ArticlePDF Available

The Development of Social Network Analysis

Authors:
THE DEVELOPMENT OF SOCIAL
NETWORK ANALYSIS
A STUDY IN THE SOCIOLOGY OF SCIENCE
Linton C. Freeman
ΣP
Empirical Press
Vancouver, BC Canada
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
© 2004 Linton C. Freeman
All rights reserved. Published in 2004
Printed in the United States of America
ΣP Empirical Press, Vancouver, BC Canada
Cover and book design by Rosemary Boyd, Document Design
Laguna Beach, CA
Library of Congress Control Number: 2004111710
Publisher: BookSurge, LLC
North Charleston, South Carolina
ISBN 1-59457-714-5
“We all connect, like a net we cannot see.”
—Mickenberg and Dugan,
Taxi Driver Wisdom, 1995
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Jacob L. Moreno
Harrison C. White
This work is dedicated to Jacob L. Moreno and Harrison White.
Without the mammoth contributions of these two scholar-
teachers there would be no field of social network analysis.
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Contents
List of Illustrations........................................................................ ix
Preface ............................................................................................ xi
Chapter 1: Introduction ................................................................ 1
Chapter 2: Prehistory: The Origins of Social Network Ideas
and Practices........................................................................... 10
Chapter 3: The Birth of Social Network Analysis I:
Sociometry .............................................................................. 31
Chapter 4: The Birth of Social Network Analysis II:
The First Harvard Thrust ..................................................... 43
Chapter 5: Social Network Analysis During
the Dark Ages I: the 1940s ...................................................65
Chapter 6: Social Network Analysis During
the Dark Ages II: the 1950s.................................................. 82
Chapter 7: Social Network Analysis During
the Dark Ages III: the 1960s .............................................. 113
Chapter 8: The Renaissance at Harvard ................................121
Chapter 9: Getting Organized .................................................. 129
Chapter 10: A Summary and Some Surprises ....................... 159
Acknowledgments ..................................................................... 169
References ................................................................................... 173
Index ............................................................................................ 195
vii
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
ix
Figures
Figure 1.1. Association between number of substantive areas specified
and year of publication of social network research................................5
Figure 2.1. Auguste Comte honored on a French postage stamp............... 11
Figure 2.2. Plaque at Comte’s Paris residence ...........................................12
Figure 2.3. Georg Simmel...........................................................................15
Figure 2.4. Hobson’s two mode data ..........................................................19
Figure 2.5. Morgan’s descent system of ancient Rome .............................. 22
Figure 2.6. Macfarlane’s images of two-step marriage prohibitions .......... 23
Figure 2.7. Hobson’s hypergraph ...............................................................24
Figure 2.8. Sir Francis Galton, 1909, with a young Karl Pearson ............25
Figure 3.1. Sociogram of a football team ....................................................38
Figure 4.1. W. Lloyd Warner, circa 1920s..................................................44
Figure 4.2. An idealized model of overlapping “cliques” ..........................46
Figure 4.3. Friendships in the bank wiring room.......................................50
Figure 4.4. Reported “cliques” in the bank wiring room ...........................50
Figure 4.5. Allison Davis honored on a U.S. postage stamp ..................... 51
Figure 4.6. St. Clair Drake ........................................................................52
Figure 4.7. Attendance of 18 women at 14 social events ...........................52
Figure 4.8. “Clique” memberships of males ...............................................53
Figure 4.9. George Caspar Homans ...........................................................56
Figure 4.10. Whyte’s image of the Nortons ............................................... 59
Figure 5.1. Kurt Lewin (before he immigrated to America).......................66
Figure 5.2. Alex Bavelas.............................................................................69
Figure 5.3. Experimental communication forms studied by the Bavelas
group ...................................................................................................70
Figure 5.4. The Festinger graph .................................................................72
Figure 5.5. Claude Lévi-Strauss.................................................................78
Figure 5.6. Directed graphs from Lévi-Strauss..........................................80
Figure 6.1. The Rashevsky seminar............................................................87
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
x
Figure 6.2. Paul Lazarsfeld ........................................................................ 91
Figure 6.3. Robert Merton..........................................................................91
Figure 6.4. Everett M. Rogers .................................................................. 100
Figure 6.5. Alfred Reginald Radcliffe-Brown...........................................101
Figure 6.6. Max Gluckman with John Barnes ......................................... 104
Figure 6.7. Karl Deutsch ..........................................................................106
Figure 6.8. Ithiel de Sola Pool .................................................................. 107
Figure 7.1. Edward O. Laumann ............................................................. 115
Figure 9.1. Influences on some founders of social network analysis ........131
Figure 9.2. Social proximities at the start of the EIES research ...............133
Figure 9.3. Direction of positional change during the EIES research ......133
Figure 9.4. Final locations of individuals in the EIES research...............134
Figure 9.5. Migrations linking social network research centers ..............137
Figure 9.6. Participants in the 1975 Dartmouth Conference .................. 144
Figure 9.7. Participants in the 1979 East-West Center Conference ........146
Figure 9.8. Barry Wellman.......................................................................147
Figure 9.9. The Sun Belt and European Meetings...................................154
Figure 10.1. Citation patterns in the Small World literature ..................166
Figure 10.2. Number of social network articles listed in Sociological
Abstracts by year ............................................................................. 167
xi
Preface
This book was produced with a great deal of help from a
lot of people. Many of the early practitioners of the social net-
work approach allowed themselves to be interviewed—either
face-to-face, or by telephone, email or regular old-fashioned let-
ter mail. These include Peter Abell, John Barnes, Russ Bernard,
Peter Blau, Liz Bott, John Boyd, Kathleen Carley, Doc Cartwright,
Jim Davis, Alain Degenne, Pat Doreian, Tom Fararo, Claude Fla-
ment, Michel Forsé, Ove Frank, Sue Freeman, Bill Garrison, Tor-
sten Hägerstrand, Maureen Hallinan, Muriel Hammer, Frank
Harary, Paul Holland, Charles Kadushin, Larry Kincaid, Al
Klovdahl, Ed Laumann, Hal Leavitt, Sam Leinhardt, Joel Levine,
Nan Lin, Larissa Lomnitz, Duncan Luce, Bob Merton, Clyde
Mitchell, Rob Mokken, Woody Pitts, Charles Proctor, Anatol
Rapoport, Ev Rogers, Arnold Simel, Tom Snijders, Frans Stok-
man, Morry Sunshine, Chuck Tilley, Barry Wellman, Doug White,
Harrison White, Al Wolfe and Rolf Ziegler. A good many of these
people went to great pains to produce detailed accounts of who
influenced them to think in the structural perspective of social
network analysis. These individuals provided most of the in-
formation that has gone into the present historical account.
In addition, I received useful feedback from a number of
people after I presented a preliminary version of this historical
review at the International Sunbelt Social Network Conference
in Vancouver, BC, on April 12, 2000. The detailed written sug-
gestions provided by Davor Jedlicka, Peter Marsden, Frans Stok-
xii
man and Barry Wellman were particularly helpful. And a good
many colleagues have read all or parts of this manuscript and
provided useful feedback. These include John Barnes, Mike
Brower, Katie Faust, Sue Freeman, Irving Louis Horowitz, Do-
minic Jesse, Charles Kadushin, Bob Merton, Cal Morrill, Woody
Pitts, Ev Rogers, M. de Lourdes Sosa, Cindy Webster, Doug
White, Robin Williams and, most of all, Morry Sunshine, who
has worked harder than anyone could reasonably expect.
Beyond this relatively recent input, I learned a great deal
from earlier discussions I had with others who were involved in
the development of social network analysis but who are no longer
available to be interviewed. Among these, my conversations with
Gregory Bateson, Jim Coleman, St. Clair Drake, Fred Kochen,
Paul Lazarsfeld, Jiri Nehnevajsa, Ithiel de Sola Pool, Tom Sch-
weizer and Bill Whyte are particularly memorable. Each had an
impact on my thinking about the field.
I am also grateful to Rosemary Boyd who designed and
edited this book. Without her wise counsel this would be a far
less polished essay.
In an important way, all of these individuals are co-authors
of this book. All provided key information and insights. With
all this help, then, I hope that the present account is mostly ac-
curate. Any inaccuracies that remain are my own.
Linton C. Freeman, Laguna Beach, CA
Holmes Beach, FL
Chapter 1
Introduction
Writing in 1968, a Columbia University sociologist, Allen
Barton, described mainstream research in social science:
For the last thirty years, empirical social research has
been dominated by the sample survey. But as usually
practiced, using random sampling of individuals, the
survey is a sociological meatgrinder, tearing the indi-
vidual from his social context and guaranteeing that
nobody in the study interacts with anyone else in it.
It is a little like a biologist putting his experimental
animals through a hamburger machine and looking
at every hundredth cell through a microscope; anato-
my and physiology get lost, structure and function
disappear, and one is left with cell biology.…If our
aim is to understand people’s behavior rather than
simply to record it, we want to know about primary
groups, neighborhoods, organizations, social circles,
and communities; about interaction, communication,
role expectations, and social control.
Barton’s statement was true when he made it, and it is still
true today. Mainstream social research was and is focused ex-
clusively on the behavior of individuals. It neglects the social
part of behavior; the part that is concerned with the ways indi-
viduals interact and the influence they have on one another.
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Fortunately for those of us who don’t want to grind up the
social world into hamburger, there is, and there has always been,
an alternative. Some social research has consistently focused on
the social relationships linking individuals rather than on the
individuals themselves. The kind of research that examines the
links among the objects of study is called structural.
This kind of structural approach is not confined to the study
of human social relationships. It is present in almost every field
of science. Astrophysicists, for example, study the gravitational
attraction of each planet in the solar system on each of the oth-
ers in order to account for planetary orbits. Molecular chemists
examine how various kinds of atoms interact together to form
different kinds of molecules. Electrical engineers observe how
the interactions of various electronic components—like capaci-
tors and resistors—influence the flow of current through a cir-
cuit. And biologists study the ways in which each of the species
in an ecosystem interacts with and impinges on each of the oth-
ers.1
In social science, the structural approach that is based on
the study of interaction among social actors is called social net-
work analysis. The relationships that social network analysts
study are usually those that link individual human beings. But
important social relationships may link social individuals that
are not human, like ants or bees or deer or giraffes or apes. Or
they may link actors that are not individuals at all. Network an-
alysts often examine links among groups or organizations—even
among nation-states or international alliances.
The social network approach is grounded in the intuitive
notion that the patterning of social ties in which actors are em-
bedded has important consequences for those actors. Network
analysts, then, seek to uncover various kinds of patterns. And
they try to determine the conditions under which those patterns
arise and to discover their consequences.
The idea that the patterning of social ties is worth examin-
ing is probably very old. For example, descent lists are stressed
1 As early as 1925, for example, the philosopher-mathematician Alfred North Whitehead
(1925, p. 297) suggested that, “A forest is the triumph of the organisation of mutually
dependent species.”
2
both in the Bible and in the education of Hawaiian nobles, who
were required to memorize dozens of generations of their fore-
bears.
Before modern social network analysis emerged, investi-
gators used one or some combination of four approaches in con-
ducting structural research on social phenomena. Some clarified
and extended the basic structural intuition. Some collected the
kind of actor-by-actor data that permits the systematic examina-
tion of social patterning. Some developed procedures for con-
structing visual images of patterns of ties. And others worked
on computation or spelled out the mathematical properties of
social patterns.
But it is only recently that these approaches have all been
integrated into an organized paradigm for research. All four of
these features are found in modern social network analysis, and
together they define the field:
1. Social network analysis is motivated by a structural in-
tuition based on ties linking social actors,
2. It is grounded in systematic empirical data,
3. It draws heavily on graphic imagery, and
4. It relies on the use of mathematical and/or computation-
al models.
Beyond commitment to these four features, however, mod-
ern social network analysts also recognize that a wide range of
empirical phenomena can be explored in terms of their structur-
al patterning. But coming to that recognition has been a diffi-
cult process. In an interview in 1996, the anthropologist Clyde
Mitchell described his own experience when he personally came
to see the generality and broad utility of the network analytic
approach. In the 1950s he had been a participant in Max Gluck-
man’s seminar at the University of Manchester. There, several of
the participants had used a structural approach to guide their
research. But, at that point, Mitchell failed to see the generality
of that approach; he viewed those research reports as interest-
ing but he failed to see their underlying identity.
Introduction
3
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Then, a few years later, in the early 1960s, Mitchell was di-
recting the research of several ethnographic fieldworkers in Zim-
babwe (Mitchell, 1969). They included David Boswell, Peter
Harries-Jones, Bruce Kapferer and Pru Wheeldon. Boswell was
collecting data on personal crisis and social support. Harries-
Jones was concerned with the importance of tribalism in politi-
cal organization. Kapferer was studying a labor conflict in a
mining company. And Wheeldon was examining the emergence
of political processes in an inter-ethnic community. On the face
of it, these were simply four unrelated research projects.
But at that point, Mitchell had a breakthrough. He began to
see that, as structural studies, all four projects shared a common
core. As he put it:
…it was then that I realized that we needed a formal
method of conducting the analyses. I had been read-
ing the journal Sociometry, so I knew something
about those procedures, but of course I knew very lit-
tle. And it was then that Doc Cartwright and Frank
Harary’s book (Harary, Norman and Cartwright,
1965) appeared that I learned to get on to it.
Once the social network research community began to rec-
ognize the generality of its structural approach, useful applica-
tions in a very wide range of empirical situations became
possible. As I indicated in a recent piece in the Encyclopedia of
Psychology (Freeman, 2000a), the network field has developed
important applications in research on:
…the study of occupational mobility, the impact of
urbanization on individuals, the world political and
economic system, community decision-making, social
support, community, group problem-solving, diffu-
sion, corporate interlocking, belief systems, social cog-
nition, markets, sociology of science, exchange and
power, consensus and social influence, and coalition
formation…primate studies, computer-mediated com-
munication, intra- and inter-organizational studies,
and marketing…health and illness, particularly AIDS.
4
The ever-widening range of applications of social network
analysis is demonstrated in a study by Otte and Rousseau (2002).
They examined network analytic articles published between 1984
and 1999. Figure 1.1 shows that, year-by-year, there has been an
almost linear growth in the number of substantive areas in which
the social network approach has been applied.
In effect, then, social network analysis is defined not only
by the four features listed above, but also by an extremely wide
and ever-growing range of applications. Its practitioners increas-
ingly share recognition of the broad generality of its structural
paradigm.
Because of this generality, network analysis cuts across the
boundaries of traditional disciplines. It brings together sociolo-
gists, anthropologists, mathematicians, economists, political sci-
entists, psychologists, communication scientists, statisticians,
ethologists, epidemiologists, computer scientists, both organi-
zational behavior and market specialists from business schools
and recently, physicists. These people come from vastly differ-
Introduction
Figure 1.1. Association between number of substantive areas
specified and year of publication of social network research
5
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
ent backgrounds but they all share a common commitment to
the structural perspective embodied in the network approach.
This shared commitment is demonstrated by the systemat-
ic accumulation of knowledge within the field. Social network
analysis is one of the few social science endeavors in which peo-
ple influence one another in such a way that they all work to-
gether to build a cumulative body of knowledge. Indeed, after
examining its citation practices, Hummon and Carley (1993) went
so far as to declare network analysis a “normal” science in the
sense described by Thomas Kuhn (1962). Kuhn defined a sci-
ence as normal if and only if it displays a systematic approach
that both generates puzzles and solves them. Since a normal sci-
ence is the product of an ordered sequence of discoveries, it is
cumulative.
Social network analysts now have an international organi-
zation, the International Network for Social Network Analysis, or
INSNA.2 INSNA holds an annual meeting in Europe or in North
America and it is involved in the publication of three profes-
sional journals, Social Networks, Connections and the electronic
Journal of Social Structure. Beginning in the early 1980s with books
by Steven D. Berkowitz (1982) and David Knoke and James H.
Kuklinski (1982), writers began to produce standard texts on
social network analysis. Texts now have been written in several
languages including English, French, German, Italian, Finnish
and Swedish. An up-to-date list is available on an INSNA web
page:
http://www.sfu.ca/~insna/INSNA/books_inf.html
There are also a large number of computer programs de-
signed specifically for the analysis and display of social networks.
See the INSNA web page:
http://www.sfu.ca/~insna/INSNA/soft_inf.html
2A Spanish-language affiliate of INSNA, REDES, was formed in the late 1990s. It
organizes meetings and publishes a social network journal. REDES also has a web
page: http://www.redes-sociales.net/
6
In addition, a number of centers for network training and
research have sprung up throughout the world. They are listed
on an INSNA web page:
http://www.sfu.ca/~insna/INSNA/schools_inf.html
All of this growth leaves us with some interesting ques-
tions. Beyond the very old intuitive idea of the importance of
social ties, where did this interest and activity begin? What are
the intellectual foundations of the field? How and when were
various kinds of structural studies brought together into a co-
herent program of research? And how did that research get or-
ganized into a recognized collective enterprise that includes
organizations, meetings, books and journals and the like?
So far, no one has produced a comprehensive history of the
origins, development and emergence of this field. There are a
few brief historical notes in print, but each of them tends to fo-
cus on only a part of the whole picture. Some (Leinhardt, 1977;
Berkowitz, 1982; Marsden and Lin, 1982; Freeman, 1989; Scott,
1992; Degenne and Forsé, 1994; Wasserman and Faust, 1994) were
concerned with the question of the origins of social network
analysis. These writers have sought to uncover how and when
the key ideas and procedures of the field were developed. Oth-
ers (Mullins and Mullins, 1973; Wolfe, 1978; Alba, 1982; Mars-
den and Lin, 1982) were primarily interested in questions of
institution building. They tried to figure out how and when the
intellectual ideas were organized into an established field of re-
search.
Some of the writers who have examined the question of the
origins of social network analysis have proposed that it all be-
gan in the early 1930s with the work of Jacob Moreno (Hummon
and Carley, 1993; Leinhardt, 1977; Marsden and Lin, 1982; Free-
man, 1989; Degenne and Forsé, 1994; Wasserman and Faust,
1994). The year in question is 1934 when Moreno’s introduction
to sociometry, Who Shall Survive?, was published. Clearly, that
publication was a signal event in the history of social network
analysis. It was a turning point for the development of the field.
But other writers have argued that social network analysis
did not begin until the early 1970s when Harrison White began
Introduction
7
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
training graduate students at Harvard (Mullins and Mullins,
1973; Berkowitz, 1982; Scott, 1992). During that era, White, along
with his students, produced an amazing number of important
contributions to social network theory and research. Contem-
porary network analysis could never have emerged without
those contributions.
Both of these events do turn out to have produced points of
critical transition in the history of the field. Both involved con-
tributions that provided turning points for the development of
social network research. I treat these two periods of critical tran-
sition in great detail. But there is much more, as well. In this
book I also highlight a great deal of important work that took
place before the 1930s, and crucial developments that occurred
between the 1940s and 1960s. Many of these are fascinating ex-
amples of the development of science, and they also underline
the importance of social ties and social gaps in the development
of a scientific specialty.
In this book I will examine contributions in four distinct
eras: (1) the period ending in the late 1920s, (2) the 1930s, (3) the
period from 1940 to the end of the 1960s, and (4) the period from
the early 1970s to the present.
Chapter 2 will examine the contributions of the people who
worked before the 1930s. These are the people who introduced
key ideas and practices that anticipated current social network
research.
Chapters 3 and 4 will review the birth of modern social
network analysis. Two, probably independent, developments all
started in the late 1920s and ran through the 1930s. One of these,
examined in Chapter 3, was the introduction of sociometry by
Jacob L. Moreno and his associates. A second, outlined in Chap-
ter 4, took place at Harvard and was led by W. Lloyd Warner.
Chapters 5, 6 and 7 will cover the 1940s, 1950s and 1960s
respectively. In those chapters, I will review social network re-
search that took place during the “dark ages” between the con-
tributions of the 1930s and those of the 1970s. And then Chapter
8 will return to Harvard and examine the transformations that
Harrison White and his students produced in the early 1970s.
8
Chapter 9 will take up the question of institution building.
I will examine what happened to the field after the 1970s. There
the issue is how the network approach became a recognized field
of research. Finally, Chapter 10 will attempt to bring it all to-
gether and reexamine some of the surprising results that were
revealed in the earlier chapters.
As a final note, I need to stress that this book is not intend-
ed as an intellectual history of social network analysis. It is, rather,
an exploration of the field from a sociology of science perspec-
tive. I am convinced that the patterning of links among the peo-
ple who were involved in the development of the field—its social
network—is a key to understanding how and why it emerged.
So I have tried to reconstruct those patterns. Of course, I could
not avoid dealing with a good many elements of intellectual his-
tory, but my primary aim here is to uncover the social processes
that were involved in the emergence of a scientific specialty. Thus,
this is a history of social network analysis written from a social
network perspective.
