 124, 131, 151-152, 153, IBM Radiotype , 95-97, 348n24
183, 190-192.229 IBM RlSC System/6000, 316, 317
IBM 608 transistor calculator, 229- IBM RT PC workstation, 316
230 IBM Selectric typewriter, 121
IBM 650 computer, 178-182, 257, IBM System/3, 312
264, 364n54, 365n56, 380n31 IBM System/38, 312
IBM 700 computer series, 174-178, IBM System/360, 263-300
219 announcement, 275-277, 301
IBM 701 computer, 167-174 manufacturing, 284-286, 289-292,
customer support, 172-174, 185- 385n15, 385n16
187, 188-190 market impact, 296-300, 381n46,
development, 167-172, 269 389n53
installations, 362n27,380n30 models, 275-276, 287, 295, 298,
name, 170-172,245 303, 307, 309,312, 385n17
performance, 172, 174, 2 16, 222 software support, 292-296
IBM 702 computer, 175-177, 189- technology choices, 282-289
190, 363n38, 367n19, 367n21, IBM System/370, 297, 304-307,
380n30 319
IBM 704 computer, 187,192, 195, IBM System/390, 391n24
236, 380n30 IBM Type IV tabulator, 51
IBM 709 computer, 177, 188,235 Information processing, 245-246
IBM'805 test scoring machine, 68-71 Information Storage Systems Com-
IBM 941 auxiliary storage unit, 154 pany, 299
IBM 1301 disk storage, 288 Institute for Advanced Study (lAS),
IBM 1302 disk storage, 288 133. 137, 138, 170
IBM 1311 disk storage drive, 288 Integrated circuits. See Solid-state de-
IBM 1400 computer serie� 270, 274 vices
vopyrighted Material
400 Index

International Computers and Tabula­ Machines Bull of France, 260, 267,

tors Limited, 261,267, 381n46 297
International Computers Limited, McPherson.John C 113 .•

381n46 EAM, 1 15 , 118- 1 19. 120-121

International Time Recording Com­ electronics, 1 19. 132-133, 138, 153-
pany, 26, 27-28 154, 157, 161, 199
Inventory control, 93-95, 224 management style, 115, 166-167
Iowa State College, 67, 86-87, 133, relations with Watsons,112-115,
158 163-164, 166- 167, 199
Magnetic-core memory, 177, 204-
Jacquard, Joseph M.,6 207, 209-210, 222. 233, 286-287
Johnson, Reynold B., 69-7 1 76, 111,
, Magnetic storage devices. See Stor­
342n6, 37 4n8 age devices
Jumbo machine, 78, 79 Mainframe computers, 311
Management styles
Kelly, Mervin j., 282 Bryce, 80-81, 145
Kettering,Charles F.,109-110, 349n2 Cary, 318-319
Kilby, Jack S., 281, 282 Forrester,212
Kirk, Charles A., 1 47-1 48 150, 166
. Haanstra, 279 294, 302-303
,

Kondolf, Frank N.,28 Hollerith, 13, 16, 19, 2 0-21 , 23, 27,
35-36, 246, 247, 314, 321-322
Lake, Clair D., 40, 47-50, 51, 55 74-
, Learson, 291-292, 307
76, 78, 101, 122, 325, 344n12 McPherson, 115, 166- 167
LaMotte, Louis H., 160, 166, 169, Palmer, 230. 235. 238, 286
264 Watson, Jr., 199, 207-208, 230. 236,
LARC (Livermore Automatic Re­ 256, 263, 291-292, 298, 302. 321-
search Computer), 231-232, 234 322
Leach, Ruth M., 59 Watson, Sr.,35, 46, 58, 70, 73. 101.
Learson, T. Vincent, 176 107, 116, 165,321-322, 377n50,
CEO, 318, 392n2 380n36
director of EDPM, 175-177, 189, Manchester University, 306
231 Manufacturing
management style, 291-292. 307 computers,219, 292
sales, 175, 177, 291-292 EAM, 60-62
System/360, 264, 267-269, 282- electronics,208-209, 280, 282, 284-
284. 291-292, 307 286, 289-291,·376n38, 385 n16
Leasing companies, 300 310, 389n53
, World War I I, 90 ,92. 105, 117,
Lemmon,Walter S., 95, 1 18 , 347n20 346n12
Leo Computer Limited, 261 Mark I computer,74-76, 101, 122-
Liberator.273 123, 124, 126, 159, 184, 193
Logue, Joseph C., 375n23 Marketing, 156-158
Los Alamos Scientific Laboratory, Market share, 250-251 , 252,311-
172, 190-192,233,236 312, 320, 356n36
Lovelace, Ada, 184 Massachusetts Institute of Technol­
Lowe, William C., 392n33 ogy (MIT), 133, 138, 199-203, 307
Luhn, Hans P 226
.• Digital Computer Laboratory, 203
Lyons,j., Company, 261 Lincoln Laboratory, 203, 209, 210,
212-213 2 16, 219, 370n13
,

