Emerging Trends in The Social and Behavioral Sciences

Close Friendships among Contemporary People

Item

Title

Close Friendships among Contemporary People

Author

Brashears, Matthew E.

Brashears, Laura Aufderheide

Research Area

The Individual and Society

Topic

Social Interactions in Everyday Life

Abstract

Do contemporary people have fewer friends than they used to? In this entry, we examine the research on strong tie networks, or networks containing people especially important to us, in order to provide a partial answer. We begin with a review of the methods employed to collect data on social network connections. We then summarize the literature on the size of Americans' social networks as well as change in this size over time. Finally, we conclude with an analysis of where the study of strong networks needs to go and some suggestions for getting it there. Overall, the research suggests that while strong networks have changed over the past 30–40 years, and may be smaller overall, people are no more isolated than they were in the past.

Related Essays

Emotion and Intergroup Relations (Psychology), Diane M. Mackie et al.

Emergence of Stratification in Small Groups (Sociology), Noah Askin et al.

Problems Attract Problems: A Network Perspective on Mental Disorders (Psychology), Angélique Cramer and Denny Borsboom

The Development of Social Trust (Psychology), Vikram K. Jaswal and Marissa B. Drell

Herd Behavior (Psychology), Tatsuya Kameda and Reid Hastie

Cultural Neuroscience: Connecting Culture, Brain, and Genes (Psychology), Shinobu Kitayama and Sarah Huff

Social Classification (Sociology), Elizabeth G. Pontikes

Social Relationships and Health in Older Adulthood (Psychology), Theodore F. Robles and Josephine A. Menkin

Emerging Trends: Shaping Age by Technology and Social Bonds (Sociology), Annette Spellerberg and Lynn Schelisch

Social Neuroendocrine Approaches to Relationships (Anthropology), Sari M. van Anders and Peter B. Gray

Identifier

etrds0085

extracted text

Close Friendships among
Contemporary People
MATTHEW E. BRASHEARS and LAURA AUFDERHEIDE BRASHEARS

Abstract
Do contemporary people have fewer friends than they used to? In this entry, we
examine the research on strong tie networks, or networks containing people especially important to us, in order to provide a partial answer. We begin with a review
of the methods employed to collect data on social network connections. We then summarize the literature on the size of Americans’ social networks as well as change in
this size over time. Finally, we conclude with an analysis of where the study of strong
networks needs to go and some suggestions for getting it there. Overall, the research
suggests that while strong networks have changed over the past 30–40 years, and
may be smaller overall, people are no more isolated than they were in the past.

INTRODUCTION
Human life takes place within a web, or network, of interpersonal relationships. Some of these relationships (i.e., network ties) are relatively weak,
ranging from the mere awareness of another’s existence to acquaintanceship. These relatively weak relationships typically do not provide us with
emotional support or other resources, but can provide us with knowledge
that our closer associates lack. Other relationships are strong, including close
friends, family members, and spouses. These relatively strong relationships
provide us with major support when we are in need and with companionship when we are lonely, but rarely give us access to knowledge that we could
not obtain elsewhere. Our relationships, both strong and weak, also impose
costs directly through the time and mental effort required to maintain contact
with others, and indirectly via the loans of goods and services we provide to
our associates. Finally, our social networks expose us to risks that would not
exist without them (e.g., we cannot be betrayed until we have an associate
whom we trust). This mixture of costs, benefits, and risks makes social networks a defining element of human life, and it is crucial to social science that
we understand them.
Emerging Trends in the Social and Behavioral Sciences. Edited by Robert Scott and Stephen Kosslyn.
© 2015 John Wiley & Sons, Inc. ISBN 978-1-118-90077-2.

1

2

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Much effort has been expended trying to understand one of the most basic
characteristics of strong tie networks: their size. How large is the average
individual’s social network of important people, and has this number been
increasing, decreasing, or holding steady over time? In other words, do contemporary people have fewer friends, or fewer important people in their
lives, than they used to? By understanding the size of strong tie networks,
and the tendency for these networks to change, we gain a better understanding of how human society functions and how individuals experience the
world. Moreover, the explosion of technologies connecting us to one another
makes such knowledge all the more critical.
This entry begins by reviewing the methods used to obtain estimates of the
size of strong tie networks. We then summarize the foundational research on
typical network size as well as longitudinal research exploring how network
size has changed over the past 30–40 years; on the whole, this research provides consistent signs that portions of our strong tie networks may be smaller
than they used to be, but these declines are probably not as severe as was once
feared, and are far from unambiguous. Lastly, we cover current cutting edge
research in the area, discuss key issues where further investigation is needed,
and conclude with our thoughts on the current state of the research.
FOUNDATIONAL RESEARCH
MEASURING FRIENDSHIP NETWORKS
How does one measure the size of a social network? In general three
approaches have been employed: ego network methods, sociometric methods, and the network scale-up approach. In an ego network data collection,
the researcher begins with a probability sample of some population. While
there are many ways of drawing such a sample, the end result is a set of
persons that reflect the group under study; thus a national sample will
reflect, or be representative of, the national population as a whole. Individuals selected by the sampling procedure (i.e., “egos”) are then invited to
participate in a survey that includes a “name generator,” or a question meant
to elicit a set of names that meet a set of criteria (e.g., “Who have you spoken
to in the last 8 h?”). Once a list of names has been obtained, a second set of
questions known as “name interpreters” are then often asked (e.g., “What is
this person’s sex?”) about each person named (i.e., “alter”), or a subset of the
people named. Because the egos are a representative sample, the resulting
data allow researchers to develop a representative estimate of certain
characteristics of personal networks (e.g., average network size and average
diversity) at the level of the population. At the same time, since the egos
are drawn from a larger population, few, if any, of the individuals sampled