Introduction
9
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Chapter 2
Prehistory: The Origins of Social
Network Ideas and Practices
In the first chapter I defined social network analysis as an
approach to social research that displays four features: a struc-
tural intuition, systematic relational data, graphic images and
mathematical or computational models. In this chapter I will
consider each of those four features in turn and try to specify its
first use in social research. My aim here is to describe the earli-
est research I have been able to find that embodied each feature.
Much of the research described here will include work that
is often cited—work done by people who are usually credited
with having paved the way for social network analysis. Other
research I will describe, however, was produced by people who
are relatively unknown in the network field, people who are not
usually recognized as having introduced an idea or practice, but
who did innovate in ways that provide precursors of current
network research practices. These are people whose contribu-
tions have not been—but perhaps should be—acknowledged.
2.1 Early Structural Intuitions
In a previous report (Freeman, 1989, p. 18), I reproduced a
descent list from the Book of Genesis. My aim there was to sug-
gest that human beings have, since the earliest days, recognized
the importance of the ties that link social actors. But that early
recognition was implicit.
10
The earliest explicit statement of a structural perspective
on social life that I have found was proposed by Isidore Auguste
Marie François Xavier Comte (usually known simply as Auguste
Comte). Comte is pictured in Figure 2.1. Although he is usually
not mentioned in reviews of the history of social network analy-
sis, I suspect that he had a large, albeit indirect, influence on the
development of the field.
Comte was born in 1798 in Montpellier in the south of
France. He was the eldest son in a middle class family. He en-
tered the lycée—the equivalent of high school—at the age of nine,
and studied mathematics with Daniel Encontre. Comte was al-
ways an outstanding student, but from his earliest days in school
he also established himself as a something of a rebel.
Despite his rebellious tenden-
cies, Comte scored fourth among
300 applicants and was admitted to
France’s top technical college, the
Ecole Polytechnique, when he was
sixteen (Boiteux, 1958). There he
continued to study mathematics
and he continued to excel. But this
was a period of political unrest in
France and as a result, the whole
ecole was closed two years later in
1816. For a year Comte supported
himself by tutoring. Then in 1817,
at the age of nineteen, he met and
went to work with Henri Saint-Si-
mon.
Saint-Simon was then sixty
years old. He was an established
utopian thinker and is generally ac-
knowledged as the founder of French socialism. According to
Coser (1977, p. 15), Saint-Simon was “creative” and “fertile,” but
at the same time was a “disorderly and tumultuous man.” Comte
brought order and discipline into the partnership and the two
worked and published together for seven years. During most of
that period Comte lived in the VIth arrondissement in Paris. His
Figure 2.1. Auguste Comte
honored on a French
postage stamp
Prehistory: The Origins of Social Network Ideas and Practices
11
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
regard in France is demonstrated by the fact that his residence is
still memorialized with the marble plaque shown in Figure 2.2.
Coser (pp. 1–17) described the collaboration between Saint-
Simon and Comte:
A number of scholars have argued the question of who
benefited the most from the close collaboration, Comte
or Saint-Simon. There is no need to take sides in this
somewhat byzantine quarrel. It suffices to say that
Comte was influenced in a major way by his patron,
even though his close contact with Saint-Simon may
have brought to fruition ideas that had already germi-
nated in Comte’s mind.…The sketches and essays that
Comte wrote during the years of close association with
Saint-Simon, especially between 1819 and 1824, con-
tain the nucleus of all his later major ideas.
Figure 2.2. Plaque at Comte’s Paris residence3
3In English, this plaque reads, “Auguste Comte, founder of positivism, lived here from
1818 until 1822. Here he conceived the sociological law of three states and formulated
the system of classification of the sciences.”
12
Comte’s primary commitment was to the development of
sociology as a science. In that respect his views are quite con-
temporary, although they were published first between 1830 and
1842 (Comte, 1830–42).4 Comte coined the term sociology and he
specified its goal as uncovering the laws of society. This goal, he
argued, required both theory and systematic observation. On
the importance of theory he said (Martineau, 1853/2000, v. II,
p. 203) “No real observation of any kind of phenomena is possi-
ble, except in as far as it is first directed, and finally interpreted,
by some theory.” And about observation, he indicated that the
study of social phenomena (Martineau, 1853/2000, v. II, p. 181)
“…supposes first, that we have abandoned the region of meta-
physical idealities, to assume the ground of observed realities
by a systematic subordination of imagination to observation.”
Comte argued also for the importance of systematic com-
parative research in sociology. He (Martineau, 1853/2000, v. II,
p. 207) went so far as to suggest comparing the social structures
of human beings with those of non-human animals, “It is a very
irrational disdain which makes us object to all comparison be-
tween human society and the social state of the lower animals.”
This view is consistent with some of the latest work in social
network analysis (Faust and Skvoretz, 2002).
In his definition of sociology, Comte specified the two main
aspects of the field, statics and dynamics. Statics, he said (Mar-
tineau, 1853/2000, v. II, pp. 192, 204 and 207), is focused on the
investigation of the “laws of social interconnection” or (v. II, p.
183) “the laws of action and reaction of the different parts of the
social system.” At the most fundamental level, according to
Comte, these parts are nuclear families. He argued (Martineau,
1853/2000, v. II, p. 234) that since
…every system must be composed of elements of the
same nature with itself, the scientific spirit forbids us
to regard society as composed of individuals. The true
Prehistory: The Origins of Social Network Ideas and Practices
4In the present work I have drawn heavily on a new edition of the translation of Comte’s
Cours de philosophie positive by Harriet Martineau (1853/2000). Comte liked Martineau’s
translation of his work so much that he substituted Martineau’s two-volume version
for his own six-volume version in his list of books that he believed should survive
forever (Standley, 1981, p. 160).
13
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
social limit is certainly the family—reduced, if neces-
sary, to the elementary couple which forms its basis.
Comte (Martineau, 1853/2000, v. II, p. 187) went on to show
how the parts of the social system are interconnected: “Families
become tribes and tribes become nations.” His view, then, is strik-
ingly contemporary. He defined society using the kinds of struc-
tural terms that are found today in social network analysis.
Whether he is given credit or not, Comte was the first scholar I
could find that proposed a way of looking at society in terms of
the interconnections among social actors.
Most other prominent nineteenth and early twentieth cen-
tury sociologists embraced Comte’s structural perspective. A
common theme involved describing differences in the pattern-
ing of social connections in traditional versus modern societies.
In his study of ancient law, for example, Sir Henry Maine (1861/
1931) did just that. He suggested that in small traditional, fami-
ly-oriented societies, individual ties were governed by univer-
sal rights and obligations; these arrangements he called status.
In contrast, he saw most ties in large modern societies as ground-
ed on negotiated agreements; in his terms, they are based on
contract.
Ferdinand Tönnies (1855/1936) made a similar distinction
when he used the word gemeinschaft to characterize the tradi-
tional social form that involved personal and direct social ties
that linked individuals who shared values and beliefs. He con-
trasted that with modern forms based on gesellschaft, where so-
cial links are formal, impersonal and instrumental.
Emile Durkheim (1893/1964) described traditional societ-
ies in which solidarité mechanique linked similar individuals with
repressive regulations. This he distinguished from modern so-
cieties in which a division of labor led individuals to form coop-
erative links based on solidarité organique.
Sir Herbert Spencer (1897) in England and Charles Horton
Cooley (1909/1962) in America both described traditional small-
scale societies in which individuals were linked by intimate, pri-
mary relations. And they both contrasted those with modern,
large-scale societies where individuals are often linked by im-
personal, secondary relations.
14
These early sociologists all tried to specify the different
kinds of social ties that link individuals in different forms of
social collectivities. Thus, since they were all concerned with
social linkages, they all shared a structural perspective.
An entirely different structural perspective was developed
at about the same time by Gustave LeBon (1897/1995). LeBon
examined the phenomenon of crowd behavior. He suggested
that when individuals become members of crowds they lose their
individual identities. As members of crowds, people imitate those
around them, he said, and ideas and behaviors diffuse from per-
son to person by a process of contagion. Thus, since LeBon
focused on the flow of infor-
mation among individuals,
his concern was also structur-
al.
Perhaps the most explic-
itly structural perspective
adopted by any of the late
nineteenth and early twenti-
eth century social thinkers
was displayed in the work of
Georg Simmel (pictured in
Figure 2.3). Simmel (1908/
1971, p. 23) said, “Society
exists where a number of in-
dividuals enter into interac-
tion.” He (pp. 24–25) went on
to specify this idea:
A collection of human beings does not become a soci-
ety because each of them has an objectively determined
or subjectively impelling life-content. It becomes a
society only when the vitality of these contents attains
a form of reciprocal influence; only when one individ-
ual has an effect, immediate or mediate upon another,
is mere spatial aggregation or temporal succession
transformed into society. If, therefore, there is to be a
science whose subject matter is society and nothing
else, it must exclusively investigate these interactions,
these kinds and forms of sociation.
Prehistory: The Origins of Social Network Ideas and Practices
Figure 2.3. Georg Simmel
15
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
In these statements Simmel expressed the core belief that
underlies modern social network analysis. For Simmel, sociolo-
gy was no more and no less than the study of the patterning of
interaction. And Simmel’s student, Leopold von Wiese (Wiese
and Mueller, 1931/1941, p. 30), went even further and talked in
contemporary terms about a “system of relations” and a “net-
work of lines between men.” The views expressed by Simmel
and his students, then, were—and remain—explicit statements
of the social network perspective.
Overall, then, there were a good many early writers—
particularly in the field of sociology—who laid out the intuitive
groundwork for network analysis. Early contributions, howev-
er, were not limited to intuitive ideas. In the next section I will
show some early work that involved the systematic collection
of structural data.
2.2 Systematic Empirical Data
At the beginning of the nineteenth century, long before
Comte defined a structural approach to sociology, systematic
data on social structure were being collected by a Swiss natural-
ist, Pierre Huber. Huber was born in Geneva, Switzerland, in
1777. His father, François Huber, was an entomologist who was
deeply involved in the study of honeybees. The elder Huber
became blind in 1773 at the age of twenty-three. Nevertheless,
he continued his research by drawing on the observational skills
of his wife, Marie Lullin, and his servant, François Burnens. With
their help as observers, Huber published a major work based on
systematic observation of honeybees (Huber and Bonnet, 1792).
That publication had enough impact that he was thereafter
known as “Huber des abeilles (Huber of the bees).” Even now
his contributions are honored with a walkway named “François
Huber” off the rue de Saint-Victor in Geneva.5
Young Pierre was raised in this household—one that was
almost totally focused on the systematic observation of insect
behavior. It is no wonder then that he went on to a distinguished
career in entomology. At twenty-four, he published a detailed
5There is even a recently published historical novel, The Beekeeper’s Pupil by Sara George
(2003), about the life of François Huber.
16
description of bumblebees (1802). He observed and reported
every aspect of their anatomy and life cycle in painstaking de-
tail. And included in that description was a thorough report of
their dominance behavior with respect to one another. This, so
far as I have been able to discover, provides the earliest example
of a report of patterned social interaction based on systematic
observation.
Pierre Huber went on to work with ants (1810) and achieved
world fame as a naturalist. In fact, Charles Darwin (1859, Chap-
ter VIII) described him “as a better observer even than his cele-
brated father.” The younger Hubers work set the stage for the
development of the ethological approach in biology. At the same
time, it was a precursor of two developments that have become
parts of contemporary social network analysis. First, studies of
social patterning among nonhuman animals continues in social
science. It was picked up first by Moreno’s sociometric commu-
nity in the 1940s (see the special issue of Sociometry, Volume VIII,
Number 1, February 1945). Moreover, recent research in network
analysis still includes observation-based studies of the pattern-
ing of social linkages among nonhumans. And second, network
analysts still conduct systematic studies of dominance. Huber’s
work, then, provided a model for later research in both biology
and social network analysis.
The earliest example of systematic data collection on hu-
mans came a half-century later. The lawyer-anthropologist Lewis
Henry Morgan was born near Aurora in upstate New York in
1818. He attended Union College where he joined a secret soci-
ety called the “Grand Union of the Iroquois.” Members wanted
to model the form of their organization after the Iroquois, but
discovered that little was known about that form.
So several members of the society, including Morgan, went
to nearby Iroquois reservations to learn what they could about
Iroquois social practices. And soon Morgan found himself deeply
involved in the fieldwork. He ended up publishing an ethnog-
raphy of the Iroquois (1851).
At that point, Morgan left off his anthropological pursuits.
He moved to Rochester and began the practice of law. But the
1856 meeting of the American Association for the Advancement
Prehistory: The Origins of Social Network Ideas and Practices
17
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
of Science took place in Albany, and according to Tooker (1997),
that meeting revived Morgan’s interest in the field. Morgan pre-
sented a paper on the Iroquois descent system at the meeting in
1858. His conclusion was that the Iroquois system of naming kin
and reckoning descent was dramatically different from that used
by western Europeans.
Later that year, on a trip to Michigan, Morgan interviewed
an Ojibwa woman and discovered that the Ojibwa had yet an-
other distinct scheme for naming kin and reckoning descent. For
Morgan, these differences provided enough motivation to de-
termine his future career. He began to travel in order to inter-
view representatives of other North American tribes. At the same
time, he went to great effort to enlist the help of others in collect-
ing data in other parts of the world. When he had amassed a
mammoth collection of data, he quit his law practice and devot-
ed his time to the task of bringing all the data together. The re-
sult was a huge volume, Systems of Consanguinity & Affinity of the
Human Family (1871/1997).
In that book Morgan reported the terms used in describing
lineages of peoples throughout the world. He presented data
and proposed that kinship terminologies embodied various pat-
terns and were correlated with forms of marriage and rules of
descent.
A century later, John Atkinson Hobson (1894/1954) devel-
oped an approach to uncovering links among organizations.
Hobson was born in 1858 in Derbyshire, England. He earned an
M.A. degree at Lincoln College, Oxford. He taught literature in
Faversham and Exeter until 1887 when he moved to London.
There, he joined the Fabian Society and found a job as a reporter
for the Manchester Guardian newspaper.
By the time he arrived in London, Hobson had already es-
tablished a reputation for his liberal views on economics. He
had published books on poverty and unemployment, and he
had produced a major work on capitalism—one that was often
cited by Lenin (Hobson 1894/1954). As part of his treatment of
capitalism, Hobson (p. 271) presented systematic data on over-
lapping directorships among members of “the small inner ring
of South African finance.” His table is reproduced as Figure 2.4.
18
It shows the five major South African companies and the six
board members that linked them.6
Thus, Hobson innovated by presenting two-mode (board
member by company) data that reveal links of individuals to
companies. At the same time, these data display links among
the companies in terms of their shared board members as well
as links among the individuals who are co-members of the same
boards. Hobson, then, was the first investigator to collect sys-
tematic data on corporate interlocks. Both of these innovations
are still major parts of contemporary social network analysis
(Levine, 1972; Stokman, Ziegler and Scott, 1985).
The next several contributions all came from developmen-
tal and educational psychologists who worked in the 1920s.
During those years, large long-term grants were made to the
child-welfare institutes at the Universities of Iowa and Minne-
sota, Yale, Columbia Teachers College and the University of
California, Berkeley (Renshaw, 1981). One result of this funding
was a huge increase in the amount of research focused on chil-
dren, and a good deal of that new research was centered on the
study of children’s interpersonal relationships.
These works are for the most part unrecognized in the field
of social network research, but they did succeed in innovating a
Prehistory: The Origins of Social Network Ideas and Practices
6Fennema and Schijf (1978/79) credited Otto Jeidels (1905) with “the first extensive
and systematic study on corporate interlocks.” They did mention that Hobson’s research
was earlier, but they described it as “very partial.” But since Jeidels opened his report
with a quotation from Hobson, the direct line of influence is clear.
Figure 2.4. Hobson’s two mode data
19
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
number of important network ideas and practices. In a study of
homophily among school children, for example, John C. Almack
(1922) developed a way of using interviews to collect network
data. He asked children in a class to name those they would like
to invite to a party. Then he examined the data in terms of the
similarities between choosers and those chosen. Thus, Almack
anticipated by more than ten years the data collection proce-
dure usually credited to Moreno. Questions of this sort contin-
ue to provide a standard method for collecting data in network
analysis. And, by examining the similarities between the choos-
ers and the chosen, Almack anticipated the very keen interest of
contemporary network analysts in homophily as a basis for so-
cial choice.
Beth Wellman7 (1926) collected network data by recording
systematic observations of who played with whom among pre-
school children during periods of free play. She pioneered, then,
in extending Huber’s ethological approach to the study of hu-
man interaction. In network analysis involving human subjects,
data generated by questioning actors are common, but observa-
tional data of social links are still gathered.
Helen Bott8 (1928) went even further, refining Beth Well-
man’s approach in several ways. First, she used ethnographic
methods to uncover the various forms of interaction that occurred
regularly among preschool children. These methods enabled her
to limit systematic observations to the specific kinds of interac-
tion relevant in the context of her research. Second, she was the
first to employ a focal-child method for collecting detailed ob-
servations of who displayed each particular form of interaction
with whom. Thus, she was able to avoid many of the biases in-
herent in other methods of observing interaction (Altmann, 1973).
And third—eighteen years before Forsyth and Katz (1946) sug-
gested using matrices to record interaction patterns—Bott re-
corded her data in matrix form. Although Bott apparently had
little direct influence on the development of social network re-
7Beth Wellman was not a relative of Barry Wellman, who is treated in later chapters.
8Helen Bott was the mother of Elizabeth Bott, a major figure in social network analysis.
Elizabeth was one of the experimental subjects in Helen Bott’s research. Interestingly,
Elizabeth Bott denies that her mother’s work had any influence on her involvement
with the social network perspective.
20
search, and is seldom if ever cited by network analysts, she cer-
tainly anticipated many modern sophisticated methods of data
collection and presentation.9
In 1933 Elizabeth Hagman brought these two approaches
to data collection—observation and interview—together. She
explicitly raised a data-related issue that is still a central con-
cern among contemporary network analysts. She observed which
children played with which others during a period of free play.
Then, at the end of the school term, she interviewed each child.
They were asked to name who their playmates had been at the
beginning of the term, in the middle of the term and at the end
of the term. She then compared the observed data with the re-
ports and examined the discrepancy between the two. that the
discrepancy she found defined a research problem that remains
a key issue to a great many recent investigators in the social net-
work field (Bernard, Killworth, Kronenfeld and Sailer, 1985; Free-
man, Romney and Freeman, 1987).
2.3 Graphic Imagery
Graphic images have had an important place in structural
studies from the earliest days (Freeman, 2000b). The earliest ex-
amples were all focused on kinship. Christine Klapisch-Zuber
(2000) showed that tree-based images were drawn as early as
the ninth century. Those early images, designed to display the
general patterning of kinship, showed pictorially the proximity
between the occupants of any two kin categories. Klapisch-Zuber
(2000, p. 37) went on to report that by the thirteenth century,
drawings of trees were commonly used to depict the lineages of
particular families. So the application of imagery in examining
kinship began very early.
The data on kinship collected by Lewis Henry Morgan were
described above. However, Morgan not only collected a huge
amount of data on kinship terminology, he drew diagrams to
try to specify the positions of equivalent relatives. His drawing
Prehistory: The Origins of Social Network Ideas and Practices
9Bott’s contributions are recognized by developmental psychologists (Renshaw, 1971),
but until her work was rediscovered by Freeman and Wellman (1995), the only citation
from any social network analyst to Helen Bott that I could locate was made by Eliot D.
Chapple (1940).
21
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Figure 2.5. Morgan’s descent system of ancient Rome
22
of the system of descent in ancient Rome is shown in Figure 2.5.
In it, positions are drawn as circles and descent is shown by
lines. Clearly, this image is an ancestor of contemporary pictures
of social networks.
Near the end of the nineteenth century, Alexander Macfar-
lane (1883) developed a formal model of the British kinship sys-
tem. Macfarlane was born in Ontario, Canada, in 1851. He
received a DSc degree from the University of Edinburgh, and
stayed on as a Lecturer in Mathematics there. In 1878 he became
a Fellow of the Royal Society of Edinburgh, but left Scotland in
1893 to become a Professor of Physics at Texas University (now
the University of Texas).
Macfarlane was an algebraist, and his work on kinship was
primarily algebraic, but a part of that work involved construct-
ing visual representations of various degrees of kinship. In the
illustration shown in Figure 2.6, Macfarlane specified all the two-
step marriage relationships that are prohibited in English law.
From left to right, the picture shows that the law prohibits a
male (designated by a +) from producing offspring with his
granddaughter, his sister or his grandmother. Put another way,
it prohibits a woman (o ) from producing a child with her grand-
father, her brother or her grandson. In the figure a short line
crossing a longer one simply designates a person of either sex in
another generation.