MacArthur, Douglas, 208 Radiation Laboratory, 129, 159

McDowell. W. Wallace, 164-165. Servomechanisms Laboratory,200
223, 229, 233. 237-239. 279- Mauchly,John w., 87, 133-141, 156,
280 184
Machine language,193 Meader, Ralph I., 106
Copyrighted Material
 Index 401

Memorex Corporation, 299 North, Simon N. D., 21-23, 37,

Memory 334n4
early devices,135, 137, 177,204, Northrop Aircraft,Inc., 153-154
222 North Street Laboratory, 55,56. See
magnetic<ore,177, 204-207, 209- also Endicott, IBM facilities at
210,222,233, 287 Noyce, Robert N., 281, 282
magnetic drum,178-182 Nutt, Roy, 195
semiconductor,305, 306
use of,136, 295, 305-307, 316,330 Object-oriented addressing, 309
Methods research,62-65,112 Office of Naval Research (ONR),239
Military Products Division, 167, 213, Office of Scientific Research and De-
263-264 velopment (OSRD), 132
Miller,H. W ismer, 264 Office Products Division, 312-313.
Minicomputers,298 See also Electric Typewriter Division
Minneapolis-Honeywell Regulator Olsen, Kenneth H., 298, 371n28
Company, 257 Opel, John R.,308,309
Mobile machine record units, 92- Operating systems, 188
93, 347n13 Origins of IBM, 19-28
Monolithics. See Solid-state devices OS/360, 294-295
Monroe Calculating Machine Com­
pany,72-73 Paging,306
Moore School. See University of Palmer, Ralph L., 151, 176
Pennsylvania computer developments, 169-172,
Moss, Larry M.,383n30 230-233
Motorola Corporation, 302 electronics pioneering. 82-86, 123,
Multiprogramming,defined,293 150-152, 279-280, 281-282
Munitions Manufacturing Corpora- management positions, 82, 150,
tion, 90, 345n3 167, 229, 238-239
management style, 230, 235, 238,
National Bureau of Standards, 132- 286
133, 138, 139,225 and World War II, 103, 105-107
National Cash Register Company Pantograph punch,9,10
(NCR) Papian, William N., 370n21,371n28
computers, 296, 297 Parker, John E., 106,142
electronics pioneering, 81-82, 105- Patent organizations, 80, 115-116,
106, 138, 196,325 250
impact on Watson, Sr.,29-31, 32- Patent policies
34, 43, 109-110, 248, 250-251 (;fR, 42-44,249-250
revenues, 53,250-251,252, 296 IBM, 226,252-253,260, 290,
National security,132 384n25
National Security Agency (NSA), University of Pennsylvania, 138-
141, 232 139
Naval Computing Machine Labora­ Patents,CTR, 25-26,42-44, 77-78,
tory, 105-106 249-250, 325
Navy Bureau of Aeronautics, 200 Patents, Hollerith,1,4,8-10, 11-12
New York City, CTR facilities at, 39- Patents,IBM
40, 45,47, 50 Bryce, 77-81, 86
New York City, IBM facilities at, 187, computer programs, 226
190 foreign, 260
Nichol, Frederick w., 91 manufacturing, 290, 386n25
NORC (Naval Ordnance Research Mark I (ASCC),76
Calculator), 161, 231 SSEC, 126, 136
Norris, W illiam C., 106 stored-program computers, 136
Copyrighted Material
402 Index