Close Friendships among Contemporary People

3

will be directly connected to each other. Thus, we cannot understand the
complete structure of any one network using ego network data collection.
Research also shows that egos have elaborate systems of categorization for
their associates, viewing some as appropriate for recreational contact, some
as appropriate for emotional support, and some as appropriate for loans of
material assistance (e.g., Wellman & Wortley, 1990). As a result, the number
of names obtained is exquisitely sensitive to the specific name generator
asked (Bernard, Shelley, & Killworth, 1987; Marin, 2004). Put differently, a
name generator asking about whom you spend free time with may produce
a very different list than one asking about whom you would borrow a large
sum of money from, even though all persons named may be strongly tied
to you. Thus, size estimates using different types of name generators must
be compared only cautiously and the most appropriate comparisons are
between studies using the same name generator(s).
In a sociometric data collection, all individuals in a specific group (e.g., a
neighborhood or a school) are sampled, with the goal of illuminating the
entire network structure. This method often involves a roster of names,
allowing respondents to simply check off those others with whom they are
connected in particular ways, although such a roster is not required. This
approach is superior to an ego network sample in revealing the overall
structure of a social network, but there are several limitations. First, a sociometric approach requires the researchers to identify all the members of a
particular group before starting data collection. Second, because this method
is not based on sampling, a much higher proportion of the study population
must participate in order for the data to be valid. In both cases, these goals
may not be practical or even possible to achieve. Third, sociometric data
are of limited use in estimating overall population characteristics because
the characteristics of a subset do not necessarily match the characteristics
of a superset (e.g., the average network size in a neighborhood is not
necessarily the same as for a nation). This final difficulty may be declining
in significance, however, as social media makes national-level network data
available to researchers.
Finally, network size has been measured with what is often known as
a “scale-up” method. This approach begins much like an ego network with
a probability sample, but the researcher asks a name generator focused on
a population segment whose prevalence is already known. For example,
the researcher could ask, “How many people do you know who are named
“Jonathon”?” Since it is possible to determine the population prevalence
of people with this name (e.g., voter registration and phone books), the
researcher can use the answer to estimate the total size of the respondent’s
network assuming random mixing. This approach to estimating network
size imposes a relatively low burden on the respondent, but the assumption

4

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

of random mixing is not necessarily valid. For example, individuals tend
to associate with those similar to themselves (i.e., “Birds of a feather flock
together”) in a process known as homophily, thus preventing random
mixing and introducing bias into this estimation method.
All three of these approaches can be implemented actively or passively.
Active data collection involves an explicit request to individuals for the
needed information (e.g., a name generator or a roster). This method
is frequently used, both because it is practical and because it gives the
researcher more control over which ties are measured, but is vulnerable
to respondent forgetting, recall biases, and simple fatigue. In passive data
collection, the researcher either directly observes individuals having contact
with others (e.g., observes interactions between students in a classroom), or
obtains technological traces of past interactions (e.g., email, Facebook, and
phone records). In either case, individuals reveal their network ties through
their behavior rather than by answering a set of explicit questions. To the
extent that an entire social network is available in an electronic form (e.g.,
Facebook), it may be possible to capture even a massive network passively,
providing a wealth of data to the researcher. However, it is often difficult
to reliably distinguish different types of relationships using passive data
collection (e.g., not all “friends” on Facebook represent an individual’s
strong ties, even though the title “friend” implies that they might), and the
resulting data are often limited to one method of contact (e.g., email).
CLASSICAL ESTIMATES OF NETWORK SIZE
Since the 1970s, considerable attention has been devoted to estimating network size. De Sola Pool and Kochen (1978) considered the problem in the first
issue of the flagship journal for social networks (i.e., Social Networks), estimating that the average network contains several hundred ties, both strong and
weak. This conclusion was based on an informal meta-analysis of existing
studies, as well as a compilation of data sources with very different characteristics, leaving considerable room for error, and distinctions were not made
between various types of associates. A more comprehensive effort to measure network size was made by Claude Fischer (1982), who collected data
from respondents in and around an urban area in California using name generators to reveal the types of benefits people could obtain through strong
ties, including mutual favors, socializing, and emotional support. While this
effort was far more rigorous than its predecessors, it was geographically limited, and thus unable to provide estimates of network size among the larger
national population.
Building on Fischer’s work, the 1985 General Social Survey (GSS), a probability sample of the US noninstitutionalized adult population conducted