Both of these early attempts produced images of graph-
like structures. They both used points to represent social actors
Prehistory: The Origins of Social Network Ideas and Practices
Figure 2.6. Macfarlane’s images of two-step marriage prohibitions
23
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
and lines to represent linkages among them. In both cases, how-
ever, the images as they were drawn were not quite graphs. The
whole meaning in a graph is in its structure—which point shares
an edge with which other point. So when they are drawn, graphs
place points in locations that are arbitrary; their relative posi-
tions—above, below, left or right—have no meaning. But both
Morgan and Macfarlane drew images that were oriented top to
bottom. Both placed genealogical antecedents higher on the page
and descendants lower. Nonetheless, their images anticipated
the pictures produced by network analysts using graph theory.
Hobson’s (1894/1954) data on interlocking directorates was
described in Section 2.2 above. But Hobson not only collected a
new kind of data, he went on to introduce a way to display two-
mode network data. He drew on the data shown in Figure 2.4
along with data from another table in his book and produced a
picture of overlaps, a hypergraph, to show his readers how a
small number of large corporations—notably the De Beers Group
and Rand Mines—could use interlocking directorates to control
many other firms. That hypergraph is shown in Figure 2.7. As
far as I can discover, it is the earliest example of an image of
24
Figure 2.7. Hobson’s hypergraph
overlaps used to display social patterning. Images of this sort
are still widely used in social network analysis.
2.4 Mathematical and Computational Models
Unlike many other
approaches to social
research, network anal-
ysis has consistently
drawn on various
branches of mathemat-
ics both to clarify its
concepts and to spell
out their consequences
in precise terms. The
earliest application of
mathematics to a prob-
lem of social structure
that I have been able to
uncover was algebraic.
Macfarlane was
mentioned above in the
discussion of early
graphic tools. Those
graphics, however,
were secondary to his main contribution. Macfarlane (1883)
worked out an explicit way of characterizing the English system
of kinship by concatenating kin terms. He began with a simple
symbolic system:
There are two fundamental relationships of the high-
est generality, namely, child and parent, the one re-
lationship being the reciprocal of the other. These can
be combined so as to express any of the complex rela-
tionships; thus grandchild is expressed as child of
child…let c be used to denote child, p to denote par-
ent, and let “of” be expressed by juxtaposition, then
grandchild will be expressed by cc.
10Pearson (1914, 1924, 1930), the famous statistician, wrote a three-volume biography
of Galton.
Prehistory: The Origins of Social Network Ideas and Practices
Figure 2.8. Sir Francis Galton, on the
right, 1909, with Karl Pearson10
25
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Macfarlane made his system more complex and more in-
teresting by adding a notation for gender, m and f. Then he went
on to designate the incest prohibitions specified under English
law using his system of symbols and drawings. An example was
shown in Figure 2.6.
At that time, one of the leading figures in the British scien-
tific community was Sir Francis Galton, pictured in Figure 2.8.
Galton worked as an African explorer, a meteorologist, a geog-
rapher, an anthropometrist, a statistician and a geneticist. Like
many of his nineteenth century colleagues, he was something of
a racist. His belief that almost all human characteristics are in-
herited led him to develop the controversial notion of eugenics.
As a consequence, he is less respected in the current world than
he might otherwise have been.
Galton was born in 1822 to a well-to-do Quaker family in
Birmingham, England. He was a gifted child and, following his
mother’s desire, he began the study of medicine as an appren-
tice at the age of sixteen. He attended medical school at Kings
College and Trinity College, Cambridge.
Galton transferred his medical studies to London, but when
his father died in 1844, he dropped out of school and used his
inheritance to become a gentleman-explorer. He explored in
Southwestern Africa, and his successes promoted his recogni-
tion by the members of the scientific community. In 1856, when
he was thirty-four, he was elected a Fellow of the British Royal
Academy.
The most profound influence in his life occurred in 1859
when his cousin, Charles Darwin, published the Origin of the
Species. As Galton later recalled:
The publication in 1859 of the Origin of Species by
Charles Darwin made a marked epoch in my own
mental development, as it did in that of human thought
generally. Its effect was to demolish a multitude of
dogmatic barriers by a single stroke, and to arouse a
spirit of rebellion against all ancient authorities whose
positive and unauthenticated statements were contra-
dicted by modern science.
26
From then on, Galton’s major pursuit involved the study of
inheritance, which led indirectly to his contributions to the de-
velopment of tools for social network analysis. It led, for exam-
ple, to a concern with kinship. So when Macfarlane presented
his paper at a meeting of the Royal Anthropological Institute in
1883, Galton was present. And the minutes of the meeting show
that Galton was the major discussant of Macfarlane’s paper. The
minutes paraphrased Galton’s remarks:
Dr. Macfarlane had attacked the problem of relation-
ship with thoroughness, ability, and success, and that
he had done a very acceptable work for all who con-
cerned themselves with genealogies of the complicat-
ed descriptions referred to by Dr. Macfarlane. The
diagrammatic form seemed to himself the most dis-
tinctive and self-explanatory.
So Galton was involved in Macfarlane’s presentation, but
soon thereafter he made his own contribution. He participated
in developing an early probability-based model in the social net-
work area. Galton had been working on the problem of heredi-
tary genius, and had noticed that families that were prominent
at one time seemed to decline—even disappear—at later times.
He had enough mathematical training to suspect that such dis-
appearances were the result of some probabilistic process, but
he was not a good enough mathematician to spell out the pro-
cess.
Consequently Galton enlisted the help of the Reverend
Henry William Watson, who was both a minister and a statisti-
cian. Galton and Watson (1875) took a stochastic approach to
the study of networks. Galton put the problem as follows:
Let p0 , p1 , p2 ,…be the respective probabilities that a
man has 0, 1, 2,…sons, let each son have the same
probability for sons of his own and so on. What is the
probability that the male line is extinct after r genera-
tions?
Drawing on probability theory, Watson built a stochastic
model of the disappearance of family names—a network process.
Prehistory: The Origins of Social Network Ideas and Practices
27
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
His model was essentially computational; his solution involved
assigning values to their basic parameters and deriving a nu-
merical solution based on those parameters.
Watson began with the notion of a population of family
names. He proposed a set of parameters involving the probabil-
ities of a given man producing 0, 1, 2,…, q male offspring. From
these properties, Watson calculated the expected proportions of
each surname in each succeeding generation. Of course, any
surname holder who produced no male offspring would con-
tribute to the reduction of representatives of his name in suc-
ceeding generations. So Watson was able to show that, simply
by a random process of reproduction, “We have a continual ex-
tinction of surnames going on.” His conclusion was that any fam-
ily name would ultimately disappear with a probability of 1.
Among network analysts, Galton and Watson are tradition-
ally credited with having been the first to solve the problem of
the disappearance of family names (Mullins and Mullins, 1973,
p. 257). But, as Kendall (1974) documented, that credit is mis-
placed. There were two important limitations of the Galton and
Watson work. First it was incomplete, and second it was actual-
ly preceded by an earlier work that had produced a more com-
plete answer.
In 1845, a French demographer and mathematician, Irénée
Jules Bienaymé, had already examined the same problem (Hey-
de and Seneta, 1977, p. vii). Bienaymé’s work was motivated by
reports by Doubleday (1842) and de Châteauneuf (1845) both of
whom had observed that the surnames of noble families tend to
disappear over time. Bienaymé developed a model designed to
explain that observation. He said (translation by Heyde and Sen-
eta, 1977, pp. 117–118):
If…the mean of the number of male children who re-
place the number of males of the preceding generation
were less than unity, it would be easily realized that
families are dying out due to the disappearance of the
members of which they are composed. However, the
analysis shows further that when this mean is equal
to unity families tend to disappear, although less rap-
idly…The analysis also shows clearly that if the mean
28
ratio is greater than unity, the probability of the ex-
tinction of families with the passing of time no longer
reduces to certainty. It only approaches a finite limit,
which is fairly simple to calculate and which has the
singular characteristic of being given by one of the
roots of the equation (in which the number of genera-
tions is made infinite) which is not relevant to the
question when the mean ratio is less than unity.
This last point illustrates the incompleteness of the later
Galton and Watson work. Watson dealt only with the case in
which the mean number of male children was equal to one. His
conclusion that names would always disappear is true only un-
der that condition or when the number is less than one.
It is interesting to speculate about why the work of Bi-
enaymé was lost and why Galton and Watson have traditionally
been given credit for solving the problem. The answer, I think,
lies in the relative prominence of Bienaymé and Galton. Bi-
enaymé spent most of his career working as a French civil ser-
vant. He did publish mathematical papers, but relatively
infrequently; there were thirty-four in all. Moreover, it was not
until late in his career that he received any academic recognition
at all: in 1848 he was awarded a temporary appointment as Pro-
fessor of the Calculus of Probabilities at the Sorbonne, and in
1852 he was elected to the Académie des Sciences de Paris (Heyde
and Seneta, 1977, p. 7).
In contrast, Galton came from a prominent family and was
a major figure in the British Academy from the time he was in
his early thirties. Moreover, he produced, “…considerably more
than a hundred and seventy publications…” (Galton, 1908, Chap-
ter XX). It is clear, then, that Galton’s relative prominence put
him in a much better position to be recognized and remembered.
Some justice is evident, however, in the fact that in current works
by probabilists the problem of the disappearance of lineages is
called the “Bienaymé-Galton-Watson process.”
The materials introduced in this chapter are all early exam-
ples of important innovations in each of the four features that
characterize contemporary network analysis. Various early writ-
ers (1) produced structural intuitions, (2) collected systematic
Prehistory: The Origins of Social Network Ideas and Practices
29
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
who-to-whom data, (3) produced graphic images of structural
forms and (4) developed both mathematical and computational
models.
Some, like Morgan (1871), Macfarlane (1883) and Hobson
(1894), produced work that embodied two of the features. Mor-
gan collected huge amounts of systematic data on kinship and
displayed his results in graphic images. Macfarlane developed
an algebraic model of kinship and he too used graphic images
to display its properties. Hobson collected systematic data on
corporate interlocks, then drew hypergraphs to reveal their ob-
served interlock patterns. By employing two of the four tools
that define modern social network analysis, these nineteenth
century investigators began to approach current practice.
From a network perspective, this chapter has raised some
interesting issues. First, of all the people discussed here, only
one, Georg Simmel, is generally recognized as having been in-
fluential in forming a foundation for social network analysis.
Maine, Tönnies, Durkheim, Spencer, Cooley and von Wiese have
been mentioned in the network literature, but rarely. There has
been at least one citation to Hobson and one to Galton and Wat-
son. This suggests that both works may have had some impact
on the development of the field. But, as far as I can discover,
even though their work could be expected to have influenced
the development of the field, all the others I have discussed in
this chapter are simply unknown in the social network commu-
nity.
In any case, in the early 1930s a broad research effort called
sociometry was introduced. It was the first work that included
all four of the defining features of social network analysis. In the
next chapter I will review its development.
30
Chapter 3
The Birth of Social Network
Analysis I: Sociometry
Jacob Levy Moreno was the force behind the development
of sociometry. He was an enigmatic figure. He was bright—
perhaps brilliant—he was wildly creative, he was entertaining
and he was blessed with boundless energy. Like many Viennese
intellectuals of the early twentieth century he had presence and
style. But, at the same time, Moreno had a dark side. He was
self-centered, self-serving and by his own description, megalo-
maniacal. He admitted hearing voices, he sometimes thought
he was God and he was convinced that others were always steal-
ing credit for ideas that were his. Moreno, then, was both a dy-
namic intellectual innovator and a severely troubled human
being. His role in the development of social network analysis
can be understood only by considering both facets of his per-
sonality.
At one level Moreno set out to create ambiguity. There are
discrepant accounts of his life, beginning with his name and the
circumstances of his birth. He was actually born in his parents’
home in Bucharest on May 18, 1889, and he was originally given
the name Jacob Levy. His biographer, Marineau (1989), has re-
produced both his birth certificate (p. 7) and the register of his
graduation from the faculty of medicine in Vienna (p. 8). Both
documents indicate that he was born in 1889 and that his name
was Jacob Levy, not Jacob Moreno.
31
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
But in his autobiography (1985, Ch. 1, p. 6) Jacob Moreno
claimed to have been born in 1892 on the 400th anniversary of
the forced exodus of Jews from Spain. He wrote:
I was born on a stormy night on a ship sailing the
Black Sea from the Bosphorus to Constanta in Rou-
mania. Dawn of the Holy Sabbath and the delivery
took place just before the initial prayer. My being born
on a ship was due to an honorable error, the excuse
being that my mother was only sixteen and little ex-
perienced in the mathematics of pregnancy. No one
knew the identity of the ship’s flag. Was she a Greek, a
Turkish, a Roumanian or a Spanish ship? The ano-
nymity of the ship’s flag started off the anonymity of
my name and the anonymity of my citizenship. When
World War I broke out in 1914 no one knew whether
I was a Turk, a Greek, a Roumanian, an Italian, or a
Spaniard because I had no birth certificate…. I was
born a citizen of the world.
The fact is that Moreno was the eldest of six children born
to Moreno Nissim Levy and his wife, Paulina Iancu. Moreno
Levy was a Sephardic Jew and a relatively unsuccessful travel-
ing salesman. Paulina was 15 years old—17 years younger than
her husband—when the marriage took place. She too was
Sephardic, but she had been educated in a Catholic convent. So
she ended up embracing both Jewish and Christian beliefs. As
Marineau (1989, p. 13) put it, “Her hero, who was to become
Jacob’s model, was Jesus Christ.”
Jacob was his mother’s favorite, particularly after a gypsy
reportedly advised her that “The day will come when this boy
will become a very great man. People will come from all over
the world to see him.” His mother passed this tale around and it
became the first of many stories that were later integrated into
the legend that was created by the adult Moreno.
Another part of that legend was reported by Moreno him-
self (1946, p. 2):
When I was four-and-a-half years old my parents lived
in a house near the river Danube. They had left the
32
house on a Sunday to pay a visit, leaving me alone in
the basement of the house with the neighbors’ chil-
dren. The size of this basement was about three times
that of an average room. It was empty except for a
huge oak table in the middle. The children said, “Let’s
play.” One child asked me: “What?” “I know”, I said,
“lets play God and his angels.” The children inquired:
“But who is God?” I replied “I am God and you are
my angels.” The children agreed. They all declared:
“We must build the heavens first.” We all dragged
chairs from every room in the house to the basement,
put them on the big table and began to build one heav-
en after another by tying several chairs together on
one level and putting more chairs above them on the
big table until we reached the ceiling. Then all the chil-
dren helped me to climb up until I reached the top
chair and sat on it. There I sat pretty. The children
began to circle around the table, using their arms as
wings, and singing. Suddenly, I heard a child asking
me: “Why don’t you fly?” I stretched my arms, try-
ing it. A second later I fell and found myself on the
floor, my right arm broken.
According to Marineau (1989, p. 17) this business of playing God
was repeated often, and it was encouraged by Moreno’s mother.
In the late 1890s the family moved to Vienna and young
Jacob entered school. He was an excellent student, but when his
parents separated he began to rebel. He dropped out of school,
wore outlandish clothes and grew an unkempt full beard—un-
heard of in that period in Vienna. He took to hanging around
the University of Vienna, and in 1909 he was finally admitted as
a temporary student in philosophy. The following year he was
able to transfer into the school of medicine. He received his med-
ical degree in 1917 with a specialty in psychiatry.
After receiving his degree Moreno somewhat reluctantly
started a medical practice. He moved to Bad Vöslau, a small town
in Austria, where in 1920 he began a serious long-term relation-
ship with a very beautiful young woman, Marianne Lörnitzo.
The Birth of Social Network Analysis I: Sociometry
33
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Moreno described Marianne as his “muse.” Marineau (1989,
p. 62) said that Marianne “…was instrumental (and perhaps the
word is too weak) in making Moreno refocus…” In any case,
with Marianne’s influence, Moreno published two works: a long
mystical poem, Das Testament des Vaters (The Words of the Fa-
ther), and a rambling precursor of present day sociodrama and
psychodrama, Das Stegreiftheater (The Theater of Spontaneity).
These were the products of the first of a series of important en-
counters between Moreno and various women. For the most part
Moreno seemed to be unfocused, but when he was involved with
a woman—one who could serve as a “muse”—he succeeded in
concentrating and was able to write.
In terms of social networks the most important aspect of
this period in Moreno’s life is that it includes a demonstration of
his concern with social structure. Das Stegreiftheater contained
his first sociometric diagrams. These show that, from his per-
spective, the therapeutic potential of drama stemmed from the
interactions among the actors. As he (1953, p. xxxv) put it in a
later description of Das Stegreiftheater:
Sociometric measurement started with things like this:
how much “time” does an actor A spend with another
actor B? He may spend half as much time with anoth-
er actor C and three times as much time with another
actor D. Or, what is the “spatial distance,” near or
far, in inches, feet or meters, between actors A, B, C
and D in the course of the same situation and what
effect have nearness or distance upon behavior and
acting? Or, how frequently do two actors appear si-
multaneously in a scene and how frequently do they
exit together?
These kinds of structural intuitions are at the core of all of
Moreno’s work.
In any case, Moreno’s career was not going well in Austria
and at the end of 1925 he decided to emigrate to America. The
plan was for Marianne to maintain their house in Bad Vöslau.
He told her that he would send for her after he became estab-
lished in the new world.
34
Moreno struggled learning English, but in 1927 he had de-
veloped enough skill to allow him to begin working with an
impromptu theater group in New York. There, in 1928, he met
and married Beatrice Beecher,11 reportedly for the sole purpose
of gaining American citizenship. He and Beecher were divorced,
as they had agreed, in 1934 when Moreno did gain his citizen-
ship. Interestingly, he never bothered to tell Marianne about this
marriage. She still expected to join Moreno once he was estab-
lished.
Most important from our present perspective, Moreno also
made another new friend through the theater group, Helen Hall
Jennings. At the time, Jennings was a graduate student in psy-
chology at Columbia University. She worked with the promi-
nent psychologist, Gardner Murphy, and specialized in research
methods and statistics. She introduced Moreno to Murphy and
helped him design and conduct sociometric studies at Sing Sing
prison (Moreno, 1932) and at the Hudson School for Girls
(Moreno, 1934). These two studies resulted in two books being
published within two years. Both studies involved systematic
data collection and analysis, and in his 1934 book he used the
term “network” in the sense that it is used today. By the end of
the Hudson study, the sociometric approach had been fully de-
veloped.
Moreno was listed as the sole author of these two books,
although Jennings was credited with the authorship of a “sup-
plement” in the 1934 volume. Nonetheless, his acknowledge-
ments do include the phrase, “I and my collaborator, Helen
Jennings,…” And in his autobiography he acknowledged her as
a guiding force (Moreno, 1985, Ch. 3, p. 8).
It is impossible to overestimate the importance of Jennings
in these works. Marineau (1989, p. 101) reports that “…it is fair
to say that the work would have lacked precision and refine-
ment without her help.” Hare and Hare (1996, p. 73) describe
her as having played “a major role in the collection and analysis
of the data.” And in his preface to her later book (Jennings, 1943)
Gardner Murphy describes Jennings as “the initiator of many of
the methods” of sociometry. In contrast to Who Shall Survive?
11She was the granddaughter of Henry Ward Beecher, the evangelist.
The Birth of Social Network Analysis I: Sociometry
35
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Jennings (1943) book, Leadership and Isolation, is far more orga-
nized and coherent. Homans (1984, p. 163) described it as “a
much better book.”
My own suspicion is that Jennings’s contributions were
immense. Moreno had no research training at all, and until this
point, he had shown absolutely no interest in systematic research.
All of his earlier publications were couched in heavy mysticism.
But his newer works—those produced during his collaboration
with Jennings—were comparatively systematic and were, for the
first time, empirically grounded. The obvious conclusion is that,
though the intuitive ideas came from Moreno, the completed
research and the publications drew heavily on the contributions
of Jennings. Jennings, it seems, was not only a collaborator but
she was, in Moreno’s terms, a very powerful “muse.”
By the mid 1930s these books, along with public support
from Jennings, Murphy and the prominent psychiatrist, Will-
iam Allison White, turned Moreno into something of a social
science celebrity. He started a journal, the Sociometric Review, in
1936. But a year later he dropped that publication and started
another journal, Sociometry. By the late 1930s he could count
among those who were involved in his work such notables as
Franz Alexander, Gordon W. Allport, Read Bain, Howard Beck-
er, Franz Boas, Emory S. Bogardus, Jerome S. Bruner, Hadley
Cantril, F. Stuart Chapin, Leonard S. Cottrell, Stuart C. Dodd,
Paul Lazarsfeld, Kurt Lewin, Charles P. Loomis, George A. Lun-
dberg, Robert S. Lynd, Margaret Mead, Karl Menninger, George
Peter Murdock, Gardner Murphy, Theodore M. Newcomb, Wil-
liam H. Sewell, Pitirim Sorokin and Samuel Stouffer. These names
are a virtual Who’s Who in American sociology and psychology
at the time.