Patterson, John H., 29,32-34,109, revenues from, 60, 254

110 small cards, 312,380n37
Peirce, J. Royd en, 42, 43-44,47-51, Punched-card mach ines. See Ta bu lat­

78, 250, 325 ing machines

Peripheral products,299-300
Personal computers, 313-314, 315 Quality control, 60-62, 290
Phelps, Byron E., 83-84, 103, 123,
124 Rabi, I. I., 129, 159
Philco Corporation, 196 Radin, George, 390n16
Phillips, J. George, 148, 160 Radio Corporation of America
Phillpotts, Raleigh,17 (RCA)
Piore, Emanuel R., 239-242,280, 282 antitrust suit, 255-256
Pitrat, Ju liu s E., 25 COBOL, 196-197
Planar processing,281 computers, 137, 257, 281, 296-297,
Plug board s,38-39, 62-63,101,134- 388n44
135, 183, 184 electronics, 81-82, 257, 281
Plug-compatible m anu facturers, 299- Radiotype See IBM Radiotype
.

300, 310-311,388n50, 388n51 RAMAC. See IBM 305 RAMAC

Pointers, 63, 183,341n18 Rand,James H., Jr., 208
Porter, Robert P., 12-13, 16-18 Rand Corporation, 187,216
Poughkeepsie, IBM fa cilities at Raytheon Manufacturing Company,
customer support, 187, 188 208, 257
electronics, 150, 230, 237, 264, 265, Relays, 101, 126, 355n31
267 Remington Rand, In c . See also
research, 238-240, 377n58 Sperry Rand Corporation
SAGE, 210 an titrust suit, 247, 260
System/360,268, 274 computers, 139-141,173-174, 196-
typewriters, 121 197, 208-209,230-231, 232
and World War II, 90 customer support, 51, 189-190
Powers, James, 37-39,78 m arket share, 89, 131-132, 250,
Powers Accounting Machine Com- 252, 256-257
pan� 38, 42-43, 48, 249-250 organization, 48, 141, 173, 208-
Powers-Samas Company, 259-260 209, 233, 259-260
Pratt and Whitney Company, 13,39 punched cards,47-48, 50
Presidents of IBM, 327-328 tabulating machines, 50-51, 378n9
Printers Remington Typewriter Company, 31
computer, 121,265,305 Research
punched-card equipment,38,40, academic, 67-76
47,119, 120-121 electronics,81-86, 119,121-126
Selectric typewriter,121, 351nlO in Endicott, 60,65, 82-86
Processor, defined, 271 magnetic storage, 82
Programming, 183-197,216-217, management of, 222-224,237-242,
292-296 378n60
Progr a m ming Research Depart ment , in New York City, 45, 62-63, 64, 81-
175-177 82
Project Grind, 211 printers, 120-121
Project Stretch, 232-237, 269 i n Poughkeepsie, 238-240,377n58
Punched cards at Watson Research Center, 240-
early concepts, 3, 5-8,15, 47-48, 242, 309, 377n58
49, 332n15 Research and Development Board,
IBM card, 48-52, 93-95, 244-245, 280-281. See also Corporate Tech-
252-253 nical Committee
manufacture of, 47, 60,254 Research and development expendi -