Close Friendships among Contemporary People

5

every year or every other year, included a single name generator that has
since become a standard in network research examining strong ties. This
name generator, known as the “important matters” item, asks respondents to
name those persons with whom they have discussed important matters during the preceding 6 months. No definition is given for an “important matter,”
because respondents might differ on what they consider to be “important,”
but they will most likely discuss important things with their strong ties. Thus,
the important matters item provides a way to measure strong tie networks.
The GSS name generator for the first time permitted an estimation of typical
strong network size nationwide. Using these data, Marsden (1987) found that
in 1985 Americans discussed “important matters” with an average of three
others, approximately half of whom were kin and half of whom were nonkin
(including friends). Obviously, these results do not mean that the average
American only has 1.5 friends, but the “important matters” item captures
those relationships that are relatively strong, and thus these results suggested
that in 1985 our strong network of important confidantes was relatively small
and evenly split between kin and nonkin.
It should be noted that other lines of research have produced different, and
sometimes much larger, estimates of strong tie network size than the GSS,
ranging from 3 to 150 (e.g., Bernard et al., 1990; Roberts, Dunbar, Pollet, &
Kuppens, 2009). However, because of the sensitivity of estimates to differences in name generator phrasing, it is difficult at best to compare many of
these results to each other. As such, we focus on the estimates deriving from
the GSS, which are both nationally representative, and have been repeated
longitudinally in several independent studies.
STUDIES OF NETWORK CHANGE
While the 1985 GSS provided a crucial look at the social environment of the
average American, and the first nationwide estimates of strong network
size, it was unable to provide any sense of how stable these environments
were. While a partial, and only semi-comparable repetition, of the network
module was included in the 1987 GSS, it was not until 2004 that the GSS
included a full repeat of the network module. Analysis of these data by
McPherson, Smith-Lovin, and Brashears (2006) proved to be quite surprising: whereas in 1985 Americans had an average of three others with
whom they discussed important matters, 1.4 of which were kin, in 2004
this number had declined to two, 0.88 of which were kin. Moreover, while
in 1985 only 10% of respondents gave no names in response to the name
generator, by 2004 approximately 25% of all respondents appeared to be
“socially isolated.” These results suggested that American strong networks
had declined in size by roughly one-third, and that social isolation had

6

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

more than doubled, in approximately 20-years. These results implied that
American social life was growing more fragile but a major puzzle remained:
what could be causing such a dramatic change in association? McPherson,
Smith-Lovin, and Brashears were unable to definitively identify a cause,
but speculated that the change was linked to the growth of the internet and
mobile phones, which might permit Americans to rely more exclusively
on a smaller number of associates. Nevertheless, the existing data were
insufficient to test their speculations and the decline remained unexplained.
One possible explanation for the decline, advanced by Claude Fischer
(2009), was that the data were simply wrong. He argued that the apparent
decline in network size in 2004 was attributable to one or a combination of
three factors: a training effect (i.e., respondents learn how to avoid lengthy
follow-up questions), a fatigue effect (i.e., respondents give incomplete
answers because they are tired), or a computer error. Using a set of similar,
but not identical, items on network size in the GSS data, Fischer argued
that the GSS “important matters” results were almost certainly an artifact of
these potential flaws. Thanks to Fischer’s concerns, a number of miscoded
responses in the original data were identified, but reanalysis of the GSS
data (McPherson, Smith-Lovin, & Brashears, 2008) using updated weighting
procedures continued to show that strong networks in 2004 consisted of
roughly two people, 1.39 of whom were kin, and that approximately 23%
of the respondents were socially isolated. McPherson, Smith-Lovin, and
Brashears (2009) likewise defended their position, showing that artifacts
almost certainly existed in both the 1985 and 2004 datasets, but that the
downward trend over time persisted even when controlling for these
artifacts. While this debate provided a public airing of many of the relevant
issues, it ultimately settled very little; the decline in network size had been
called into question, but no “smoking gun” had been identified.
CUTTING-EDGE RESEARCH
Recently several attempts have been made to resolve the uncertainty over
network size. While McPherson et al. speculated that the rise of the internet and other communications technologies might have led to a reduction
in strong network size, accumulating evidence suggests that this is not the
case. Tufecki (2010) finds that while online interaction does not consistently
increase sociability, it also does not consistently decrease it. Hampton, Sessions, and Her (2011) used a nationally representative phone survey conducted in 2008 to show that internet and communications technology usage
either has no effect, or a slight positive effect, on both network size and sociability. It thus is unlikely that the reduction in strong network size, if real, is
the result of the increasing usage of communications technologies.