In a period in which most mainstream sociology had be-
come psychologistic, it is remarkable that it took a psychiatrist,
Moreno, and a psychologist, Jennings, to reintroduce a perspec-
tive that was distinctly structural. Like others at the time, Moreno
and especially Jennings were concerned with psychological prop-
erties of individuals; they used sociometric questions to deter-
mine people’s feelings about one another. They asked the
members of a defined population to name the others with whom
36
they would like to live, for example, or work or spend leisure
time. The responses, then, were individual choices. But since
the Moreno-Jennings approach used these choices to uncover
groups and positions of individuals within those groups, their
approach was essentially structural.
According to the New York Times (1933) Moreno first called
his approach “psychological geography.” But by 1934 he had
renamed it sociometry. As Moreno (1934, pp. 10–11), described
it, sociometry was based on an “…experimental
technique…obtained by application of quantitative
methods…which inquire into the evolution and organization of
groups and the position of individuals within them.”
This is certainly a structural concern, and one that falls at
the very heart of social network analysis. A great deal of con-
temporary work is involved in the specification of both groups
and positions. Examples of recent work on groups can be found
in Alba (1973), Breiger (1974), Freeman, (1992), Mokken (1979),
Winship (1977) and Yan (1988). Examples of work on positions
can be found in reports by Breiger and Pattison (1986), Lorrain
and White (1971), Sailer (1978), White and Reitz (1983) and Win-
ship and Mandell (1983).
The question, then, is whether Moreno and his supporters
exhibited all the features that define modern social network anal-
ysis. The Hares (1996, p. 30) reported that, “Moreno’s theory of
society focused on the networks of interpersonal relations that
join individuals.” That point is documented throughout Moreno’s
1934 edition of Who Shall Survive? There, Moreno not only wrote
explicitly about “networks,” but he referred to “the effects be-
yond the two persons and the immediate group” of an interact-
ing pair (p. 347). He (Moreno et al., 1960) illustrated the point by
showing that an epidemic of runaways among the girls at the
Hudson school could be explained by chains of social ties that
linked all of those who had left. This is a clear and dramatic
example of structural thinking.
The book includes systematic empirical data collected in a
nursery, several classrooms and the cottages and work groups
in the Hudson school for girls. Graphics were used throughout.
An example, shown in Figure 3.1, shows the positive and nega-
The Birth of Social Network Analysis I: Sociometry
37
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Figure 3.1. Sociogram of a football team
tive sociometric choices made by members of a football team (p.
213).
There was one defining element of contemporary social
network analysis, however, that was missing in these two books.
The 1932 and 1934 volumes included no mathematical or com-
putational models at all. But Moreno (or, more likely, Jennings)
soon recognized that deficiency and enlisted the help of the
eminent mathematical sociologist at Columbia, Paul Lazarsfeld.
In response, Lazarsfeld created a base-line model for sociomet-
ric choice. Given random choices, he worked out the probability
that any particular individual would be chosen by any other
particular individual. Moreno and Jennings then published Laz-
arsfeld’s model in an article in volume 1, number 2 of the new
38
journal Sociometry (1938, pp. 342–374). According to Nehnevajsa
(1956), the Lazarsfeld model “…stands at the roots of probabi-
listic analysis in the field.”
By 1938, then, the work of Moreno—with the help of Jen-
nings and Lazarsfeld—had displayed all four of the features that
define contemporary social network analysis. It is clear, more-
over, that they recognized the generality of their approach. They
collected data on positive and negative emotional choices and
on who was acquainted with whom. They observed interaction
patterns linking individuals. They discussed kinship ties. And
they examined social roles. As Moreno (1937) put it:
The first decisive step in the development of Sociome-
try was the disclosure of the actual organization of
the group. The second decisive step was the inclusion
of subjective measures in determining this organiza-
tion. The third decisive step was a method which gives
to subjective terms the highest possible degree of ob-
jectivity, through the function of the auxiliary ego.
The fourth decisive step was the consideration of the
criterion (a need, a value, an aim, etc.) around which
a particular structure develops.
Every one of these steps was focused on the structural pattern-
ing of the data.
Clearly, with a great deal of help from Jennings and Laz-
arsfeld, Moreno had developed an approach that included all of
the defining properties of social network analysis. It was based
on structural intuitions, it involved the collection of systematic
empirical data, graphic imagery was an integral part of its tools
and it embodied an explicit mathematical model. That structur-
al perspective, moreover, was generalized to a range of phenom-
ena. Thus, the group led by Moreno displayed all four of the
features that define social network analysis.
At this point, sociometry had the attention of the elite of
the American social science community. So the question is, why
didn’t the sociometric model of the social network approach take
off after the late 1930s? What led to the marginalization of so-
The Birth of Social Network Analysis I: Sociometry
39
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
ciometry and the necessity for the reintroduction of social net-
work analysis at a later time?
The answer, I think, can be found in the other aspect—the
dark side—of Moreno’s character. For a brief period Jennings
was able to “clean up his act” and present him to the world as a
creative social scientist who was capable of conducting system-
atic research. But she was unable to suppress the fundamental
flaws in his character. So although Jennings introduced us to
Moreno as Dr. Jeckyll, as time went on, Mr. Hyde kept reappear-
ing.
During his university years Moreno (1985, Ch. 1, p. 2) con-
fessed, “I began to believe that I was an extraordinary person,
that I was here on the planet to fulfill an extraordinary mission.”
So as a student he formed a new religion, became its one proph-
et and began to gather disciples. By 1920 he reported hearing
voices—both as a child and as an adult. But he hastened to add
that that he didn’t hear voices “…as a mental patient does.”
He went on to claim that he experienced “direct encoun-
ters with God” (Hare and Hare, 1996, p. 26). And, he insisted, he
often “played” God. In particular, he played God with his sec-
ond wife. In 1938, Moreno had married a Columbia student
named Florence Bridge. She apparently adored her husband and
she encouraged his God playing. But despite her adoration, it
turned out that Bridge was unsatisfactory as a “muse.”
In 1941, when he met another beautiful young woman,
Celine Zerka Toeman, he finally found the “integrated partner”
he had been seeking. He married Zerka in 1949, and according
to Marineau (1989, p. 107):
Zerka was his partner in all his publications, confer-
ences, workshops, and many other ventures. There is
no doubt that she was not only his inspiration, but
also full-time administrator, organizer, co-trainer and
co-therapist.
Zerka particularly supported Moreno’s ideas about thera-
py, and the two of them began to put increasing effort into the
development of group psychotherapy, psychodrama and socio-
drama as therapeutic techniques. Thus, by the end of the 1940s,
40
Moreno had quit making contributions to sociometry and turned
his attention almost completely to various forms of therapy.
In 1953, however, Moreno did publish a second edition of
his 1934 book, Who Shall Survive? That book embodied several
new sections, including in its first 100 pages his “Preludes to the
sociometric movement.” These preludes take the form of a long
rambling narrative. In it he referred often to God. He laid claim
to having originated virtually every important idea in social sci-
ence. He demeaned the contributions of others including Marx
and Freud. And he finally wound up with several pages of dia-
tribe condemning Kurt Lewin and especially Lewin’s students,
whom he accused of having a problem “of interpersonal ethics.”12
He laid particular blame on Ronald Lippitt, Alvin Zander, John
R. P. French, Alex Bavelas, Leland P. Bradford, C. Hendry, Mar-
garet Barron and Kenneth D. Benne. He accused these and other
students of Lewin of crediting Lewin for ideas they learned while
attending Moreno’s workshops. As he put it, they used “a tech-
nique of quoting only each other, that is, those who belong to
their clique, and not quoting any of my close associates or my-
self.”
Moreno’s increasing preoccupation with therapeutic pro-
cedures, along with his apparent abandonment of systematic
sociometric work, tended to drive many early supporters away.
For those who were interested in empirically based structural
research, the attraction of the sociometric paradigm was dulled
by Moreno’s insistence that sociometry had to do, not with struc-
tural research, but with ambiguous links to God and to psycho-
therapy. Then too, his increasing megalomania put people off.
The kinds of statements described above led research-oriented
scholars to have serious misgivings about the whole sociomet-
ric enterprise. In fact, it is reasonable to conjecture that the un-
willingness of the Lewin students to associate their names with
Moreno was, for the most part, a consequence of his own bi-
zarre behavior.
My own memories of Moreno date from the 1950s when he
was still presenting his work at annual sociology meetings. Those
12Italics in the original.
The Birth of Social Network Analysis I: Sociometry
41
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
presentations always took the form of demonstrations of im-
promptu theater. His style was bombastic and overbearing but
his theater was always entertaining. Although Moreno’s dramas
might have provided entertainment, they seemed to me to lack
serious intellectual content—they failed to cast any light on so-
ciological questions.
My reactions were not unique. Moreno’s work attracted ever
fewer sociological followers. After 1955 Moreno turned his jour-
nal, Sociometry, over to the (then) American Sociological Society,
and it became a general social psychology journal (renamed So-
cial Psychology Quarterly). As the years went by, the earlier ro-
mance between sociology and sociometry faded completely
(Hare and Hare, 1996, p. 107).
Jacob Moreno, then, was—at least for a short time—a ma-
jor intellectual force. With support from various allies he de-
fined an approach that contained all the features of modern social
network analysis. And in the 1930s and 40s he succeeded in at-
tracting the interest of most of the major figures in empirical
social research.
But his commitment to mysticism, his bombastic personal
style and his megalomania drove most of his early supporters
away. These features of Moreno’s persona were too much for
regular members of the academic community to bear. They sim-
ply could not be accommodated in the day-to-day academic or
scientific world. As his biographer Marineau (1989, p. 121) put
it, “Very few young students of sociology or social psychology
today would ever have suspected the impact that Moreno had
on this field of research and practice more than fifty years ago.”
So in the field of sociometry, the techniques of contempo-
rary social network analysis were developed and then aban-
doned. Most people simply backed off from Moreno the man,
and in so doing they refused to recognize the importance and
generality of Moreno’s approach. Social research, then, contin-
ued to lack a unified structural paradigm. In the next chapter I
will describe another attempt to provide such a paradigm.
42
Chapter 4
The Birth of Social Network
Analysis II: The First Harvard
Thrust
A research effort that focused on the study of social struc-
ture began at Harvard in the late 1920s. Centered in the Gradu-
ate School of Business Administration on one side of the Charles
River and the Society of Fellows on the other, it involved a rela-
tively large number of faculty, including William Lloyd Warner,
George Elton Mayo, Fritz Roethlisberger, T. North Whitehead
and Lawrence J. Henderson. A number of students from a vari-
ety of disciplines were also involved, including Eliot Chapple,
Conrad Arensberg, Allison Davis, Elizabeth Davis, Burleigh
Gardner, George Caspar Homans and William Foote Whyte.
The main intellectual thrust for the study of social struc-
ture at Harvard came from Warner (shown in Figure 4.1), who
called himself “W. Lloyd Warner.” Warner was born in Colton,
California in 1898.13 After high school, he started college at the
University of Southern California. But after his freshman year
he transferred to the University of California, Berkeley, where
he majored in English. Although in 1923 he dropped out and
spent a year working in New York and acting on Broadway14,
13Both Collins (1994) and Scott (2000, p. 16) have incorrectly characterized Warner as
Australian.
14He had a walk-on part in the George Arliss play, The Green Goddess (Warner, 1988,
p.10).
43
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
the following year he returned to Berkeley. With the help of the
anthropologist Robert Lowie he was readmitted to the university.
Having gained Lowie’s support, Warner switched to an-
thropology and received his B.A. in that field in 1925. He en-
tered graduate school at Berkeley and worked with Lowie and
with the British anthropologist Bronislaw Malinowski who was
visiting in the Berkeley department. But then another British
visitor arrived for a brief stay. Alfred Reginald Radcliffe-Brown
stopped in at Berkeley en route from Oxford to a new position
at the University of Sydney.
During his Berkeley stay, Radcliffe-Brown attended a din-
ner put on by the Anthropology Club. At one point in the pro-
ceedings, he called across the table, “I say, Warner, how would
you like to come to Australia with me?” Warner gave a joking
reply, but Radcliffe-Brown convinced him that the offer was gen-
uine, and Warner accepted it. Radcliff-Brown sent him off to
Harvard for a short stay so that he could learn something about
physical anthropology
from Ernest Hooton.
Then Warner sailed to
Australia, arriving in
Sydney in January of
1927. After a little more
preliminary training,
Radcliffe-Brown sent
him out to the bush to
do field work.
Warner worked
among the Murngin for
more than two years. He
collected ethnographic
data, and following a
suggestion by Radcliffe-
Brown, he concentrated
particularly on deter-
mining their kinship
patterns and rules of
descent (Warner, 1937).
Figure 4.1. W. Lloyd Warner, circa 1920s
44
Then, in 1929, again with Radcliffe-Brown’s support, he returned
to Harvard, this time as an Instructor in anthropology and a Tutor
at Kirkland House.
Even before leaving Australia, Warner had described to
Radcliffe-Brown his desire to use ethnographic field methods in
the study of industrial communities (Warner, 1988, p. 41). And
once he got to Harvard, he immediately set out to do just that.
He chose the relatively small industrial city of Newburyport,
Massachusetts, as a study site, and recruited a collection of Har-
vard graduate and undergraduate students to collect data. An
undergraduate senior, Eliot Chapple, was appointed field direc-
tor. Others involved included a married couple, Allison and Eliz-
abeth (Liddy) Davis, Conrad Arensberg, Solon (Sol) Kimball, Leo
Srole, Buford Junker, J. O. Low, Paul Lunt and Burleigh Gard-
ner.
The “Yankee City” study, as it was called, focused primari-
ly on the study of stratification. The orientation of the research
was influenced by Durkheim (Warner and Lunt, 1941, p. 10),
and by Simmel (pp. 12–13). It centered on studying interaction
among individuals (Warner and Lunt, 1941, p. 12):
Throughout our research we have employed the con-
cepts of interaction between two or more individuals
and the social interrelationships within which these
interactions take place. The explicit, overt behavior of
individuals, verbally or bodily, as well as “mental at-
titudes or psychological occurrences within the minds
of the individuals” studied, have been understood by
us “as a product of mutual determinations and recip-
rocal influences.”…The larger systems of interrela-
tions which compose the extremely complex and highly
elaborate society of Yankee City were studied in spe-
cific detail, as were the interactions, direct and indi-
rect, of the individuals who constituted the biological
units of the group.
Thus, the focus of this research was on interpersonal net-
works. Moreover, over a period of data collection that lasted sev-
eral years, as many as a dozen graduate students (Warner, 1988,
The Birth of Social Network Analysis II: The First Harvard Thrust
45
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
p. 86), produced “literally tons” of empirical data (Roethlisberg-
er, 1977, p. 55). And, throughout the report, graphic images were
employed. Figure 4.2 is a diagram of an idealized version of the
hierarchical organization of overlapping cliques where the ver-
tical dimension is social status (Warner and Lunt, 1941)
Warner s organization of the Yankee City project, however,
was just the beginning of his contribution to spreading the gos-
pel of structural analysis at Harvard. Almost immediately upon
his arrival he met George Elton Mayo (known simply as Elton
Mayo). Mayo was from Australia. He had trained in medicine,
but quit before he received his degree. Instead, he switched to
psychology and received a bachelor’s degree from the Universi-
ty of Queensland. Mayo taught at Queensland until 1922, when
he migrated to the United States. He was supported by a Rock-
efeller grant and went to work at the Wharton (business) School
of the University of Pennsylvania doing industrial research. In
1926 he accepted an appointment in the Harvard Business School.
By the time Mayo got to Harvard, engineers and personnel
managers at the Western Electric Corporation in Cicero, Illinois,
were already deeply involved in the first major research study-
ing industrial productivity. Western Electric was at that time a
15This research was conduced before Luce and Perry (1949) formally defined cliques.
The term “clique” was used in the Yankee City study in an informal intuitive sense.
Figure 4.2. An idealized model of overlapping “cliques”15
46
subsidiary of the Bell Telephone Company, responsible for the
manufacture of telephones and phone equipment. The project
had begun in 1924. It grew out of an attempt to determine the
degree to which the use of artificial lighting might affect worker
productivity on the production line (Gillespie, 1991, p. 38). The
electrical industry encouraged Western Electric to do the study
in the hope that with the use of electric lights productivity would
rise and industry would, therefore, be encouraged to consume
more power. But from the perspective of the electrical industry,
the early results were bad. They showed that lighting had little
if any effect on productivity. The type of supervision, it turned
out, was a far more important factor.
In 1927 the director of the Western Electric research, George
Pennock, attended a lecture by Elton Mayo. Pennock was anx-
ious to extend the research into new areas and he was impressed
with Mayo’s talents. So, beginning in 1928, Western Electric en-
listed Mayo’s participation in their research effort. Adams and
Butler (2000, p. 126) described Mayo’s first visit to the Western
Electric plant:
The company put him up at a Chicago landmark, the
Palmer House hotel, where he was extended the same
courtesy (or privilege) as Hawthorne Works manager
C. L. Rice. A bemused Mayo wrote his wife: “Every
morning at 8:30 the doorman clears the taxis away
from the Wabash St. entrance of the hotel—-and a large
limousine with a uniformed chauffer slides noiseless-
ly in. The door is opened and Elton Mayo, formerly of
South Australia, gets in and glides off to his alleged
industrial researches.”
Once established, Mayo changed the focus of the Western
Electric research. The earlier investigations had examined the
effects of lighting and other physical interventions on produc-
tivity. As an alternative, Mayo suggested concentrating on the
effects of the psychological characteristics of the workers on their
productivity. He put together a team from the Harvard Gradu-
ate School of Business Administration. The Harvard team in-
The Birth of Social Network Analysis II: The First Harvard Thrust
47
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
cluded T. North Whitehead16 and one of Mayo’s earlier students,
Fritz Jules Roethlisberger. These two worked on the project along
with Mayo and two Western Electric employees, William Dick-
son and Harold Wright.17
In 1930, Mayo recruited Warner as a half-time member of
the Business School faculty and as an advisor to the Western
Electric project. Warner s (1988, p. 45) perspective on the research
was expressed in an early memo to Mayo:
An important problem before us at the present time in
the Western Electric research is to study and under-
stand the total social organization of each of the test
rooms. This includes not only the formal industrial
structure which the company has created but also the
organizations formed by the employees in their con-
scious or unconscious attempts to form themselves into
a group of their own….The first step necessary is to
itemize the number of social personalities found in each
place—by this I mean the types of occupations per-
formed by the workers found in the two test rooms….
All of these social personalities are integrated by a set
of primary and secondary relationships which can be
analyzed into separate parts. [Sixteen categories are
listed]…Following the analysis of the social situation
into its primary reciprocals, it will be necessary then
to combine the various reciprocals into secondary as-
sociation to see how these various social bonds react
on each other and finally to integrate the social situa-
tion in the test rooms on the basis of social relation-
ships which exist there.
From this statement it is clear that, by then, Warner had
fully embraced a structural perspective. In it he stressed exam-
ining the patterning of the informal ties among the workers, and
48
16T. North was a faculty member in the Harvard Business School. He was the son of
the eminent philosopher, Alfred North Whitehead. Both Whiteheads were at that time
teaching at Harvard.
17Mayo’s research was famous for producing the “Hawthorne effect” in which the
mere presence of observers seemed to enhance worker productivity.
that kind of emphasis on studying informal links is still found
in contemporary social network analysis. His notion of “second-
ary reciprocals” undoubtedly came from his earlier research on
Murngin kinship in Australia, designed to call attention to the
implications of social ties for one another. Both of these ideas
expressed particularly sophisticated views for the period.
Mayo got the message. He was convinced that Warners
proposed focus on the social aspects of behavior was important.
He (Mayo, 1933, pp. 116–117) later reported his reaction to Warn-
er’s views:
A representative of the Harvard Department of An-
thropology had called attention to the logical insuffi-
ciency of a merely psychological study of the
individuals in the department (the wiring room). Lab-
oratory and clinical psychological studies are inter-
ested in the individual…but they do no more than
touch the fringe of human inquiry.
Consequently, because of Warner’s input to the Western
Electric project, the research deviated from Mayo’s strict psy-
chological focus on individual characteristics; instead, it focused
on social structure. The result was a study of the patterning of
interaction among the workers in the bank wiring room. That
research was described in detail by Roethlisberger and Dickson
(1939) in their book, Management and the Worker.
Following Warner s suggestion, data on interpersonal in-
teraction were collected using systematic observation. An ob-
server was placed in the workroom. He recorded six different
kinds of interpersonal links: (1) who played games with whom,
(2) who engaged in controversies with whom, (3) who traded
jobs with whom, (4) who helped whom, (5) who displayed friend-
ly behavior toward whom and (6) who was antagonistic toward
whom. Since it was based on systematic observation of links
among individuals, the bank wiring room study generated so-
cial network data. These data were reviewed by Mayo, Roethlis-
berger, T. N. Whitehead, Warner and George Homans along with
several Western Electric Executives (Roethlisberger, 1977, p. 49).