Remington Rand, 47-48 �0 t].lre,s ' 164-165 237

,

vopynghted Marena,
.
 Index 403

Research Division, 240-242 Solid-state devices, 228-230

Revenues, of IBM, 321, 323-325 integrated circuits, 281-283, 301 -
computers, 219,326 304
electric typewriters, 121 monolithics. 301-304
mili tary, 90, 1 17 131, 326, 345n5
, semiconductors. 268,280. 304. 305,
punched cards, 60, 254 306
Reynolds, Carl H., 387n34 transistors. 222, 228-230. 234. 235
RISC architecture, 314-316 Songs of IBM. 35, 58, 337n17
Rochester, Nathaniel, 187 Sorters, 38, 155-156
Rochester, IBM facilities at, 264, 312 SOS (SHARE Operating System),
188
Saarinen, Eero, 240, 241 Spectra. 70, 296-297
SABRE, 219, 293 Speedcoding, 190-193
SAGE, 199-219, 298,372n42 Sperry Rand Corporation. See also
Salesmen, 16, 29, 31-33, 62, 174- Remington Rand, Inc.
175,222,248 computers, 136, 1 96,256-257, 266-
Samuel, Arthur L., 374n 21, 377n58 267,289
San Jose, IBM facilities at, 223-224, magnetic tap e 289
,

227. 228, 264, 373n3 market share. 256-257. 267, 296,

Scientific Data Systems, 299 297
Seagate Technology Corporation. organization, 233. 256
300 SPREAD Task Group, 270-273, 277,
Seaton, Charles w., 332nl0 284, 293, 382n16, 383n24, 384n34
Seeber, Robert R 124-126
.• Sputnik, 235
Selectric typ ewr it er See IBM Selec-
. SSEC (Selective Sequence Electronic
tric typewriter Calculator), 124-127, 136, 159,
Semiconductors. See Solid-state de- 18 4 190, 3 52n 20
,

vices Steinmetz, Charles P., 80 82

,

Service, 51,172-173, 246-247. See Stibitz,George R., 355n31

also Customer support Storage devices. See also Memory;
Service Bureau Corporation, 255, Punched cards
264,298, 392n 2 magnetic disks, 222, 223-228, 287 -
Seven Dwarfs, 296-297 288, 308, 3 16, 330, 374n8
SHARE, 187-188,366nll magnetic drums, 82, 105-106, 178-
Shockley, William, 228 182
S hu gart, Alan F., 299-300 magnetic tape, 82, 156, 175, 221-
Signal Corps, 96, 98, 101-102 222, 288-289
Simplex Time Recorder Company, 28 paper tape, 4-6, 8, 74, 94-95, 1 78
Simulation of computers, 236, 27 3, Storage Technology Corporation,
274, 376n43 299,388n51
Singer Sewing Machine Company, Stored-program computers, 135-
336n3 142,185,221, 354n17, 354n18
Single-level store, 309, 390n16 Stretch. See Project Stretch
Slade, Bernard N., 386n24 Subroutines, 184-186
SLT (Solid Logic Technology), 268, Supercomputers, 73-76, 124-127.
273-274,282-286, 301, 305, 1 58-161, 230-237, 298, 376n42,
385n13 388n47
SMS (Standard Modular System), Sylvania Electric Products Company,
234-235, 281-282, 305, 376n38 235
Social Security Administration, 57, System/360. See IBM System/ 360
63-64,89,317 System Development Corporation,
Software support, 292-296, 387n 31 . 216, 2 18-2 19 , 372n48
See also Customer support; Pro- Systems Development Division, 302,
gramming 313
Copyrighted Material
404 Index