Close Friendships among Contemporary People

7

In addition to evaluating the effect of internet usage on sociability, Hampton et al.’s (2011) study also included the same “important matters” name
generator that appeared on the 1985 and 2004 GSS administrations. Examination of these data reveals that only 12% of respondents reported no discussions of important matters in the preceding 6 months, a substantial decline
from the 23% detected by McPherson et al. (2008) over a period of only 4
years. These same data, however, indicated an average strong network size of
approximately two, with 1.19 of these being kin. Thus, while Hampton et al.
found a different level of social isolation, the estimated average size of the
network in 2008 was very similar to that detected by McPherson et al. (2006,
2008). A second study by Hampton, Goulet, Marlow, and Rainie (2011) using
a similar methodology and data collected in 2011 also yielded an average
strong network size of approximately two.
Research by Brashears (2011) investigated the artifacts proposed by Fischer
(2009) directly. This study used a general population experiment to evaluate the potential for training and fatigue effects to have reduced apparent
network size. In addition, because the data were collected from a representative sample of the national population in 2010, it is possible to use these
data to construct national estimates. The results failed to support the training or fatigue effect explanations for the reduction in network size, and once
more found an average strong network size of approximately two. However,
this research also found that only a little more than 4% of the population was
“socially isolated,” which is two-thirds less than Hampton et al.’s (2011) findings from 2008 and roughly one-fifth the level of social isolation observed by
McPherson et al. (2008) in the 2004 GSS. This suggests that while the average
strong network size is relatively stable over time, the level of social isolation varies wildly and is quite likely unreliable. Brashears (2011) presented
additional evidence to this effect, finding that of those who reported no discussions of important matters, roughly two-thirds said that this was because
they had no “important matters” to talk about, rather than no one with whom
to discuss them.
Finally, research by Paik and Sanchagrin, currently 2013, takes a new and
innovative approach to the debate by suggesting that the decrease in strong
network size may have been an artifact of fatigue among the interviewers
rather than fatigue among the respondents. They find that high levels of
social isolation are not distributed evenly by interviewer, but rather are significantly more likely to be observed by specific interviewers. They suggest
that perhaps interviewers guided respondents into answering that they had
no discussions of important matters in order to shorten the interviews. This
novel explanation is consistent with existing research showing that respondent artifacts probably did not produce the shrinking networks finding. At
the same time, however, it is unclear why fatigue would suddenly be an

8

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

issue in 2004, but not in 1985 when interview procedures were more cumbersome overall. Likewise, it does not account for Hampton et al.’s (2011)
results using a much shorter instrument, or Brashears’ (2011) results using
a computer-assisted self-interview method that does not include an interviewer.
This recent research has thus advanced our understanding of strong network size, but has not completely answered the question: are strong networks smaller now than they were before?
KEY ISSUES FOR FUTURE RESEARCH
The size of the average strong network is a matter of profound concern for
social science, but it remains, for the moment, in considerable doubt. It is
therefore also uncertain whether strong networks have declined in size in
recent decades, although several independent data collection efforts have
given us reason to think that they have. The controversy over so fundamental
an issue serves as a pointed reminder that we may not know as much about
how network items map onto social reality as we thought, and that much
basic research remains to be done.
One of the most crucial questions is: how do respondents mentally represent their social networks? While multiple studies indicate that strong
networks are smaller now than in the past, convincing evidence indicates
that people with whom we have strong ties may not all provide us with
the same types of benefits (Wellman & Wortley, 1990). In other words, if the
way respondents classify their alters changes, it may give the appearance
of a change in network size, without the substance of the network structure
actually changing very much at all. This concern is underscored by the
fact that discussions of “important matters” are apparently often regarding
topics that many would consider trivial (Bearman & Parigi, 2004). This
implies that the name generator may, to some extent, not be capturing actual
discussions, but rather a subset of the network that the respondent views as
suitable for important discussions. Although new media technologies, such
as Facebook and Twitter, have not led to rampant social isolation, they may
be contributing to a transformation of how we categorize our associates,
including our friends and discussion partners. Research in social psychology
has indicated that the meaning of social categories can both change over
time, and differ between social groups, and technology may play a role in
these differences. Thus, natural changes in how relationships are understood
may result in respondents and researchers defining ties in disparate ways.
Without similar definitions, researchers may end up with biased estimates,
possibly explaining the confusion over network size and social isolation.