The Birth of Social Network Analysis II: The First Harvard Thrust
49
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
But in addition to structural intuitions and systematic data
on social links, this research also generated graphic images of
network ties. The patterning of each of the social ties was dis-
18See footnote 15 above.
Figure 4.4. Reported “cliques” in the bank wiring room18
Figure 4.3. Friendships in the bank wiring room
50
played using directed graphs, and a generalized overall “clique”
structure was produced in the form of a hypergraph. These fea-
tures are reproduced in Figures 4.3 and 4.4.
Overall, then, the bank wiring room research was quite so-
phisticated in its approach. It embodied three of the four ele-
ments found in current social network research. It missed only
by not including any mathematical or computational tools. But
in every other respect it came very close to the contemporary
standard.
In 1933 Warner continued his research on stratification by
organizing the “Deep South” project. Deep South was a study
of the impact of race differences on social stratification that was
conducted in Natchez Mississippi. Natchez was chosen because
it was roughly the same size as Newburyport and it was a total-
ly segregated city with a large black population.
Warner chose a black couple,
Allison and Elizabeth Davis, and a
white couple, Burleigh and Mary
Gardner, to settle down in Nachez
and do fieldwork in the black and
the white communities. Then Alli-
son Davis recruited my own under-
graduate mentor and his former
undergraduate student, St. Clair
Drake, to help in the collection and
analysis of the data in the black
community.19 Davis and Drake are
pictured in Figures 4.5 and 4.6 re-
spectively. Their effort, along with
that of the Gardners, produced a
book titled Deep South (Davis, Gard-
ner and Gardner, 1941).
As in the Yankee City research,
Davis, Gardner and Gardner were
concerned with the question of the degree to which members of
various social classes limited their interaction to others at ap-
Figure 4.5. Allison Davis
honored on a U.S. postage
stamp
The Birth of Social Network Analysis II: The First Harvard Thrust
51
19 The rumor was that Davis turned out to be uncomfortable interviewing lower class
informants and recruited Drake to do that job.
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
proximately the same social class level. To find out, they record-
ed the observed patterns of interaction that linked collections of
individuals. They set out to specify the “clique” memberships
of individuals in terms of who interacted with whom. They col-
lected systematic two-mode network data on co-attendance
among a collection of eighteen white women. These data are
shown in Figure 4.7. And with the help of Drake they collected
similar data linking black males and produced structural imag-
Figure 4.6. St. Clair Drake
Figure 4.7. Attendance of 18 women at 14 social events
52
es in the form of hypergraphs showing “clique” memberships
(see Figure 4.8).
The Davis, Gardner and Gardner report, then, displayed
the same structural perspective that was present in all of the
work by Warner and his students. They were motivated by struc-
tural ideas, they collected systematic who-to-whom data and
they used graphic images to display their results. But nowhere
in any of this early work is there any use of mathematical or
computational tools.
During this period Warner organized regular seminars at
Harvard that aimed at organizing and coordinating these com-
munity-study projects. The seminars were attended by those
faculty members who were involved in the Western Electric re-
search: Elton Mayo, T. North Whitehead and Fritz Roethlisberg-
20See footnote 3 above.
The Birth of Social Network Analysis II: The First Harvard Thrust
Figure 4.8. “Clique” memberships of males20
53
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
er. Another frequent attendee was Lawrence Joseph Henderson
of the Business School.
At that time, Henderson was a major figure at Harvard. He
was a direct descendant of an old-line family in Salem, Massa-
chusetts. He had attended Harvard College and the Harvard
Medical School where he earned an M.D. degree. He went to
work at Harvard where he became a prominent professor of bi-
ological chemistry. His work on blood chemistry was recognized
throughout the world. Henderson was, moreover, a close per-
sonal friend of the Harvard President, A. Lawrence Lowell.21
Henderson had been teaching in the College of Arts and
Science, not only in biological chemistry but in philosophy and
the history of science as well. In 1927 he shocked the Harvard
community when he moved across the Charles River and joined
the faculty of the Graduate School of Business Administration.
Henderson, it seems, had very wide interests.
Along with Alfred North Whitehead, Henderson was trou-
bled by the limitations in standard graduate education. They
got together with some colleagues and proposed that Harvard
develop a system of fellowships. Their idea was that a few Se-
nior Fellows could guide the work of a small number of out-
standingly talented Junior Fellows. This would free up the Junior
Fellows from the usual dull routine of graduate study, examina-
tions and the like, and allow them to pursue their own inter-
ests—wherever those interests might lead.
In 1932 President Lowell instituted and personally funded
the plan. He appointed Henderson as its first chairman. At that
time Henderson was excited about sociology, or rather, the soci-
ology of the Italian economist, Vilfredo Pareto. A colleague from
entomology, William Wheeler, had just suggested that Hender-
son read Pareto’s work. Henderson was so completely taken with
Pareto’s sociological views (1916/1963) that he introduced a sem-
inar on Pareto that continued for several years.
Despite his achievements in economics, Pareto’s (1916/1963)
sociology is made up almost entirely of psychobabble. For the
most part he tried to account for the existence of various—often
21Harvard at the time was still dominated by the old Massachusetts aristocracy. A.
Lawrence Lowell was one of the Lowells who, as the saying goes, “speak only to God.”
54
questionable—“social facts” in terms of a complex collection of
presumed cognitive characteristics of individuals. Pareto’s “so-
cial facts” included, for example, the fact that people are able to
choose numbers in lotteries. Their choice is “explained by the
human instinct for combinations” (p. 516). Moreover the fact that
people intone religious chants is a result of their “need of ex-
pressing sentiments by external acts” (p. 517).
Notwithstanding the intellectual poverty of Pareto’s soci-
ology, Henderson’s seminar attracted a stellar collection of par-
ticipants. They included the Business School faculty: Warner,
Mayo, Roethlisberger and T. North Whitehead. And they also
brought in Joseph Schumpeter, Crane Brinton, Bernard DeVoto,
Charles Curtis, Hans Zinzer, Talcott Parsons, Kingsley Davis,
Robert Merton and for a brief period, the chairman of Harvard’s
Sociology Department, Pitirim Sorokin. Most important in the
present context, Henderson hired a recent Harvard graduate, a
young aspiring poet, George Caspar Homans, as the seminar’s
administrator. This was Homans’ first introduction to sociology.
Homans reaction to that seminar is documented by the fact
that it led to his first book, written with Charles Curtis (1934).
That book was simply the presentation of Pareto’s ideas written
for an American audience. Homans’ lifelong focus on interac-
tion and sentiments, moreover, turns out to have been inspired
by Pareto (1916/1963, p. 614):
Whatever the causes, groups came to be formed among
many peoples. Presumably they were bound to the soil
and endured in time, the dead being one by one re-
placed by successors. It also happened that the nucle-
us of such groups was constituted by individuals
bound to one another by ties of kinship. The existence
of such groups stands in a relationship of interdepen-
dence with the existence of sentiments tending to make
the groups permanent…
Homans, shown in Figure 4.9, was born in 1910 to a wealthy
old-line Boston family. In fact, his mother was an Adams. Like
Lowell and Henderson, then, Homans was a Boston Brahmin—
and quite proud of it. At one point, according to John Barnes,
The Birth of Social Network Analysis II: The First Harvard Thrust
55
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Homans had been talking about American society with the prom-
inent Indian sociologist, M. N. Srinivas. After listening to
Homans’ views, Srinivas—who was a Brahmin—accused
Homans of presenting a Brahmin view of American society.
Homans thought for a moment, and then replied, “Yes, by God,
I’m a true Boston Brahmin.”
True to his Brahmin heritage, Homans had attended St.
Paul’s preparatory school and then enrolled at Harvard. He re-
ceived his bachelor’s degree in 1932. His field was poetry. But
his work with the Pareto seminar led to a change in direction.
In 1934 Homans was chosen as a Junior Fellow in the Har-
vard Society of Fellows and he remained in that position until
1939. Junior Fellows were expected to work with Senior Fellows
and Homans worked primarily with L. J. Henderson. And, at
Henderson’s urging, he also worked extensively with Elton
Mayo.
Figure 4.9. George Caspar Homans
56
Fortunately for Homans, Mayo insisted that he read some
social science beyond Pareto. Mayo introduced him to a good
deal of the literature in psychology as well as to the works of the
anthropologists Malinowski and Radcliffe-Brown (Homans,
1984, pp. 135–166). Moreover, Homans’ association with Hend-
erson and Mayo—both Business School faculty—led him to meet
Warner. Through that contact, Homans met and became friends
with two of Warner s students, Eliot Chapple and Conrad Arens-
berg. Both were influential on his thinking. Although Arensberg
introduced Homans to Moreno’s ideas (Homans, 1984, p. 163),
and this led Homans to cite Moreno favorably in his later work
(Homans, 1950, pp. 40–42), there is no indication that Moreno
had much effect on his intellectual development.
Finally, in 1939 Sorokin, who was the Chairman of the Soci-
ology Department at Harvard, hired Homans as an instructor.
Although he was never particularly close to Sorokin, Homans
speculated that he was hired because he had resisted the lure of
Parsons’ “fuzzy thought” and “sloppy writing” (Tilly, 1989). At
Harvard, Homans worked his way up from Instructor until he
was appointed Professor in 1953. In that year he visited Britain
and taught at Manchester University. In subsequent years he
went to Cambridge where he received his highest degree, an
M.A., in 1955. He returned to his job at Harvard in 1956.
While he was still a Fellow, Homans began to develop what
he (Homans, 1984, p. 164) called his “threefold classification.”
He proposed that interaction frequency, sentiment and joint ac-
tivity (first called “function”) are all interrelated. This idea was
first introduced in print in the last chapter of his book on thir-
teenth century English village life (Homans, 1941). There (p. 405)
he went on to suggest that:
…it is possible to observe that the action of one man is
a stimulus for the action of a second, and that this
action of the second is in turn a stimulus for the ac-
tion of the first. Or the action of the second becomes a
stimulus for the action of a third, and so forth. We
have seen that a society can be defined as any group of
people interacting in this way. Within a society so
defined, the chains of interaction are infinitely com-
The Birth of Social Network Analysis II: The First Harvard Thrust
57
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
plex and cover the society in a number of different
ways.
These remarks capture the essence of social network anal-
ysis. And his early ideas on the threefold classification provid-
ed a précis of his much more ambitious volume, The Human
Group (Homans, 1950). In that book he developed and refined
his classification and showed how it might help to explain some
structural phenomena.
Specifically, The Human Group focused on showing how
the structure of groups and the positions of individuals in those
groups might emerge from the interrelations among his three
variables. He examined structural data that others (including
his colleagues at Harvard, Davis, Gardner, and Gardner (1940),
Roethlisberger and Dickson (1939) and William Foote Whyte
(1943) had collected. Moreover, he reproduced images of graphs
and hypergraphs from Roethlisberger and Dickson. But, like
the Hawthorne group, the Yankee City group and the Deep
South group, Homans simply failed to develop or to draw upon
any mathematical or computational tools.
This failure might be surprising since his mentor, Hend-
erson, had insisted that Homans learn some mathematics dur-
ing his years as a junior fellow (Homans, 1984, p. 122).
Unfortunately, Henderson made Homans read the calculus
rather than exposing him to algebra or probability theory. The
result was that Homans’ mathematical training was entirely
inappropriate for his chosen work. The calculus could not pro-
vide any useful tools for refining his ideas about social struc-
ture.
William Foote Whyte and Conrad Arensberg were added
to the collection of Junior Fellows in 1936. By that time, Arens-
berg had a great deal of experience at Harvard, and he took the
new Harvard student, Whyte, under his wing. Whyte’s under-
graduate degree from Swarthmore was in economics. He had
planned to continue in economics at Harvard when he arrived
there, but Chapple and particularly Arensberg convinced him
that he should focus on studying social structure (Whyte, 1994,
p. 63; 1997, pp. 19–20). For that study, he chose a slum commu-
nity in Boston’s North End that he called “Cornerville.” He en-
58
tered that community as a participant observer and came out
with a classic book, Street Corner Society (1943).
Whyte’s book contains no systematic data, but it is rich in
ethnographic detail. Its aim was to detail the social structure of
the community by observing interaction patterns among its cit-
izens, and it provides a full description of that structure. It in-
cludes graphic images, like the one reproduced in Figure 4.10,
that are designed to depict structural arrangements. This work,
then, is clearly in line with the whole structural thrust at Har-
vard in the 1930s. Though it lacked systematic data and it did
not use mathematical tools, it did lay out the structural form of
the interaction patterns linking individuals.
The Birth of Social Network Analysis II: The First Harvard Thrust
Figure 4.10. Whyte’s image of the Nortons
59
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
By the mid-1930s, then, the efforts at Harvard had come
close, but had never included all four of the features that define
social network analysis. Structural intuitions dominated all of
this work. Two of the projects, the Western Electric research and
the Deep South study, produced systematic relational data. All
of it also drew on visual images. But none of the work described
so far developed or used any mathematical or computational
tools.
Chapple and Arensberg, however, moved to fill that gap.
They had returned from collecting data on the Yankee City project
somewhat disillusioned. As Homans (1984, p. 162) reported,
“What came to bother Chapple and later Arensberg was the lack
of rigor with which the results of social research could be made
and reported.” So they set themselves the task of providing a
framework for the systematic collection and analysis of struc-
tural data. As Whyte (1994, p. 63) described it:
Their framework was designed to focus on objectively
observable behavior, thus bypassing the subjectivity
inherent in methods that focused on the interpreta-
tion of what people say. Arensberg argued that, when-
ever the same individuals interact over a period of time,
a structure for that interaction emerges. That struc-
ture can be determined through direct observation, and
that structure will strongly influence what people say,
think, feel, and do.
They began to develop a rigorous operational definition of
the variable “interaction.” Chapple (Chapple, 1940) devised a
“recording typewriter” to collect carefully timed data on the
interaction pattern linking a pair of individuals. And although
he had agreed to work with Warner on the analyses of the Yan-
kee City data, he backed down. Warner s wife, Mildred Warner
(1988, p. 121), described Chapple’s withdrawal:
…he had developed a system of analysis inconsistent
with that used in the research, indeed, inconsistent
with the purposes for which the research material had
been gathered, and he felt that he could not meet his
commitment since his report would conflict with oth-
60
ers. His primary interest had become a system of quan-
titative analysis on which he and Conrad Arensberg
were working.
In any case, Chapple and Arensberg developed ways to
collect and analyze detailed data about social linkages. As a part
of that effort, they (Chapple, 1940, p. 79) faced a problem:
I order to describe a system of relations quantitative-
ly, it is necessary to deal with a large number of rela-
tions simultaneously. Otherwise we have to take each
individual in turn and work out the equations describ-
ing his relations, with no way of synthesizing the sys-
tem as a whole.
To solve that problem, they consulted with Harvard math-
ematician, Willard Quine. With Quine’s help they developed an
algebraic model generalized from kinship. They defined unit
relations (foreman to worker or father to daughter) and calcu-
lated relative products (foreman to daughter of worker) and their
inverses (woman to foreman in father’s department). They or-
ganized these concatenations into matrices in order to deal with
large sets of compound relations.
Quine, Chapple and Arensberg, then, provided the miss-
ing feature. Following their contribution, the Harvard group
displayed all of the features of contemporary social network
analysis. Unfortunately these efforts did not result in the emer-
gence of a recognized “school” of structural analysis. Indeed,
almost immediately after Chapple and Arensberg introduced
their formal model, the collective effort was abandoned.
Warner, Gardner, Davis and Drake all left for the Universi-
ty of Chicago in 1935. Following Arensberg’s advice to go study
with Warner, in 1940 Whyte followed them to Chicago, where
he completed his PhD. Gardner also completed his PhD at Chi-
cago. Then he started a market research firm and recruited Warn-
er as an active partner in that effort. Davis, Drake and, later,
Whyte also completed their PhDs at Chicago. Afterwards Davis
accepted an appointment in Chicago’s Department of Education.
Drake specialized in black studies and went to work first at
Roosevelt College and finally at Stanford. Whyte decided that
The Birth of Social Network Analysis II: The First Harvard Thrust
61
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
he would rather work with Everett C. Hughes in Chicago’s Soci-
ology Department than with Warner in Anthropology. He fol-
lowed Hughes’ lead and devoted the rest of his career to
participant observation in various settings. He wound up teach-
ing at Cornell. Chapple completed his PhD at Harvard and was
appointed to a faculty position in Sociology and in the Harvard
Business School. He soon quit, however, to become a consultant
to industry. Arensberg took a position at the Massachusetts In-
stitute of Technology in 1934 and from there went on to Colum-
bia in 1946 where he remained until his retirement.
Henderson died in 1942. Mayo retired in 1947. And, accord-
ing to Roethlisberger (1977, p. 308) the previously unified Har-
vard approach to the study “human relations” had, by then,
divided into two parts: (1) the “case method” associated with
Mayo and Roethlisberger’s own work, and (2) the “applied an-
thropology” approach associated with Chapple, Arensberg and
Whyte. Since Whyte took a very different tack than Chapple and
Arensberg, Roethlisberger’s grouping them all together is diffi-
cult to understand., Roethlisberger (1977, p. 57) was apparently
troubled by the rigor and formalism of Chapple and Arensberg’s
work:
Eliot’s “breakthrough” seemed to me to be a “break
with” a more pedestrian approach. The development
of the social sciences, it seemed to me, would require
for a long time just ordinary “tillers of the soil” (cli-
nicians and field workers), not fancy “new models.”
In this sense Eliot was a threat to the development of
the area as I thought it should take place at the Busi-
ness School.…So Eliot had me disturbed plenty; my
emerging career at the Business School, as I then en-
visaged it, was at stake.
At the same time, Whyte (1994, p. 63), whom Roethlisberg-
er had grouped together with Chapple and Arensberg, also ob-
jected to their approach:
Impressed as I was with the micromeasurements of
this methodology, I was not inclined to pursue this
62
type of research myself. I was more interested in field
research focusing on larger units of interaction.
So, once Warner was gone and Chapple and Arensberg were
cast out, the Harvard thrust was lost. Only Chapple seemed to
understand the full implications of the earlier collective effort. A
few years later he (Chapple, 1953, p. 304), put it succinctly, “We
can, in fact, use a modified form of the kind of network analysis
used in electrical work…and we can determine the effects of any
change in the quantitative values assigned to any link on its
neighbors in the network pattern.”
I conclude, then, that although the work at Harvard con-
tained all four of the features that define social network analy-
sis, it failed to generate enough long-range commitment to be
able to form, and to convince others that they had formed, a
general paradigm for research. The group disintegrated because
people—in particular, Warner—left and went on to other col-
laborations on other kinds of research. Moreover, another rea-
son the Harvard group failed to form an identifiable paradigm
for research probably stems from the fact that it developed in-
ternal conflicts. The other members of the Harvard group were
apparently not ready to move off in the new direction proposed
by Chapple and Arensberg: they were not prepared to embrace
the rigor in procedures for data collection and the formal alge-
braic modeling suggested by Chapple and Arensberg. In fact, in
the quotation above, Roethlisberger was troubled enough to
describe the Chapple-Arensberg initiative as “a threat.”
So the Harvard effort never “took off.” It never provided a
general model for a structural paradigm. As a matter of fact, the
efforts at Harvard are almost never recognized in historical re-
views of social network analysis.22 Of the Harvard group,
Homans is sometimes mentioned as an intellectual antecedent
and the studies by Roethlisberger and Dickson and by Davis,
Gardner and Gardner are sometimes credited with having gen-
erated early network data. But the organized collective work
centered in the Business College and the Society of Fellows has
The Birth of Social Network Analysis II: The First Harvard Thrust
22The one exception I can find is in Scott’s (2000) text
63
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
simply not been acknowledged as a central part of the history of
the field.
Following Moreno’s work and that at Harvard, social net-
work analysis seemed to pass into a period that might be called
“dark ages.” Chapters 5, 6 and 7 deal with this period. For a
span of thirty years, from about 1940 until about 1970, no major
centers of social network research emerged. However, as we shall
see in these three chapters, there were quite a number of small
efforts that kept the structural perspective alive.
64
Chapter 5
Social Network Analysis During the
Dark Ages I: the 1940s
By the 1940s, much of the excitement that Moreno and Jen-
nings had generated in the 1930s had already started to wane.
Moreover, the Harvard group had broken up and its members
had drifted away from structural analysis. So the period in ques-
tion was essentially a kind of “dark ages” for social network
analysis. There was no generally recognized approach to social
research that embodied the structural paradigm. Social network
analysis was still not identifiable either as a theoretical perspec-
tive or as an approach to data collection and analysis.
That fact, however, did not mean that no social network
research was done during the period in question. As we shall
see, a number of people at a number of universities kept the
structural perspective alive. None of these efforts, however, had
enough impact to be widely adopted as a general paradigm for
the structural approach.