System service women,58-60 University of California Radiation

Systems Manufacturing Division, 302 Laboratory,231
University of Pennsylvania, 87, 133,
Tabulating Machine Company,16, 138-139
25,26-28,34,38 USE (Univac Scientific Exchange),
Tabulating machines,244-245. See 187
also EAM; Hollerith machines Utility programs,185-186
alphabetic,49, 50-51,53-54
early devices,7-10, 38-39, 40 Vacuum tubes. See also Electronics
electric versus mechanical, 6, 38, pioneering
43,62-63, 78-80, 253 Atanasoff computer, 86-87
Jumbo machine,78, 79 Atlas computer, 141
printing,38, 40,47, 53,119, 120- Colossus,353n11
121 EDVAC, 354n19
wartime use,40, 92-95,98-103 Endicott circuits, 83-86
Tabulator Limited, 17. See also Brit- ENIAC, 83,134-135, 353n11
ish Tabulating Machine Company IBM 603 and 604, 152
Taft-Peirce Company, 39 IBM versus Remington Rand, 209
Talcott, Edmund, 20-21 SAGE,215
Talcott, Lucia,19-20 SSEC, 126, 134,353nll
Tape Processing Machine. See T PM Valley,George E.,202-203
Teleprocessing,93-94 Virtual memory, 305-307
Telex Corporation,299 Von Neumann, John,135-138, 170,
Teller, Edward, 231-232 362n25,365n3
Test scoring machines, 67-71
Texas Instruments Company (TI), Walker, Francis A., 3-4
280, 282, 283,302, 384n3 Watson, Arthur K. (Dick), 148,264,
Thomas, Llewellyn H., 127 291-292, 302, 318
Time Equipment Division, 335n21, Watson, T homas j.,Jr.,145-150,
381n4 163, 263,291-292,318,392nl
Time Recording Division, 40 early computers,155,167-172,175-
T PM (Tape Processing Machine), 182,230-232,236-237
168-169,174-175 ea rly training,145-148
Transistors. See Solid-state devices electronics, 149-150, 155, 163
Tripp, C. A., 43-44, 250 engineering, 163-167, 237-238
Tub file, 224 management positions, 163, 253,
Tucker, Stuart G., 383n30 318
Turing, Alan M., 184-185 management styl e, 199, 207-208,
230,236, 256, 263, 291-292, 298,
Underwood, Francis 0., 382n6 302, 321-322
Union Col lege, 80, 82, 83 System/360,275, 279, 282, 291-
Union of Soviet Socialist Republics, 292, 301, 302, 318
199,218, 235 team members, 163-167,175,237-
United Aircraft Corpora tion,194 238
UNIVAC (UNIVersal Automatic Com- Watson, Thomas j., S r 29-36, 243,
.,

puter) 244, 337n16

customer support, 187, 189-190 early computers, 73-76, 124-127,
development, 139, 369n9 158-161, 178,182
impact,155-156, 174-175, 230- early years, 31-32
231 electric typewriters, 96, 119, 150,
installations,356n34 348n23
performance, 175, 185, 222,289, electronics, 149-150
363n29 engineering,37, 39-42, 45-47, 109,
Univac File Computer, 22 e
opyrighfed McM!ncltl8-121, 178,182,226
 Index 405

general manager of CTR, 28, 34-35

grooming two sons, 145-148, 172,
377n50
legacy, 243-261, 378nl
management style,35, 46, 58, 70,
73, 101, 107, 116, 165,321-322,
377n50,380n36
NCR experience, 29-31, 32-34, 43.
109-110, 248,250-251
patent policy,42-44, 249-250
research support, 67-69,70-71, 73,
76,95,121-129
sales, 31-33,142-143,247-248
Watson Laboratory, 127-129, 157,
159-161, 169,239
Watson Research Center, 240-242,
308-309, 377n58
Weapon System Evaluation Group
(WSEG), 202
Wesley,Felix j.,62-64, 91
Western Electric Company,13,39
Wheeler, Earl F., 387n40
Whirlwind computer, 199-207. 210-
212
Wilkes,Maurice V., 185, 271. 355n22
Williams, Albert L., 164, 207, 264,
292
Williams tube, 168, 177
Wilson, Lawrence A., 391n26
Winchester file, 228
Women employees, 58-60
Wood, Benj amin D., 67-69
World Headquarters. See Corporate
headquarters
World Trade Corporation,148. 258-
259,261,264. 380n32,380n37
World War II.89-107,114-115,117,
121, 131. 345n5. 346n6, 346nl2

Yourke, Hannon S., 234

Ziller,Irving, 194-195
Zurich, IBM laboratory at, 239
Zuse, Konrad. 343n19

Copyrighted Material