Close Friendships among Contemporary People

9

Social isolation, or the lack of any contacts, has received considerable attention, particularly given its extraordinarily high level in the 2004 GSS data.
However, subsequent data collections have failed to replicate this level of
isolation even while finding very comparable average network sizes. This
suggests that what may determine the level of social isolation is less the availability of associates with whom to discuss important matters, but rather the
availability of topics to have discussions about. When topics are relatively
hard to come by, social isolation may appear to increase. Researchers therefore need to work harder to distinguish measurements of the existence of a
relationship from measurements of particular uses of that relationship.
The instability in the estimates of social isolation emphasizes an additional
issue: that the concern over network size in general, and isolation in particular, may be somewhat misguided. The real question is not whether an
individual has any associates, but whether they have enough to provide the
support that they need. Our attention should be on social poverty, or an inadequate level of social support, rather than on social isolation, or no social
support whatsoever. It is likely the case that few individuals are truly isolated, but many may suffer under some degree of social poverty.
As the debate over the shrinking of American strong networks has made
painfully clear, we have a poor understanding at present of how networks
vary naturally in size over time. Certainly networks change over time with
individuals forming new friendships and losing old ones. At the population
level, there may be natural trends to the gain and loss of associates, and
these trends may vary over long periods of time, possibly in cyclical patterns. While many network studies have been carried out over the decades,
they often involve different populations, different name generators, or both,
and thus can tell us little about these natural variations. As such, even if the
decline in strong network size observed in much of the above research is real,
that does not mean that there is cause for alarm: over a span of decades, average strong network size may vary naturally by a considerable amount. The
simple reality is that, at the present time, we do not know if such natural variation exists and, if so, how extreme it may be, and thus cannot know if any
particular increase or decrease in network size is a virtue or a vice.
A promising medium for collecting such data is the utilization of passive data collection using social media and other electronic traces. These
approaches have real advantages in that they avoid many of the potential
biases introduced by question interpretation that have plagued network
research to date, and thus represent a valuable approach to measuring
network size. At the same time, however, it is unclear how relationships
recorded in social media (e.g., Facebook or Twitter) connect to the important
relationships of individuals. Some important associates may be contacted
rarely via social media, while more trivial contacts are interacted with

10

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

frequently using these technologies. It is thus critical that researchers
develop both expertise with these new sources of data, as well as explore the
connections, or differences, between them and more traditional measures.
CONCLUSION
A considerable amount of research effort has been expended on measuring
the size of modern strong tie networks and determining whether they have
grown smaller over the previous several decades. Unfortunately, the best
that we can say with the existing research is that they may have shrunk, but
that we cannot be sure. The early research on long-term change in network
size suggests that networks are smaller, and several independent data collection efforts since have also shown smaller networks, but the distinctive
social isolation finding has failed to replicate, and there are growing signs
that methodological artifacts may be partly to blame. All we can say with
certainty at this point is that more research, both fundamental and applied,
is needed in order to resolve this question. It is the privilege of authors,
however, to offer their own opinions and ours is this: strong tie networks
are different now than they were nearly three decades ago, but they are not
necessarily smaller. As our society has changed, so too have our definitions
of friend, confidant, and associate transformed to fit the new demands, and
constraints, of our lives. It may indeed be the case that we have fewer of
those persons with whom we might have once discussed “important matters,” be they kin or friends, but it is our view that modern Americans remain
intensely social, even if our relationships have been evolving into endless
forms most wonderful.
REFERENCES
Bearman, P., & Parigi, P. (2004). Cloning headless frogs and other important matters:
conversation topics and network structure. Social Forces, 83, 535–557.
Bernard, H. R., Johnsen, E. C., Killworth, P. D., McCarty, C., Shelley, G., & Robinson, S. (1990). Comparing four different methods for measuring personal social
networks. Social Networks, 12, 179–215.
Bernard, H. R., Shelley, G. A., & Killworth, P. (1987). How much of a network does
the GSS and RSW dredge up? Social Networks, 9, 49–61.
Brashears, M. E. (2011). Small networks and high isolation? A reexamination of
American discussion networks. Social Networks, 33, 331–341.
De Sola Pool, I., & Kochen, M. (1978). Contacts and influence. Social Networks, 1, 5–51.
Fischer, C. S. (1982). To dwell among friends: Personal networks in town and city. Chicago,
IL: University of Chicago Press.
Fischer, C. S. (2009). The 2004 GSS finding of shrunken social networks: An artifact?
American Sociological Review, 74, 657–669.

Close Friendships among Contemporary People

11

Hampton, K. N., Goulet, L. S., Rainie, L., & Purcell, K. (2011). Social networking sites and our lives. Pew Internet and American Life Project, June 16,
2011. Retrieved from http://www.pewinternet.org/Reports/2011/Technologyand-socialnetworks.aspx.
Hampton, K. N., Sessions, L. F., & Her, E. J. (2011). Core networks, social isolation,
and new media: Internet and mobile phone use, network size, and diversity. Information, Communication & Society, 14, 130–155.
Marin, A. (2004). Are respondents more likely to list alters with certain characteristics? Implications for name generator data. Social Networks, 26, 289–307.
Marsden, P. V. (1987). Core discussion networks of Americans. American Sociological
Review, 52, 122–131.
McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2006). Social isolation in America: Changes in core discussion networks over two decades. American Sociological
Review, 71, 353–375.
McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2008). Erratum: Social isolation in America: Changes in core discussion networks over two decades. American
Sociological Review, 73, 1022.
McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2009). Models and marginals:
Using survey evidence to study social networks. American Sociological Review, 74,
670–681.
Paik, A., & Sanchagrin, K. (2013). Social isolation in America: An artifact. American
Sociological Review, 78, 339–360.
Roberts, S. G. B., Dunbar, R. I. M., Pollet, T. V., & Kuppens, T. (2009). Exploring variation in active network size: Constraints and ego characteristics. Social Networks,
31, 138–146.
Tufecki, Z. (2010). Who acquires friends through social media and why? ‘Rich Get
Richer’ versus ‘Seek and Ye Shall Find’. Proceedings of the 4th International Conference on Weblogs and Social Media (ICWSM). AAAI Press.
Wellman, B., & Wortley, S. (1990). Different strokes from different folks: Community
ties and social support. The American Journal of Sociology, 96, 558–588.