Doubtless many of these people were influenced by the
earlier work at Harvard or, even more likely, by the early popu-
larity of the contributions by Moreno and Jennings. But at least
some of these latter day network scholars more or less indepen-
dently came up with the social network paradigm. In the next
several sections I will review a series of settings in which inves-
tigators working in the 1940s produced research that was essen-
tially network analytic.
65
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
5.1 A Mobile Group: From Iowa to MIT to Michi-
gan in the 1930s and
1940s
The next develop-
ment in research that
embodied the social net-
work perspective was
originated by a psychol-
ogist, Kurt Lewin.23
Lewin, shown in Figure
5.1, was born in Germa-
ny in 1890. He received
a PhD in experimental
psychology from the
University of Berlin in
1916. He served in the
German army during
World War I. After the
war he accepted a posi-
tion in the Psychological
Institute of the Berlin
University where he worked with the eminent Gestalt psychol-
ogists Max Wertheimer and Wolfgang Koehler. In 1933, like many
German Jews of that era, he fled the Nazi regime and came to
the United States.
Lewin’s first job in America was in the Home Economics
Department at Cornell. But, in 1935 he was hired by the Child
Welfare Research Station at the University of Iowa. When he ar-
rived at Iowa, Lewin’s research was focused on what he called
“field theory,” or “topological psychology.” It stressed the im-
portance of internal and external “forces” as they might impinge
on an individual’s behavior. Thus, his approach considered the
impact of situational factors in determining behavior (Lewin,
1936).
23He pronounced his name the German way, Leveen, with the accent on the second
syllable.
Figure 5.1. Kurt Lewin (before he
immigrated to America)
66
Among the situational factors that Lewin took into account
in understanding an individual’s behavior, were the effects of
the behaviors of other people (Patnoe, 1988, p. 5). Beginning at
the time he moved to Iowa, however, he began to shift his inter-
est from the study of individual behavior to the study of inter-
personal relations and group processes.
Above all, Lewin was a gifted teacher. For him, research
and teaching were wrapped together into a single integrated
package. The way he worked was by talking to other people, his
colleagues, his students—anyone and everyone. He made no
status claims and this contributed to his attraction as a teacher.
So, at Iowa, Lewin trained a whole generation of American so-
cial psychologists. All were imbued with a respect for empiri-
cal—particularly experimental—research. He was a charismatic
figure, and because of that, he managed to attract a great many
talented young psychologists as graduate and post-doctoral stu-
dents. They included Alex Bavelas, Dorwin (Doc) Cartwright,
Leon Festinger, John R. P. French, Jr., Ronald Lippett, Marian
Radke and Alvin Zander.
In 1945 Lewin left Iowa to form the Research Center for
Group Dynamics at MIT. He was joined by four of his former
students, Dorwin Cartwright, Leon Festinger, Ronald Lippitt and
Marian Radke. In addition, Lewin also brought along a fifth
young man who was still a graduate student, Alex Bavelas.
The moment this group arrived at MIT they began to at-
tract more students. The early recruits included Kurt Back, Mor-
ton Deutsch, Gordon Hearn, Harold Kelley, Albert Pepitone,
Stanley Schachter and John Thibaut.
This was an immensely productive period for the Research
Center for Group Dynamics. Patnoe (1988, p. 9) described just
how productive it was:
During the three years they were in residence at MIT,
research was designed and carried out on such issues
as leadership (Lippitt & French, 1948), group cohe-
siveness (Back, 1951; Thibaut, 1950), group produc-
tivity (French, 1950), the effects of group membership
on it’s members (Schachter, 1951), cooperation and
Social Network Analysis During the Dark Ages I: the 1940s
67
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
competition (Deutsch, 1949), intergroup relations
(Lippitt & Radke, 1946), communication and the
spread of influence within groups (Festinger, Schachter
& Back, 1950; Festinger & Thibaut, 1951) and social
perception (Kelley, 1950; Pepitone, 1950).
Amid all this productivity, disaster hit the program in 1947.
Lewin died suddenly and unexpectedly. He was replaced by one
of his former students, John (Jack) French, Jr.24 But President
Compton of MIT soon decided that, without Lewin, the Research
Center was no more than an expensive frill in an engineering
school, one that they could do without. So Cartwright and Fest-
inger set out to find another home.
Before and during the war, Cartwright had worked with
Rensis Likert, a sociologist at the University of Michigan. So,
through Likert’s intervention, the whole Research Center for
Group Dynamics was invited to move to the University of Mich-
igan. The move was made in 1948. Cartwright, Festinger, French
and Lippitt took with them several graduate students, includ-
ing Harold Kelley, Stanley Schachter and John Thibaut. And,
before leaving, both Cartwright and Festinger helped Lewin’s
graduate student Alex Bavelas to finish his degree and get start-
ed on his own research.
Cartwright took over for Lewin and directed Bavelas’ dis-
sertation. Bavelas completed his PhD at MIT in 1948 and was
immediately invited to join the faculty. He did, and he designed
a landmark study of the implications of social network struc-
ture.
When he took the job at MIT, Bavelas (shown in Figure 5.2)
was already armed with an important structural intuition (Bave-
las, 1948). He believed that in any organization the degree to
which a single individual dominates its communication net-
work—the degree to which it was centralized—affected its effi-
ciency, its morale and the perceived influence of each individual
actor. Since the Research Center for Group Dynamics was leav-
ing for Michigan, Bavelas created a new—presumably less ex-
24French had received his PhD from Harvard, but he too had been trained by Lewin
during Lewin’s visiting appointments at Harvard in l938 and 1939.
68
pensive—center, the Group
Networks Laboratory, as a part
of MIT’s established Lincoln
Laboratory.
Festinger helped in the de-
velopment of Bavelas’ new lab
by introducing him to R. Dun-
can Luce. Luce recalled the
event (1978, p.247):
The actual start of my ca-
reer in psychology was, in
a sense, sharply defined.
One afternoon, Albert Per-
ry, a graduate student in
electrical engineering at
MIT, and I were modifying
a military surplus radio into what then passed for high-
fidelity equipment, when my roommate William
Blitzer returned from Leon Festinger’s class in social
psychology. He described to us some of the combina-
torial problems they faced in dealing with social net-
works. Soon Perry and I were busy trying to translate
these into questions about matrices, and a few days
later Blitzer introduced us, with some theorems in
hand, to Festinger. By the end of the summer we had a
paper ready for submission, and another paper on the
same topic followed shortly.
The paper that was “ready for submission” was the classic pa-
per—one of the most important in social network analysis—in
which Luce and Perry (1949) formally defined the notion of
“clique.”
At that time Luce was working on his PhD in mathematics
at MIT and was looking for a job. Festinger introduced Luce to
Bavelas and Luce was hired as “Bavelas’ captive mathematician”
(Luce, 1978, p. 248). Bavelas put Luce to work, along with a col-
lection of social science students. The job was to work out a way
to study the consequences of communication structure. The crew
Figure 5.2. Alex Bavelas
Social Network Analysis During the Dark Ages I: the 1940s
69
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
included Bavelas and Luce, along with Robert P. Abelson, Harold
Leavitt, Lois Rogge, Arnold Simmel and Sidney Smith. Under
Bavelas’ leadership, they developed a formal model, drew graph
theoretic images of social structures, designed an experiment,
and collected experimental data on efficiency, morale and the
recognition of leadership. Their work indicated that Bavelas’
original intuition had been correct in every respect (Bavelas, 1950;
Leavitt, 1951; Smith, 1950).
The work of the MIT group, then, displayed all the elements
that are found in contemporary social network analysis. It em-
bodied a structural intuition. It involved the collection of sys-
tematic experimental data. It used the graphs shown in Figure
5.5 to represent the patterns of communication that were stud-
ied. And the Bavelas group developed a formal model for their
main independent variable, centrality.
At the time, Bavelas’ research generated widespread inter-
est. People from several fields, including Claude Flament (1956)
from psychology, Harold Guetzkow (Guetzkow and Dill, 1957)
from political science and the Nobel Laureate Herbert Simon
(Guetzkow and Simon, 1955) from economics wrote follow-ups
extending the Bavelas approach. But that interest quickly waned.
Bavelas left MIT in 1950 to work on a State Department project
and turned the research laboratory over to Lee S. Christie and
Duncan Luce. Bavelas worked in industry for years and never
returned to his experimental structural analyses. Luce and Sim-
mel moved to Columbia. So the impetus at MIT dried up. Oth-
Figure 5.3. Experimental communication forms studied by the
Bavelas group
70
ers did pick up parts of the MIT idea but very few really got the
whole picture. As I summarized the problem (Freeman et al.,
1980), “…subsequent investigators tended to be concerned ei-
ther with conceptual problems of centrality or with the conse-
quences of communication structure on problem solving, not
both.” This meant that the integration developed at MIT was
lost. Each of the newer investigators focused on part of the total
social network picture, but lost its generality.
The MIT part of the story, then, was over. But the same lev-
els of energy and creativity were kept alive at Michigan. The
structuralists at Michigan continued to be impressively produc-
tive. Their work maintained the Lewin style; it was usually ex-
perimental and it continued to involve both faculty and graduate
students.
Once they were established at Michigan, Cartwright and
Festinger both recognized a need for a mathematical collabora-
tor—one who could help them explore the potential of “new
mathematical techniques” in behavioral sciences. They received
a grant from the Rockefeller Foundation, and Cartwright con-
tacted the Chair of Michigan’s Mathematics Department to seek
a collaborator. The Chair thought the idea of mathematical so-
cial or behavioral science was crazy, but he did provide a name,
Frank Harary, a new PhD teaching in the Department who had
not yet chosen his area of specialization.
Since he had some mathematical training (and perhaps be-
cause he had provided the link between Bavelas and the mathe-
matician, Luce, at MIT), Festinger made the first contact with
Harary. They sat down together and Festinger drew the graph
of the simple social structure shown in Figure 5.4. Harary re-
ports that the moment he saw Festinger’s graph his whole ca-
reer appeared before him like a vision; he began a long and
distinguished career focusing for the most part on applications
of graph theory to problems of social structure (Hage and Harary,
1983; Harary, 1953, 1955; Harary and Norman, 1953; Harary and
Ross, 1957).
At first Harary worked with Festinger, but when Festinger
moved to Minnesota in 1951 and was no longer available, Harary
and Cartwright teamed up. Their earliest major joint effort drew
on the theory of signed graphs. They developed a formal state-
Social Network Analysis During the Dark Ages I: the 1940s
71
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
ment of the notion of cognitive balance proposed by the psy-
chologist Heider (Cartwright and Harary, 1956). And, from that
start, a career-long collaboration was born.
Various subsets of the Michigan group worked on inter-
personal influence and on rumor transmission (Back, 1951; Back,
Festinger, Hymovitch, Kelley, Schachter and Thibaut, 1950; Fes-
tinger, Schachter and Back, 1950). In addition, in the mid 1950s
Newcomb (1961) and Nordlie (1958) began collecting data for
their path-breaking multi-year study of interpersonal attraction
in a university residence hall.
There is a legitimate question, however, about the indepen-
dence of all of this Lewin-inspired work from Moreno’s sociom-
etry. On several occasions in 1935 Moreno had met with Lewin
and several of his students (Morrow, 1947). After those meet-
ings, according to Moreno (1953, p. lxiv), Lewin exhibited a
“…change in focus and attention from individual and Gestalt
psychology to the consideration of group and action methods.”
Moreno argued, then, that the whole impetus for the MIT Re-
search Center for Group Dynamics was built on his ideas.
Given Moreno’s tendency toward paranoia, it would be easy
to dismiss this argument as a paranoid fantasy. Renshaw (1981),
for example, argues that Lewin did shift focus, but that his shift
was the result of his association with established research prac-
tices at the University of Iowa’s Child Welfare Institute:
25Harary wears a picture of this graph on his cap to this day.
Figure 5.4. The Festinger graph25
72
Lewin’s innovative research on group climates (Lewin,
Lippit and White, 1939) was foreshadowed at Iowa in
the research by Jack (1934), Page (1936), and H.
Anderson (1937) on children’s style of
interaction.…Thus, while Lewin’s contribution should
be acknowledged as a new and creative one, it appears
that his going to Iowa in 1936 and reading the re-
search already completed there on styles of peer inter-
action was a significant influence on the development
of his ideas.
In addition, Patnoe (1988, p. 31) quotes Cartwright about how
Lippit, one of Lewin’s early students at Iowa, influenced his
mentor’s thinking:
This impact of Lippitt’s cannot be overemphasized. His
role was very important. Lippitt had a tremendous
influence on Lewin to get him interested in groups
specifically.
On the other hand, Wech (1996), who is a Lewin supporter,
made a careful examination of all the relevant data. Her conclu-
sion was that the meetings with Moreno did affect Lewin’s work.
And Lewin, moreover, was an early publisher in Moreno’s jour-
nal (Lewin and Lippitt, 1938).
We end up, then, with no sure answer. Lewin’s perspective
did change direction when he went to Iowa. But whether that
was due to the influence of researchers already working at Iowa
or the result of the influence of Moreno is not clear.
To my mind, the question of influence here is beside the
point. The Iowa-MIT-Michigan group went far beyond the so-
ciometry of Moreno. The two research traditions shared the in-
tuitive notion that people’s social ties to others have important
consequences for their lives. But beyond that intuitive founda-
tion, the two went off in different directions. Moreno’s work was
based primarily on subjective responses to questionnaires. Lewin
and his students manipulated social ties and collected experi-
mental data about their consequences. And while mathematics
was peripheral to most of Moreno’s research, mathematics was
Social Network Analysis During the Dark Ages I: the 1940s
73
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
at the core of the work at MIT and Michigan. Finally, while both
projects used graphics, Bavelas, Luce, Festinger, Cartwright and
Harary used them explicitly in the context of the formalisms of
graph theory. The differences, I think, are large enough to sug-
gest that Lewin’s research—and particularly that of his stu-
dents—was a good deal more than simply derivative.
All in all, the Iowa-MIT-Michigan years were a highly pro-
ductive period for an impressive collection of research scien-
tists. They worked on a wide range of structural problems and
collected an immense amount of systematic relational data. They
used graph theoretic images regularly. They developed formal
models. And it is clear from their easy transition from applica-
tion to application that they recognized the generality of their
social network approach. This recognition was explicitly made
years later in the publication Structural Models: An Introduction
to the Theory of Directed Graphs (Harary, Norman and Cartwright,
1965). That book set out a generalized abstract representation of
social networks in precise terms.
Thus, the Iowa-MIT-Michigan axis succeeded in produc-
ing a huge amount of important theory and data over an ex-
tended period. Their structural approach was generalized to a
wide range of empirical phenomena. It had a profound impact
on research in social psychology. Moreover, in retrospect, their
work is generally recognized as an important influence in the
genesis of contemporary social network research.
In a very muddled statement, Mullins and Mullins (1973)
characterized this Lewin-led effort as completely unrelated to
the study of social networks. They defined the Lewineans as fo-
cused exclusively on the study of small groups, and they listed a
rag-tag collection of any others who had ever talked about small
groups as members of Lewin’s “theory group.” The list includ-
ed, for example, Elliot Chapple, George Homans and William
Foote Whyte, as well as Jacob Moreno and Helen Jennings,
among the members of the Lewin group.26 Chapple, Homans
and Whyte were, as we saw, associated with Warner, not Lewin,
26Most remarkably, they consistently referred to Jacob Moreno as “José Moreno.”
74
and Moreno and Jennings were at the core of the sociometric
approach. At the same time, the Mullins and Mullins list did not
include Harrison White or any of his students. So, in effect, they
grouped Warner’s people, Moreno’s people and Lewin’s people
together into a single clump, but they saw no similarities be-
tween the research efforts of any of these people and the struc-
tural approach of Harrison White.27
In any case, Mullins and Mullins concluded that the Lewin
team had failed. According to their analysis, it failed because
after Lewin’s death, it lacked both intellectual and organization-
al leadership and was unable to attract students. But when we
consider the major intellectual and organizational contributions
of Bavelas, Cartwright, Festinger and Newcomb, this argument
falls apart. As a freshman faculty member at MIT, Bavelas de-
signed a breakthrough experiment, organized a new research
laboratory and recruited students like Robert P. Abelson, Harold
Leavitt and Arnold Simmel, all of whom went on to stellar ca-
reers. Cartwright organized the movement of Lewin’s laborato-
ry from MIT to Michigan and went on to be one of the authors of
a book that is still the most authoritative text on the application
of graph theory to social network analysis (Harary, Norman and
Cartwright, 1965). Festinger is often described as “the father of
experimental social psychology” (Patnoe, p. 255), and Newcomb
(1961) conducted a study of the acquaintanceship process that is
still cited by people working in social network analysis more
than forty years later. This group, moreover, trained a whole
generation of productive social psychologists, including such
notables as Elliot Aronson, Kurt W. Back, Judson Mills, Stanley
Schachter and John W. Thibaut.
The Lewinean perspective came to dominate the field of
social psychology, and the people involved are still cited regu-
larly by network analysts. But it failed to be picked up as a gen-
eral paradigm for other social research disciplines of the time.
27Nicholas C. Mullins was one of White’s social network students.
Social Network Analysis During the Dark Ages I: the 1940s
75
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
5.2 Michigan State College in the mid 1940s
In Chapter 3, I suggested that most of Moreno’s early sup-
porters quickly drifted away. But a few did keep the sociometric
faith. Notable among these was a rural sociologist, Charles P.
Loomis. Loomis had grown up in a farming family in Las Crus-
es, New Mexico. After he received a bachelor’s degree in 1928
from the New Mexico College of Agriculture and Mechanical
Arts, he went on for a master’s at North Carolina State College.
Then, as a contemporary of Warner’s students, Loomis went to
Harvard for his PhD and received a degree in sociology in 1933.
Loomis was in graduate school, then, during the time when
Warner was introducing the structural perspective at Harvard.
There is no evidence, however, that Loomis had any signif-
icant contact with Warner or his colleagues and students at that
time. Warner s efforts were concentrated in the Business School
and the Anthropology Department and Loomis apparently
worked entirely in the Sociology and Economics Departments.
In particular, he worked with the Chairman of Sociology, Pitirim
Sorokin, who, as we saw in Chapter 3, had very little to do with
the structuralists centered in the Business School. Loomis’s doc-
toral thesis, moreover, was a fairly standard demographic study
of North Carolina farmers.
After receiving his degree, Loomis was hired into a research
position at the United States Department of Agriculture. There
he began using sociometric tools to do research in rural commu-
nities (Loomis and Davidson, 1939). Then, in 1944, he accepted a
position as Head of the Sociology and Anthropology Depart-
ment at Michigan State College.
At Michigan State, Loomis trained a large number of grad-
uate students in the use of sociometric research tools (Driscoll et
al., 1993). With their help, he conducted a series of comparative
studies of small villages and rural areas throughout the world.
(Examples are Loomis, 1946; Holland and Loomis, 1948; Loomis
and Powell, 1949; Loomis and Proctor, 1950.)28
28The Holland cited here was the father of the prominent social network statistician,
Paul W. Holland. John B. Holland received his PhD at Michigan State with Loomis,
then took his family and went to work on a Loomis project in Cuba where he died of a
sudden and unexpected heart attack.
76
At the same time, Loomis recognized the importance of
mathematics in structural research and sought out colleagues in
mathematics to help him deal with the complexities of network
analysis. Prominent among those was the mathematical statisti-
cian Leo Katz, who had received his PhD from the University of
Michigan in 1945. Reportedly, Loomis showed Katz some of the
statistical work in sociometry and Katz was intrigued, because
until then, the field had developed only very primitive models.
According to Katz’s student Charles Proctor,29 “Katz jumped at
the chance to clean things up and to advance them.” Proctor
went on to add that, as a recent PhD from the University of Mich-
igan, Katz was motivated further by a spirit of “…competition
with theorists at U of M…such as Leon Festinger and Frank
Harary.”
Over the next few years, then, Katz produced a series of
papers that made major contributions to sociometry and, in the
long run, to social network analysis (Forsyth and Katz, 1946; Katz,
1947; Katz and Powell, 1955; Bhargava and Katz, 1963). In addi-
tion Katz led several students in mathematics to work on ap-
plied problems in structural analysis. Among these, Charles
Proctor and T. N. Bhargava made important contributions.
Thus, the Loomis/Katz group at Michigan State helped keep
sociometry alive. And they contributed to the development of
social network analysis in two ways. Led by Loomis, they con-
ducted high quality research on rural areas and small villages.
Those studies are still cited by network analysts. And, led by
Katz, they developed an important collection of new probabili-
ty-based formal models. These projects both extended the so-
ciometric approach, but apparently their efforts were not enough
to re-kindle widespread interest in sociometry.
5.3 The University of Chicago and the Sorbonne in
the late 1940s
Another early effort that involved a structural perspective
also emerged in the 1940s. It was generated by Claude Lévi-
29Personal communication.