FURTHER READING
Fischer, C. S. (2011). Still connected: Family and friends in America since 1970. New York,
NY: Russell Sage Foundation.
Kadushin, C. (2012). Understanding social networks: Theories, concepts and findings.
Oxford, England: Oxford University Press.
Putnam, R. D. (2001). Bowling alone: The collapse and revival of American community.
New York, NY: Simon and Schuster.
Rainie, L., & Wellman, B. (2012). Networked: The new social operating system. Cambridge, MA: MIT Press.

MATTHEW E. BRASHEARS SHORT BIOGRAPHY
Matthew E. Brashears is an Assistant Professor of Sociology at Cornell University. In addition to his work on ego networks and average

12

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

network size, he is interested in the connections between social networks and social psychology. His current funded research includes an
NSF-sponsored study linking cognition and memory to social network
structure and a Department of Defense-sponsored effort to develop new
ways of detecting terrorist groups preparing chemical, biological, radiological, and nuclear attacks. His newest research is on the mutation of
information as it diffuses through social networks. Personal webpage:
http://www.soc.cornell.edu/faculty/brashears.html
LAURA AUFDERHEIDE BRASHEARS SHORT BIOGRAPHY
Laura Aufderheide Brashears is a Visiting Scholar at Cornell University.
A sociological social psychologist, her research sits at the intersection of
inequality and the self. She is currently examining how the strength of
one’s racial and ethnic identities influences prejudice toward racial/ethnic
out-groups and preference for one’s racial/ethnic in-group. Her past research
includes substantial study into the effects of self-esteem on academic identities and academic achievement in African American, Mexican American,
and White youth, as well as the influence of nonparental significant others
on adolescent self-esteem.
RELATED ESSAYS
Emotion and Intergroup Relations (Psychology), Diane M. Mackie et al.
Emergence of Stratification in Small Groups (Sociology), Noah Askin et al.
Problems Attract Problems: A Network Perspective on Mental Disorders
(Psychology), Angélique Cramer and Denny Borsboom
The Development of Social Trust (Psychology), Vikram K. Jaswal and Marissa
B. Drell
Herd Behavior (Psychology), Tatsuya Kameda and Reid Hastie
Cultural Neuroscience: Connecting Culture, Brain, and Genes (Psychology),
Shinobu Kitayama and Sarah Huff
Social Classification (Sociology), Elizabeth G. Pontikes
Social Relationships and Health in Older Adulthood (Psychology), Theodore
F. Robles and Josephine A. Menkin
Emerging Trends: Shaping Age by Technology and Social Bonds (Sociology),
Annette Spellerberg and Lynn Schelisch
Social Neuroendocrine Approaches to Relationships (Anthropology), Sari M.
van Anders and Peter B. Gray



Close Friendships among
Contemporary People
MATTHEW E. BRASHEARS and LAURA AUFDERHEIDE BRASHEARS

Abstract
Do contemporary people have fewer friends than they used to? In this entry, we
examine the research on strong tie networks, or networks containing people especially important to us, in order to provide a partial answer. We begin with a review
of the methods employed to collect data on social network connections. We then summarize the literature on the size of Americans’ social networks as well as change in
this size over time. Finally, we conclude with an analysis of where the study of strong
networks needs to go and some suggestions for getting it there. Overall, the research
suggests that while strong networks have changed over the past 30–40 years, and
may be smaller overall, people are no more isolated than they were in the past.

INTRODUCTION
Human life takes place within a web, or network, of interpersonal relationships. Some of these relationships (i.e., network ties) are relatively weak,
ranging from the mere awareness of another’s existence to acquaintanceship. These relatively weak relationships typically do not provide us with
emotional support or other resources, but can provide us with knowledge
that our closer associates lack. Other relationships are strong, including close
friends, family members, and spouses. These relatively strong relationships
provide us with major support when we are in need and with companionship when we are lonely, but rarely give us access to knowledge that we could
not obtain elsewhere. Our relationships, both strong and weak, also impose
costs directly through the time and mental effort required to maintain contact
with others, and indirectly via the loans of goods and services we provide to
our associates. Finally, our social networks expose us to risks that would not
exist without them (e.g., we cannot be betrayed until we have an associate
whom we trust). This mixture of costs, benefits, and risks makes social networks a defining element of human life, and it is crucial to social science that
we understand them.
Emerging Trends in the Social and Behavioral Sciences. Edited by Robert Scott and Stephen Kosslyn.
© 2015 John Wiley & Sons, Inc. ISBN 978-1-118-90077-2.

1

2

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Much effort has been expended trying to understand one of the most basic
characteristics of strong tie networks: their size. How large is the average
individual’s social network of important people, and has this number been
increasing, decreasing, or holding steady over time? In other words, do contemporary people have fewer friends, or fewer important people in their
lives, than they used to? By understanding the size of strong tie networks,
and the tendency for these networks to change, we gain a better understanding of how human society functions and how individuals experience the
world. Moreover, the explosion of technologies connecting us to one another
makes such knowledge all the more critical.
This entry begins by reviewing the methods used to obtain estimates of the
size of strong tie networks. We then summarize the foundational research on
typical network size as well as longitudinal research exploring how network
size has changed over the past 30–40 years; on the whole, this research provides consistent signs that portions of our strong tie networks may be smaller
than they used to be, but these declines are probably not as severe as was once
feared, and are far from unambiguous. Lastly, we cover current cutting edge
research in the area, discuss key issues where further investigation is needed,
and conclude with our thoughts on the current state of the research.
FOUNDATIONAL RESEARCH
MEASURING FRIENDSHIP NETWORKS
How does one measure the size of a social network? In general three
approaches have been employed: ego network methods, sociometric methods, and the network scale-up approach. In an ego network data collection,
the researcher begins with a probability sample of some population. While
there are many ways of drawing such a sample, the end result is a set of
persons that reflect the group under study; thus a national sample will
reflect, or be representative of, the national population as a whole. Individuals selected by the sampling procedure (i.e., “egos”) are then invited to
participate in a survey that includes a “name generator,” or a question meant
to elicit a set of names that meet a set of criteria (e.g., “Who have you spoken
to in the last 8 h?”). Once a list of names has been obtained, a second set of
questions known as “name interpreters” are then often asked (e.g., “What is
this person’s sex?”) about each person named (i.e., “alter”), or a subset of the
people named. Because the egos are a representative sample, the resulting
data allow researchers to develop a representative estimate of certain
characteristics of personal networks (e.g., average network size and average
diversity) at the level of the population. At the same time, since the egos
are drawn from a larger population, few, if any, of the individuals sampled

Close Friendships among Contemporary People

3

will be directly connected to each other. Thus, we cannot understand the
complete structure of any one network using ego network data collection.
Research also shows that egos have elaborate systems of categorization for
their associates, viewing some as appropriate for recreational contact, some
as appropriate for emotional support, and some as appropriate for loans of
material assistance (e.g., Wellman & Wortley, 1990). As a result, the number
of names obtained is exquisitely sensitive to the specific name generator
asked (Bernard, Shelley, & Killworth, 1987; Marin, 2004). Put differently, a
name generator asking about whom you spend free time with may produce
a very different list than one asking about whom you would borrow a large
sum of money from, even though all persons named may be strongly tied
to you. Thus, size estimates using different types of name generators must
be compared only cautiously and the most appropriate comparisons are
between studies using the same name generator(s).
In a sociometric data collection, all individuals in a specific group (e.g., a
neighborhood or a school) are sampled, with the goal of illuminating the
entire network structure. This method often involves a roster of names,
allowing respondents to simply check off those others with whom they are
connected in particular ways, although such a roster is not required. This
approach is superior to an ego network sample in revealing the overall
structure of a social network, but there are several limitations. First, a sociometric approach requires the researchers to identify all the members of a
particular group before starting data collection. Second, because this method
is not based on sampling, a much higher proportion of the study population
must participate in order for the data to be valid. In both cases, these goals
may not be practical or even possible to achieve. Third, sociometric data
are of limited use in estimating overall population characteristics because
the characteristics of a subset do not necessarily match the characteristics
of a superset (e.g., the average network size in a neighborhood is not
necessarily the same as for a nation). This final difficulty may be declining
in significance, however, as social media makes national-level network data
available to researchers.
Finally, network size has been measured with what is often known as
a “scale-up” method. This approach begins much like an ego network with
a probability sample, but the researcher asks a name generator focused on
a population segment whose prevalence is already known. For example,
the researcher could ask, “How many people do you know who are named
“Jonathon”?” Since it is possible to determine the population prevalence
of people with this name (e.g., voter registration and phone books), the
researcher can use the answer to estimate the total size of the respondent’s
network assuming random mixing. This approach to estimating network
size imposes a relatively low burden on the respondent, but the assumption