Social Network Analysis During the Dark Ages I: the 1940s
77
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Strauss and drew on the mathematical skills of André Weil.30
Lévi-Strauss is shown in Figure 5.5.
Lévi-Strauss was born in 1908 in Belgium to French par-
ents. His family returned to France when he was a year old and
so he grew up in Paris. He was an outstanding student and was
admitted to the École Normale Superieure that was, as its name
implies, a very good university. Lévi-Strauss first studied law,
then switched and received his degree in philosophy. His fellow
students included Simone Weil and Jean-Paul Sartre, both of
whom later became famous philosophers.
After receiving his degree Lévi-Strauss got a job teaching
philosophy. In 1933 he began reading ethnography and in it
found his natural intellectual home. So in 1935, he accepted a
professorship at the University of São Paulo in Brazil. His pres-
ence in South America permitted him to follow his interest in
ethnography by doing fieldwork among the Caduveo and Bororo
Indian tribes of the Amazon basin.
Figure 5.5. Claude Lévi-Strauss
30He pronounced his name “Vay” in the French manner.
78
He returned to France in 1939, but the war drove him out
again. In 1941 he took a job teaching at the New School for
Social Research in New York. While he was teaching at the
New School, Lévi-Strauss was trying to make sense of the wide
range of differing kinship patterns. The rules of who was con-
sidered to be related to whom and who could marry whom in
various societies seemed, at first, to be arbitrary. But then one
of his colleagues, a linguist named Roman Jakobson, introduced
him to the work of Ferdinand de Saussure. Saussure et al. (1916)
had studied the structure of languages. He proposed that
though the rules of grammar vary from language to language,
the words are always related to each other according to some
system of rules.
Lévi-Strauss thereafter set out to uncover the systems of
rules that various peoples used to determine kinship. When
the war was over, he returned to France and in 1948 defended
his PhD dissertation at the University of Paris. In his disserta-
tion he took a structural approach to the study of kinship. A
year later he turned that dissertation into a major book, Les
Structures élémentaires de la Parenté [The Elementary Structures
of Kinship] (1949/1969). In that book, Lévi-Strauss examined
“…preferential marriage based on definite kinship relations.”
(Josselin de Jong, 1970, p. 2). He concluded that all preferen-
tial marriages are based on some form of exchange (Lévi-
Strauss, 1949/1969, pp. 478–479). Such exchange creates
alliances (Hénaff, 1998, p. 89). And alliances are essential to
ensure “…the integration of partial units within the total
group.” (Lévi-Strauss, 1949/1969, p. 481). They are “the only
way of maintaining the group as a group…” (Lévi-Strauss,
1949/1969, p. 479).
It is clear, then, that Lévi-Strauss took a general structur-
al approach to the study of kinship and marriage. More than
that, he based his conclusions on his analysis of dozens of data
sets from societies throughout the world. He used visual im-
ages consistently, including genealogical charts and directed
graphs like the one shown in Figure 5.6. And he prevailed on
André Weil to produce an algebraic appendix for his book.
Social Network Analysis During the Dark Ages I: the 1940s
79
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
André Weil was the brother of the philosopher, Simone Weil,
who had attended school with Lévi-Strauss. Weil was born in
Paris in 1906. He had always been intrigued by mathematics,
and he received his Ph.D. in that subject in 1928 from the Uni-
versity of Paris. He taught at various foreign universities, but
finally he obtained a regular appointment at the University of
Strasbourg in France in 1933. There, he was the principal force
behind the publications of Nicholas Bourbaki. An economist,
Steven E. Landsburg (http://www.netacc.net/~fairplay/weil.htm),
described how this happened:
In 1934, Bourbaki sprang full-blown from the head of
Andre Weil. Weil was teaching at Strasbourg and
engaged in endless discussions with his colleague
Henri Cartan about the “right” way to present vari-
ous mathematical concepts to students. It occurred to
him these discussions were probably being duplicated
by his friends in other universities all over France.
Weil proposed that they all meet to settle these ques-
tions once and for all. “Little did I know,” wrote Weil,
“that at that moment Bourbaki was born.”
Bourbaki, then, was a pseudonym used by a collection of young
French mathematicians. They worked together, and together
made a series of major contributions to the foundations of math-
ematics.
In 1947 Weil accepted an appointment at the University of
Chicago, where he remained until 1958. At that point he left
Figure 5.6. Directed graphs from Lévi-Strauss
80
Chicago, bound for Princeton, New Jersey, where he joined the
faculty of the Institute for Advanced Study. Throughout his ca-
reer he continued to make major contributions to mathematics.
Arguably, Weil was the French mathematician of his generation.
Lévi-Strauss provided both the intuitive background and
the data for Weil’s algebraic modeling. Weil, in turn, developed
a model for one of the most complicated kinship systems, the
Australian Murngin, who had been studied by W. Lloyd Warn-
er. Lévi-Strauss described his basis for asking Weil to build a
model (1960):
I fail to see why an algebraic treatment of, let us say,
symbols for marriage rules, could not teach us, when
aptly manipulated, something about the way a given
marriage system actually works, and bring out prop-
erties not immediately apparent at the empirical level.
Both Lévi-Strauss and Weil used graphic images in their
treatment of kinship. Jointly they produced a work that includ-
ed all of the properties of social network analysis. Although they
did draw on the earlier work of Radcliff-Brown, there is no rea-
son to believe that they were even aware of Moreno’s work or
the efforts of the Harvard group. My guess is that theirs was
largely an independent development of a social network per-
spective.
Despite the fact that Lévi-Strauss and Weil provided a gen-
eral model for social network analysis, it never seemed to cap-
ture the imagination of people working in other areas of social
research. Subsequent follow-ups (Kemeny, Snell and Thomp-
son, 1957; White, 1963; Courrège, 1965; Boyd, 1969; Ballonoff,
1974) all extended and refined the Lévi-Strauss-Weil models,
but their extensions were confined to the analysis of kinship
data. Lévi-Strauss himself went on to study mythology, and
dropped his interest in social behavior.
Social Network Analysis During the Dark Ages I: the 1940s
81
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Chapter 6
Social Network Analysis During the
Dark Ages II: the 1950s
6.1 Lund University in the early 1950s
Early in the 1950s an effort was started at Lund University
in Sweden. It was led by a Swedish geographer, Torsten Häger-
strand. Hägerstrand grew up living above his schoolteacher fa-
ther’s schoolhouse in a rural area called Småland. He entered
the University of Lund in 1937 where he studied both geogra-
phy and art history. In 1938 he was appointed as a librarian and
teaching assistant, but he gradually drifted into geography as a
full-time study.
Hägerstrand and his future wife decided to trace the lives
of every person who had lived in the Swedish rural area of Asby
from 1840 to 1940. All in all, they recorded the lives and migra-
tion patterns of more than 10,000 people.
But Hägerstrand was uncomfortable with the purely de-
scriptive, archival approach of most geographers of the time.
He read widely in other fields and was particularly impressed
by the astrophysicist, Sir Arthur Eddington; the sociologist,
George Lundberg and the mathematician-biologist, A. J. Lotka.
He also read and was influenced by the works of Jacob Moreno
and Kurt Lewin. So he set out to do theory-based structural work
in geography.
82
His theory-based research got started when his old friend,
the Swedish mathematician/computer scientist Carl-Erik
Fröberg, returned from a year of work in the U.S. In his travels
Fröberg had discovered that some American mathematicians
were working with random numbers. He returned to Sweden
with a small pamphlet that described a 1949 symposium on the
Monte Carlo Method. Fröberg presented this idea in a colloquium
at Lund, and reportedly, most of his colleagues ridiculed it. But
the presentation did capture Hägerstrand’s imagination and he
set out to do a computer simulation of the spread of innovations
across space and time.
By means of a time-dated sequence of maps, Hägerstrand
was able to show that innovations spread from their points of
origin in a wave-like pattern. His research objective, then, was
to identify the process underlying that pattern. He guessed that
a process of pair-wise communication—from adopter to adopt-
er—could account for the kind of patterning he had observed.
So he set himself the task of building a computer model to de-
termine whether those patterns could be produced by that kind
of person-to-person diffusion process. His plan was to use ran-
dom numbers to develop a Monte Carlo simulation of his as-
sumed pair-wise diffusion process. Interestingly enough, Lund
did not yet have a computer, so in his first attempts he had to
emulate a computer simulation using only a desk calculator.
Hägerstrand assumed that a person was more likely to pass
information to another who was geographically close than to
one who was distant. But he needed a way to estimate the likeli-
hood of contact at various distances. To do that, he used existing
data on the distances between the addresses of pairs of people
who married in the region he was studying. The marriage data
did show that there was a decrease in the probability of mar-
riage with increasing geographic distance between the address-
es of the bride and groom. He used these marriage figures to
estimate the probabilities of interpersonal contacts at various
distances. In addition, he used demographic data on the distri-
bution of the population to specify the number of potential tar-
gets for diffusion at each location in the region he was studying.
Social Network Analysis During the Dark Ages II: the 1950s
83
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
His simulation began by specifying an innovation along
with a map of some region for which he had sequential data on
the spread of that innovation through space. He located the ac-
tual point of origin and used that to begin his simulation of spa-
tial diffusion. The simulation embodied a discrete time
process—a series of cycles. At each cycle, each established adopt-
er passed the innovation on to one new target person. The target
was randomly chosen, but choices were biased by the distance-
decay statistics produced from the marriage data. The simula-
tion, then, generated a sequence of maps of the spread of an
innovation through space and over time. And the maps produced
by simulation showed the same wave-like patterning displayed
by the empirical maps (Hägerstrand, 1952).
Clearly, this work was motivated by an important structur-
al intuition—the pair-wise passage of information through a
population. Hägerstrand created images of the actual and the
simulated spread of various innovations. They were so similar
that his computer simulation of a computational model lent cre-
dence to his pair-wise diffusion hypothesis.
Hägerstrand’s work was structural. It displayed all the fea-
tures of social network analysis. He provided a model for a ge-
ography that explored theoretical issues and attempted to explain
the distributions of objects in physical space. That approach had
a tremendous impact on the field of geography and led a whole
generation of geographers to do similar kinds of structural work.
They include Brian J. L. Berry (Garrison, Berry, Marble, Morrill
and Nystuen, 1959; Berry, 1964), Lawrence A. Brown (1981),
Michael Dacey (1964), William Garrison (1960), Peter R. Gould
(Gould and White, 1974), Duane F. Marble (Garrison, Berry,
Marble, Morrill and Nystuen, 1959), Richard Morrill (Garrison,
Berry, Marble, Morrill and Nystuen, 1959), Forrest R. Pitts (1965,
1979) and Waldo R. Tobler (1965). Unfortunately, the impact of
Hägerstrand’s structural research was pretty much confined to
the field of geography. Apparently, the links between social ge-
ographers and other social scientists were too weak to encour-
age adoption of this approach as a general model for structural
research.
84
6.2 The University of Chicago in the early 1950s
Another center of structural studies was established at the
University of Chicago in the 1950s. At that time, Nicolas Ra-
shevsky was the Chair of the Committee on Mathematical Biol-
ogy at Chicago. Rashevsky was born in Russia in 1899 and
educated as a mathematical physicist. He came to the United
States in 1924 and began working at the Westinghouse Electric
and Manufacturing Company and teaching as an Instructor in
Physics at the University of Pittsburgh. While teaching at Pitts-
burgh, he came up with the radical notion of building a field of
mathematical biophysics patterned after mathematical physics.
Meanwhile, in 1929, at the age of 30, Robert Maynard Hutch-
ins left the Yale Law School to become President of the Universi-
ty of Chicago. Hutchins was an innovator who objected to the
rigidity and specialization he saw in higher education. As a col-
lege president, then, he sought out innovative academic pro-
grams. Thus, Hutchins hired Rashevsky in 1935.
Chicago’s flexibility under Hutchins made it easy for Ra-
shevsky to develop a Committee on Mathematical Biophysics
(later renamed the Committee on Mathematical Biology). The
Committee—essentially only Rashevsky at the beginning—de-
veloped a doctoral program that drew support from the Rock-
efeller Foundation. Rashevsky regularly taught a seminar that
was attended by both faculty and students. He founded a jour-
nal, the Bulletin of Mathematical Biophysics. And, from the begin-
ning he succeeded in attracting notable associates.
Perhaps the most noteworthy of Rashevsky’s early students
was Walter Pitts, who is generally recognized as the founder of
neural net modeling. Writing in the Encyclopedia of the Cogni-
tive Sciences at MIT, Lettvin (n.d.) described Pitts as follows:
Pitts appeared as a penniless 14-year-old at the Uni-
versity of Chicago in 1937, attended various classes,
though unregistered, and was accepted by Rashevsky’s
coterie as a very talented but mysterious junior. All
that was known of him was that he came from De-
troit, and that would be all that was known
thereafter.…In 1938 he appeared at the office of Ru-
Social Network Analysis During the Dark Ages II: the 1950s
85
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
dolf Carnap, whose most recent book on logic had ap-
peared the previous year. Without introducing him-
self, Pitts laid out his copy opened to a section
annotated marginally, and proceeded to make critical
comments on the material. Carnap, after initial shock,
defended his work and engaged with Pitts in an hour
or so of talk. Pitts then left with his copy. For several
weeks, Carnap hunted through the university for “that
newsboy who understood logic,” finally located him,
and found a job for him, for Pitts had no funds and
lived only on what he could earn from ghosting pa-
pers for other students.
A great deal of the early work of the Committee was fo-
cused on modeling neural connections, but by the 1950s most of
Rashevsky’s own work had shifted to applications of mathemat-
ics to sociological issues. This focus emerged because he (Ra-
shevsky, 1949) believed that:
…a logical extension of the mathematical biology of
behavior leads into mathematical sociology. The envi-
ronmental parameters, which determine some of our
reactions, are themselves determined by the reactions
of other individuals. Thus we are led from the “prob-
lem of one individual” to the more general “problem
of n individuals,” and thence into the domain of the
social sciences.
At that time, the faculty in Rashevsky’s Committee includ-
ed Herbert D. Landahl, Alfonso Shimbel, Hyman G. Landau and
Anatol Rapoport. And their students included Ray J. Solomonoff
and Lionel I. Rebhun. Solomonoff went on to a distinguished
career in algorithmic probability, and Rebhun became a promi-
nent biologist. Figure 6.1 shows the attendance at a session of
Rashevsky’s seminar in 1951.
Both the faculty and the students followed Rashevsky’s lead
and developed models of social behavior. In the early 1950s Rap-
oport (Rapoport, 1953, 1954, 1957; Rapoport and Rebhun, 1952;
Solomonoff and Rapoport, 1951) followed up Rashevsky’s (1951a,
1951b) work on the interpersonal process underlying the diffu-
86
sion of information. Solomonoff and Rapoport (1951), Landau
(1952), Landau and Rapoport (1953) and Landahl (1953a, 1953b)
all reported work on the same problem. At the same time, first
Rapoport (1949a, 1949b) and then Landau (1951a, 1951b, 1953)
worked on developing formal models of dominance hierarchies.
Landau’s efforts were so successful that his 1953 paper is still
cited regularly by mathematicians who work in the area of “score
sequences.”
Rapoport was a key participant in all of this work. He was
born in Russia in 1911. He came to the United States in 1922 and
became a naturalized citizen in 1928. After high school, he spent
two years as a student of piano in Chicago. Then, in 1929, he
went to Europe and studied at the State Academy of Music in
Vienna. He became a concert pianist and from 1933 through 1937
he appeared regularly on the concert stage in Europe, the Unit-
ed States and Mexico.
He then returned to school as a student at the University of
Chicago. Because he had heard that musical talent and mathe-
matical aptitude were highly correlated, he chose to study math-
ematics. As part of that pursuit he introduced himself to
Figure 6.1. The Rashevsky seminar (Rashevsky is seated in the front
row nearest the camera. Rapoport is next, followed by Shimbel and
Landau. The lecturer is Landahl).
Social Network Analysis During the Dark Ages II: the 1950s
87
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
Rashevsky and began to sit in on his seminars. In any case, his
studies went well and he received his PhD in mathematics on
December 5, 1941. After the Pearl Harbor attack two days later,
Rapoport immediately joined the U. S. Air Corps and served as
an officer during World War II. After the war he got a job teach-
ing mathematics at the Illinois Institute of Technology, but Ra-
shevsky soon hired him to teach mathematics and biology at the
University of Chicago.
Landau and Landahl were also major contributors. Landahl
was born in the United States and graduated from St. Olaf Col-
lege in Northfield, Minnesota, in 1934. He was awarded a PhD
in Mathematics at the University of Chicago in 1941 and he joined
the Committee on Mathematical Biology as a faculty member in
1942. Landau was born in Poland in 1909 and came to the Unit-
ed States as a child. He was trained as a mathematician, first at
Carnegie Tech and finally at the University of Pittsburgh where
he received his PhD in 1946. He interrupted his studies during
the war, and worked for the government on problems of ballis-
tics. After the war he received his PhD and was hired by Ra-
shevsky.
The members of the Committee—both faculty and stu-
dents—were all essentially applied mathematicians. They fo-
cused almost exclusively on structural phenomena and
speculated about the processes that might generate them. They
produced both mathematical and computational models deal-
ing with the patterning of links connecting humans or other an-
imals. And although their use of graphic images was rare, they
were sensitive to their importance. In discussing a dominance
matrix, Landau (1951a), for example, said, “a geometric (topo-
logical) description of the structure can also be given by n points
with lines connecting every pair of these points and a direction
assigned to every line.” Thus, he described a directed graph.
As mathematicians, then, members of the Committee fo-
cused on constructing models of social structure. And they cer-
tainly were aware of the generality of their structural models.
Rashevsky (1968) made that point clear:
…history may be considered as the result of interac-
tions between individuals…what a certain person
88
wrote, said, or did is largely, perhaps even completely,
determined by the effects on his life produced by his
fellow men…
But in addition to their mathematical modeling, these Com-
mittee members regularly drew on a variety of data sets. In one
article, for example, Landau discussed his model in terms of
Warder Clyde Allee’s (1938) data on hens. And in the next, he
talked about the same model in the light of Ruth Benedict’s (1934)
data on the Zuñi.
So it is clear that Rashevsky, Rapoport, Landau, Landahl
and the rest of the Chicago mathematical biologists developed a
generalized social network perspective. And given the fact that
they were applied mathematicians, it is also likely that their work
was largely independent of the work of Moreno and the other
earlier social scientists.
Unfortunately, the whole effort at Chicago dried up just as
it was getting established. Rapoport (2000, p. 106) described what
happened:
Financial support for the University of Chicago (sus-
pected of harbouring subversion) was drying up. To
revive it, some assurance of loyalty to “American val-
ues” had to be shown. That meant that Hutchins had
to go. A skilled money raiser unstained by either “in-
tellectual elitism” or softness on Communism replaced
him. The U.S. senatorial Jenner Committee (one of
the committees investigating “un-American activi-
ties” on campuses) was welcomed at the University
of Chicago.
A group formed including people who were likely to
be called to the hearings.…I was not called. Three of
my colleagues, members of the Committee on Mathe-
matical Biology, were called…Another colleague,
(Landau) who had done research in ballistics during
the war was asked whether he was engaged in espio-
nage at that time. He turned to an attorney and asked
him whether if he answered that question in the nega-
tive, he would be obliged to answer other questions.
Social Network Analysis During the Dark Ages II: the 1950s
89
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
The attorney said he would. Then my colleague in-
voked the Fifth Amendment and was excused from
further testimony. He was fired. So was a third mem-
ber of the Committee on Mathematical Biology. It was
done delicately. Their contracts were simply not re-
newed. I never found out why I was not called.
In any case, with the departure of Hutchins, the University
of Chicago became an inhospitable home for many of its faculty.
Landau went to work in Mechanical Engineering at Columbia.
And Rapoport moved to Palo Alto, California, for a year to be-
come a fellow at the Center for Advanced Study in the Behav-
ioral Sciences. During his year there, his closest contacts included
Alex Bavelas and Duncan Luce (Rapoport, 2000, p. 112), both of
whom had been part of the earlier MIT efforts.
Meanwhile, a general exodus from Chicago was planned.
The psychologist, James G. Miller, led a group of faculty who
moved from Chicago to the University of Michigan. They in-
cluded David Easton, Ralph Gerard and Anatol Rapoport.31 So
with both Landau and Rapoport involved in other pursuits, the
Committee on Mathematical Biology was gutted. It never revived
its thrust toward structural modeling of social phenomena. More-
over, because of its name and almost certainly because of its
mathematical bent, it never captured the attention of most so-
cial scientists. So again we see an effort that failed to open the
social network approach to the larger world of social scientists.
6.3 Columbia University in the mid 1950s
In the mid 1950s, a collection of sociologists at Columbia
University developed a general social network conception. The
leaders of this effort were the lifelong collaborators Paul Lazars-
feld and Robert K. Merton. They are pictured in Figures 6.2
and 6.3.
Lazarsfeld and Merton were unlikely collaborators. As
Merton (1998) put it, they were an “odd couple.” Their social
and intellectual backgrounds were completely different and their
31And one of my own graduate school mentors, Donald T. Campbell, left the Psychology
Department at Chicago for Northwestern at the same time.
90
Figure 6.2. Paul Lazarsfeld
Figure 6.3. Robert Merton
Social Network Analysis During the Dark Ages II: the 1950s
91
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
styles of work were almost polar opposites. They were, in fact,
hired by Columbia primarily because of their differences.
Merton (1998) described something of the background of
the Columbia department:
Rather than being a matter of design, Paul’s and my
collaboration was a wholly unanticipated and delayed
consequence of a deep division in the Columbia De-
partment of Sociology that emerged in the late 1930s
and early 1940s. The senior professors—Robert M.
MacIver, the political and social theorist, and Robert
S. Lynd, co-author of the famous Middletown se-
ries—had been at odds, both intellectually and per-
sonally, for years. As a result, they could not agree on
any new senior appointment, and brought the depart-
ment to a virtual standstill.
Lazarsfeld (1975) described how he and Merton came to be
hired into that context:
I had been director of a Rockefeller Foundation project
to study the social effects of radio. Originally, the head-
quarters were in Princeton; but in 1939 the funds were
transferred to Columbia, where I had been given the
nominal title of lecturer, without faculty status. A year
later, a full professorship in sociology became vacant,
but the department could not agree on a nomination.
The issue was whether the appointment should go to
someone who emphasized social theory or to someone
primarily concerned with empirical research. Finally,
the professorial line was divided into two lower facul-
ty positions, which were filled, respectively, by Mer-
ton and by me.
So Columbia simultaneously hired these two young sociol-
ogists who were about as different as possible. They came from
different continents, different social strata and different educa-
tional backgrounds. Moreover, they had different interests and
they approached social research with vastly different styles.
92
Paul Lazarsfeld was born to intellectual, politically liberal,
middle class parents in Vienna in 1901. His father was an attor-
ney and his mother was a psychologist. Lazarsfeld attended the
University of Vienna and received a PhD in mathematics in 1925.
But his real interest was in psychology. So, after receiving his
PhD, he took a part-time job working with Charlotte and Karl
Bühler who had just founded a new department of psychology
at the university.
Working in Vienna, Lazarsfeld began studies of both con-
sumer behavior and the impact of unemployment. This work
came to the attention of the Rockefeller Foundation, and they
awarded him a fellowship in 1933 to visit the U.S. Once here, the
political chaos in Austria convinced him to stay. He continued
to work on developing what later came to be known as market
research. This early work focused particularly on examining the
impact of radio broadcasting on people’s behavior.
Merton, in contrast, was born in a south Philadelphia slum
in 1910. His birth name was Meyer R. Schkolnik. He grew up
above his father’s dairy products store. As a teenager, young
Schkolnik began a career as a stage magician. With the hope of
enhancing that career, he made up a stage name, “Robert King
Merton.” The name stuck and he used it after high school when
he enrolled at Temple.
Calhoun (2003) described Merton’s college career:
At Temple—a school founded for “the poor boys and
girls of Philadelphia” and not yet fully accredited or
matured into a university, he chanced on a wonderful
undergraduate teacher. It was serendipity, the mature
Merton insisted. The sociologist George E. Simpson
took him on as a research assistant in a project on race
and the media and introduced him not only to sociol-
ogy but to Ralph Bunche and Franklin Frazier. Simp-
son also took Merton to the ASA annual meeting where
he met Pitirim Sorokin, founding chair of the Har-
vard sociology department. He applied to Harvard,
even though his teachers told him this was usually
beyond the reach of those graduating from Temple. And
Social Network Analysis During the Dark Ages II: the 1950s
93
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
when he arrived, Sorokin took him on as a research
assistant. By Merton’s second year they were pub-
lishing together.
Merton, then, was trained by Sorokin to be a theorist; he did not
share Lazarsfeld’s background in mathematics nor his commit-
ment to methodology.
Thus Merton’s background was the polar opposite of that
of Lazarsfeld. Merton grew up in the U.S. in a working class
family and was trained as a sociological theorist. Lazarsfeld came
from a sophisticated Viennese middle-class upbringing. He was
trained as a mathematician and he worked primarily as a socio-
logical methodologist.32 As Merton (1998) put it:
we had been working in wholly disparate fields
and had not even published in the same jour-
nals.…we also drew on entirely different tradi-
tions of social and psychological thought.
According to Merton, when he and Lazarsfeld met, only 1.3% of
their citations were to the same authors.
In the early 1940s, when he first came to Columbia as part
of the radio research project, Lazarsfeld immediately began
working with his new colleagues, Bernard Berelson, Hazel Gau-
det and William McPhee. Together they launched a major study
of voting behavior. Jerábek (2001) described that project:
…the researchers formulated the hypothesis of the two-
step flow of communication, which argued against
the idea of a universal direct effect of the mass media
on everybody…[and they] also defined and identified
opinion leaders—the people who take an interest in
public goings-on and in information from the print,
radio, and other mass media, and who then mediate
the opinions and attitudes of the people around them
(their followers).
Since both of these ideas called attention to communications link-
ing individuals, they both embody a structural perspective. They
32Some Columbia students of the period have suggested that, in the long run, Merton
was the better methodologist and Lazarsfeld the better theorist.
94
show that Lazarsfeld and his colleagues at Columbia were al-
ready thinking in structural terms in the early 1940s (Lazars-
feld, Berelson and Gaudet, 1944; Lazarsfeld, Berelson and
McPhee, 1953).
When Merton first joined the Columbia department and
Lazarsfeld was officially appointed, they had virtually no con-
tact. But fairly soon that changed. A New Yorker writer, Morton
M. Hunt, wrote a piece in 1961 that described their initial con-
tact (quoted in Lazarsfeld, 1975):
In November of 1941, Lazarsfeld felt that, as the old-
er man, he ought to do the graceful thing and ac-
knowledge the existence of his opposite number. He
invited Merton to dinner, but on the afternoon of the
engagement he got an urgent call from the Office of
Facts and Figures (the predecessor of the O.W.I.),33
requesting him to conduct an audience-reaction test
that evening on a new radio program that had been
devised as part of the agency’s pre-war morale-build-
ing effort. When the Mertons arrived, Lazarsfeld met
them at the door of his apartment and said, as the
guests recall it, “How nice, how nice that you are
here at last. But don’t take off your coat, my dear
Merton. I have a sociological surprise for you. We
will leave the ladies to dine alone together, and we
will return as soon as we can.” Then he bustled Mer-
ton off to a radio studio where a score of people were
listening to a recorded broadcast of “This Is
War”…After the program, when an assistant of Laz-
arsfeld’s questioned the audience on the reasons for
its recorded likes and dislikes, Merton perked up; he
detected theoretical shortcomings in the way the ques-
tions were being put. He started passing scribbled
notes to Lazarsfeld.…As a second batch of listeners
entered the studio, Lazarsfeld asked Merton if he
would do the post-program questioning.
33Office of War Information, the government propaganda agency.
Social Network Analysis During the Dark Ages II: the 1950s
95
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
That was the beginning of the Lazarsfeld-Merton collaboration.
As Calhoun (2003) put it, “Merton and Lazarsfeld formed an
enormously productive partnership, training generations of stu-
dents and developing a program of theoretically informed but
empirically rigorous research.”
That training included the education of a good many grad-
uate students, but, as Calhoun (2003) described it, it went much
farther:
As important as each was as an individual intellectu-
al, both Merton and Lazarsfeld may have been even
more important as mentors and animators of an intel-
lectual community at Columbia—and indeed beyond,
at the Social Science Research Council, the Center for
Advanced Study in the Behavioral Sciences, and the
Russell Sage Foundation.
In general both Lazarsfeld and Merton influenced their stu-
dents and colleagues to think in structural terms. In a personal
communication, Charles Kadushin, who was a Columbia grad-
uate student at the time, described the process:
…it was Merton who had us read Simmel (Georg, not
Arnold who was in that seminar!) line by line. Sim-
mel of course had nascent network ideas and it was
the combination of Lazarsfeld’s interest in personal
influence and Merton’s interest in Simmel that lead
me to combine the two into a network theory of social
circles.
Together Lazarsfeld and Merton worked on a range of
projects. They produced an article on communication as early
as 1948 (Lazarsfeld and Merton, 1948). And they wrote a major
article on the formation of friendships six years later (Lazarsfeld
and Merton, 1954). But perhaps their biggest contribution was
in producing students who were prepared to develop projects
on the cutting edge of social network research. First Menzel and
Katz (1956), then Coleman, Katz and Menzel (1957) examined
interpersonal factors in their classic study of the diffusion of drug
information among physicians. Peter Blau (1977) developed the
96
notion of homophily in which he argued that interaction is more
likely among individuals with similar characteristics.34 Charles
Kadushin (1966) extended and specified Simmel’s concept of
“social circles.” All of these Columbia students continued to work
in a structuralist mode, and many of them produced another
generation of students who themselves became major contribu-
tors.35
So the Columbia Department of Sociology and the Bureau
of Applied Social Research became centers of structural thought.
Their research was based on network data. Graphic images—
particularly those based on matrix permutations—were used
(Coleman and MacRae, 1960). And most of their work was de-
signed to produce mathematical models. That the Columbia
group recognized the generality of their network approach is
demonstrated in the following final word included by Coleman,
Katz and Menzel (1957):
A word should be added about the significance of re-
search of this kind, aside from the possible interest in
its specific substantive findings. It exemplifies a meth-
odological approach which will, we feel, assume a larger
role in the social research of the next decade: namely,
making social relationships and social structures the
units of statistical analysis.
It is impossible to argue that this research at Columbia was
independent of earlier work. First of all, Lazarsfeld had worked
with Moreno and Jennings to develop a structural model of in-
terpersonal choice (Moreno and Jennings, 1938). And Merton
had been a graduate student at Harvard when the structural
perspective was dominant there. Lazarsfeld (1975) described
Merton’s commitment to structural thought in these words,
“Coming to the Bureau, he was asked to work on studies that
certainly portrayed people as part of a social network and im-
34This work led to the concept of “Blau space” that is now standard in social network
research (McPherson and Ranger-Moore, 1991).
35James Coleman, for example, taught Ronald Burt and Scott Feld. Peter Blau taught
Danching Ruan and Terry C. Blum. And Charles Kadushin taught Gwen Moore and
Richard Alba. All of these students became prominent contributors to social network
analysis.
Social Network Analysis During the Dark Ages II: the 1950s
97
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
bedded in a historical context; but this was, so to say, taken for
granted.”
Thus, it is clear, then, that both Lazarsfeld and Merton had
been exposed to structural thinking before they arrived at Co-
lumbia. But they were not the only ones; Duncan Luce was at
Columbia in 1951 and Arnold Simmel moved from MIT to Co-
lumbia as a graduate student in 1952.36 Both Luce and Simmel
had been involved in the original Bavelas experiments at MIT.
My conclusion, then, is that although the Columbia effort
was guided by the earlier work, it made important new contri-
butions and added a great deal to the earlier approaches. The
research at Columbia documented the importance of the struc-
tural approach by producing a whole series of new substantive
findings. Columbia, moreover, was the first strictly sociological
effort. It provided the model for a great many later develop-
ments in that field.
6.4 Iowa State University/Michigan State Universi-
ty in the mid 1950s
Some years after the unofficial demise of sociometry, there
were a few younger scholars who rediscovered it. One of these
was Everett M. Rogers. In the mid 1950s Rogers was a graduate
student in rural sociology at Iowa State University where he re-
ceived bachelors, masters and PhD degrees. For his dissertation,
Rogers used some sociometric procedures in the study of the
diffusion of innovations.
Rogers, shown in Figure 6.4, was born in the town of Car-
roll in western Iowa. In 1955, while working on his dissertation,
he discovered the writings of Jacob Moreno. His data were pro-
duced by interviews with 155 Iowa farmers; he recorded infor-
mation on the spread of several agricultural innovations. Rogers
was looking for a way to uncover patterning in their spread.
Faced with that problem, he concluded that Moreno’s sociomet-
ric approach might be helpful. As he described it:
So I plotted the network links on a huge piece of paper
(a map of the community), about 3 feet square. There
36Luce and Coleman, moreover, shared an office during Luce’s short tenure at Columbia.
98
was spatial clustering, but the sociogram was so busy,
that little other pattern emerged. I spent the entire
summer working with these data, and eventually tried
using chemical colored balls and sticks to show the
degree of opinion leadership of certain farmers, with
the height and size of the ball indicating the number
of sociometric nominations. Many people came to see
my sociogram, but eventually I destroyed it, frustrat-
ed that I could not better understand the nature of the
network. I did not think of using indices of density,
etc. And computer programs were not yet available. I
could not find any faculty member to help me with
the problem, but I remember talking to several about
it, and showing them my data.37
Rogers, then, discovered sociometry long after the general in-
terest in it had waned. But, because he was unaware of some of
the computational developments introduced by Katz and oth-
ers, his first attempt was limited to the use of Moreno’s graphi-
cal procedures. And it turned out that those graphical procedures
did not provide the sort of analytic power that he was seeking.
But, because of his interest in diffusion, Rogers continued
to think in structural terms. He worked with the idea of stages
in the diffusion process and partitioned adopters in terms of
when they embraced an innovation. He was later influenced by
Lazarsfeld and his colleagues at Columbia, and began to pro-
duce sophisticated structural interpretations of his data on Iowa
farmers. In a paper with Beal (Rogers and Beal, 1958) Rogers
wrote:
An appropriate modification of Lazarsfeld’s model
(which he developed on the basis of his study of voting
behavior) would be: technological farming ideas of-
ten flow from the impersonal sources to the innova-
tors and early adopters and from them to the late
majority and laggards.
After a few years teaching rural sociology at Ohio State
University, Rogers went to work at Michigan State, this time in
37Personal communication.
Social Network Analysis During the Dark Ages II: the 1950s
99
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
the Department of Communication Science. He continued to
work from a structural perspective and he turned out an im-
pressive collection of students, many of whom have become
important figures in social network analysis. Included among
these are George Barnett, James Danowski, Richard Farace, Pe-
ter Monge, Nan Lin, and William Richards. These are all prom-
inent contributors to social network research. Richards is
currently the president of the International Network for Social
Network Analysis.
Following his stint at Michigan State, Rogers moved to Stan-
ford University where he worked with his student D. Lawrence
Kincaid on a major network study in rural Korea (Rogers and
Kincaid, 1981). And he continued his role as an educator by train-
ing Kincaid and another major contributor to network analysis,
Ronald Rice. Thus, Rogers was, and continues to be, a major
contributor to the field.
6.5 Manchester University/London School of Eco-
nomics in the mid 1950s
In the 1920s and 30s British social anthropology had been
dominated by two major figures, Bronislaw Malinowski and
Alfred Reginald Radcliffe-Brown. Malinowski did display some
interest in structural phenomena in his study of gift exchange in
Figure 6.4. Everett M. Rogers
100
the Trobriand Islands
(Malinowski, 1922). But
it was Radcliffe-
Brown—or R-B as he
became known—who
had a major impact on
the development of a
structural perspective.
He is shown in Figure
6.5.
Radcliffe-Brown
was born in Birming-
ham, England, in 1881.
His father died when
he was five, and his
mother was left desti-
tute. Alfred Brown (his
name until he changed
it at age forty-five) attended the Royal Commercial Travelers
School in Middlesex until 1896 when he became a Foundation
Scholar at King Edward’s School. With financial support from
his elder brother, he attended Trinity College, Cambridge. His
performance was considered brilliant and he was awarded a
college prize. From 1906 to 1908 he conducted fieldwork in the
Andaman Islands. In 1909–1910 academic year he lectured at
the London School of Economics. And in 1910 he returned to the
field, this time in Australia.
In 1920 Radcliffe-Brown was awarded the newly created
Chair of Social Anthropology at the University of Cape Town.
From there he moved to the Anthropology Chair at Sydney.38
And in 1930 he moved again—this time to the University of
Chicago. He remained at Chicago until 1937 and overlapped with
Warner, Davis, Whyte and Drake, who had all been involved in
the earlier Harvard effort. At that point, he departed to become
the first Chair in Social Anthropology at Oxford.
38One consequence of the move to Sydney was R-B’s recruitment of W. Lloyd Warner,
described in Chapter 3 above. In addition, at Sydney, Gregory Bateson worked under
R-B’s direction. Bateson went on to embrace a network perspective applied, for the
most part, to cognitive structures (Bateson, 2002).
Social Network Analysis During the Dark Ages II: the 1950s
Figure 6.5. Alfred Reginald Radcliffe-Brown
101
THE DEVELOPMENT OF SOCIAL NETWORK ANALYSIS
R-B was an admirer of Durkheim. He built from Durkheim’s
structural perspective and extended those ideas. In R-B’s obitu-
ary, Firth (1956) commented on R-B’s work in that area:
As time went on, he developed the concept of social
structure as a central theme in his analysis, and though
he did not explore the concept itself systematically, he
applied it with increasing success to the explanation
of social phenomena. This success remains outstand-
ing in the field of kinship, where the bulk of his contri-
butions must remain as a part of the enduring fabric
of anthropological studies.
As an eloquent spokesman for the structural perspective,
Radcliffe-Brown had few peers. In the spring of 1937, the Dean
of the Division of Social Sciences at Chicago organized a faculty
seminar. One of the early speakers was Hutchins’ neo-Thomist
associate, Mortimer J. Adler. Adler presented the argument that
psychology was the only possible social science. R-B asked for
the opportunity to present a rebuttal. His request was granted
and he presented his own lecture series published twenty years
later, after his death, as A Natural Science of Society (Radcliffe-
Brown, 1957).
In those lectures, R-B set out the foundations for the devel-
opment of a natural science of society. Social relations were cen-
tral in his thinking. He (1957, p.43) said, for example, “The
relations between individuals in a social system are social rela-
tions.” He (p. 44) generalized that idea and talked of social rela-
tions linking individual bees and ants into social structures. And
he (p. 49) indicated that human society could be understood only,
“by an investigation of human beings arranged in a certain order.”
It is evident, then, that R-B saw the full generality of the struc-
tural perspective.
Radcliffe-Brown went on to argue that in order to build a
science based on the relations between people, we would need
certain tools. As he (1957, p. 69) put it,
A natural science is possible, first, wherever measure-
ment can be applied to phenomena, and second, wher-
ever relational analysis is possible, i.e., systems are
102
identifiable and characterizable. Relational analysis,
even if not metrical, may be mathematical, in the sense
that it will apply non-quantitative, relational mathe-
matics. The kind of mathematics which will be required
ultimately for a full development of the science of so-
ciety will not be metrical, but will be that hitherto
comparatively neglected branch of mathematics, the
calculus of relations, which, I think, is on the whole
more fundamental than quantitative mathematics.
In this statement Radcliffe-Brown anticipated exactly the
developments that took place almost forty years later in the
emerging field of social network analysis (Freeman, 1984a). His
statement is an explicit description of the kinds of models that
have been introduced to aid in the development of the field.39
For most of his career, R-B drifted from University to Uni-
versity, spreading the word about structural analysis. In their
Foreword to R-B’s book (1952), Evans-Pritchard and Eggan de-
scribed his teaching, “He has taught social anthropology at Cam-
bridge, London, Birmingham, Pretoria, Johannesburg, Cape
Town, Sydney, Yenching, Oxford, Sao Paulo, Alexandria and
Grahamstown, and in each of these places he is remembered
with affection and respect.”40 Since he was always on the move,
R-B never remained in any place long enough to produce stu-
dents. He did work with Raymond Firth in Sydney and with
Meyer Fortes at Oxford. It is evident that Firth and Fortes, along
with the Indian sociologist, M. N. Srinivas, and the British social
anthropologists, Edward E. Evans-Pritchard and Max Gluckman,
considered themselves to be disciples of Radcliffe-Brown. All of
them, particularly Evans-Pritchard (1940) and Gluckman, reflect-
ed R-B’s structural approach in their own research.
Gluckman, then, owed his structuralist viewpoint to his
exposure to Radcliffe-Brown (Cocks, 2001). Gluckman was orig-
inally South African, born in the Transvaal. He was a Rhodes
39 These lectures were not widely available until their publication in 1957 after R-B’s
death. But R-B made similar pleas for structural thought in his presidential address to
the Royal Anthropological Institute (Radcliffe-Brown, 1940) and in the introduction
to his book on African kinship (Radcliffe-Brown and Forde, 1950).
40It is an interesting footnote that, although R-B taught at the University of Chicago
with Eggan, Chicago is not among the institutions listed by Evans-Pritchard and Eggan.
Social Network Analysis During the Dark Ages II: the 1950s
103