4

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

of random mixing is not necessarily valid. For example, individuals tend
to associate with those similar to themselves (i.e., “Birds of a feather flock
together”) in a process known as homophily, thus preventing random
mixing and introducing bias into this estimation method.
All three of these approaches can be implemented actively or passively.
Active data collection involves an explicit request to individuals for the
needed information (e.g., a name generator or a roster). This method
is frequently used, both because it is practical and because it gives the
researcher more control over which ties are measured, but is vulnerable
to respondent forgetting, recall biases, and simple fatigue. In passive data
collection, the researcher either directly observes individuals having contact
with others (e.g., observes interactions between students in a classroom), or
obtains technological traces of past interactions (e.g., email, Facebook, and
phone records). In either case, individuals reveal their network ties through
their behavior rather than by answering a set of explicit questions. To the
extent that an entire social network is available in an electronic form (e.g.,
Facebook), it may be possible to capture even a massive network passively,
providing a wealth of data to the researcher. However, it is often difficult
to reliably distinguish different types of relationships using passive data
collection (e.g., not all “friends” on Facebook represent an individual’s
strong ties, even though the title “friend” implies that they might), and the
resulting data are often limited to one method of contact (e.g., email).
CLASSICAL ESTIMATES OF NETWORK SIZE
Since the 1970s, considerable attention has been devoted to estimating network size. De Sola Pool and Kochen (1978) considered the problem in the first
issue of the flagship journal for social networks (i.e., Social Networks), estimating that the average network contains several hundred ties, both strong and
weak. This conclusion was based on an informal meta-analysis of existing
studies, as well as a compilation of data sources with very different characteristics, leaving considerable room for error, and distinctions were not made
between various types of associates. A more comprehensive effort to measure network size was made by Claude Fischer (1982), who collected data
from respondents in and around an urban area in California using name generators to reveal the types of benefits people could obtain through strong
ties, including mutual favors, socializing, and emotional support. While this
effort was far more rigorous than its predecessors, it was geographically limited, and thus unable to provide estimates of network size among the larger
national population.
Building on Fischer’s work, the 1985 General Social Survey (GSS), a probability sample of the US noninstitutionalized adult population conducted

Close Friendships among Contemporary People

5

every year or every other year, included a single name generator that has
since become a standard in network research examining strong ties. This
name generator, known as the “important matters” item, asks respondents to
name those persons with whom they have discussed important matters during the preceding 6 months. No definition is given for an “important matter,”
because respondents might differ on what they consider to be “important,”
but they will most likely discuss important things with their strong ties. Thus,
the important matters item provides a way to measure strong tie networks.
The GSS name generator for the first time permitted an estimation of typical
strong network size nationwide. Using these data, Marsden (1987) found that
in 1985 Americans discussed “important matters” with an average of three
others, approximately half of whom were kin and half of whom were nonkin
(including friends). Obviously, these results do not mean that the average
American only has 1.5 friends, but the “important matters” item captures
those relationships that are relatively strong, and thus these results suggested
that in 1985 our strong network of important confidantes was relatively small
and evenly split between kin and nonkin.
It should be noted that other lines of research have produced different, and
sometimes much larger, estimates of strong tie network size than the GSS,
ranging from 3 to 150 (e.g., Bernard et al., 1990; Roberts, Dunbar, Pollet, &
Kuppens, 2009). However, because of the sensitivity of estimates to differences in name generator phrasing, it is difficult at best to compare many of
these results to each other. As such, we focus on the estimates deriving from
the GSS, which are both nationally representative, and have been repeated
longitudinally in several independent studies.
STUDIES OF NETWORK CHANGE
While the 1985 GSS provided a crucial look at the social environment of the
average American, and the first nationwide estimates of strong network
size, it was unable to provide any sense of how stable these environments
were. While a partial, and only semi-comparable repetition, of the network
module was included in the 1987 GSS, it was not until 2004 that the GSS
included a full repeat of the network module. Analysis of these data by
McPherson, Smith-Lovin, and Brashears (2006) proved to be quite surprising: whereas in 1985 Americans had an average of three others with
whom they discussed important matters, 1.4 of which were kin, in 2004
this number had declined to two, 0.88 of which were kin. Moreover, while
in 1985 only 10% of respondents gave no names in response to the name
generator, by 2004 approximately 25% of all respondents appeared to be
“socially isolated.” These results suggested that American strong networks
had declined in size by roughly one-third, and that social isolation had

6

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

more than doubled, in approximately 20-years. These results implied that
American social life was growing more fragile but a major puzzle remained:
what could be causing such a dramatic change in association? McPherson,
Smith-Lovin, and Brashears were unable to definitively identify a cause,
but speculated that the change was linked to the growth of the internet and
mobile phones, which might permit Americans to rely more exclusively
on a smaller number of associates. Nevertheless, the existing data were
insufficient to test their speculations and the decline remained unexplained.
One possible explanation for the decline, advanced by Claude Fischer
(2009), was that the data were simply wrong. He argued that the apparent
decline in network size in 2004 was attributable to one or a combination of
three factors: a training effect (i.e., respondents learn how to avoid lengthy
follow-up questions), a fatigue effect (i.e., respondents give incomplete
answers because they are tired), or a computer error. Using a set of similar,
but not identical, items on network size in the GSS data, Fischer argued
that the GSS “important matters” results were almost certainly an artifact of
these potential flaws. Thanks to Fischer’s concerns, a number of miscoded
responses in the original data were identified, but reanalysis of the GSS
data (McPherson, Smith-Lovin, & Brashears, 2008) using updated weighting
procedures continued to show that strong networks in 2004 consisted of
roughly two people, 1.39 of whom were kin, and that approximately 23%
of the respondents were socially isolated. McPherson, Smith-Lovin, and
Brashears (2009) likewise defended their position, showing that artifacts
almost certainly existed in both the 1985 and 2004 datasets, but that the
downward trend over time persisted even when controlling for these
artifacts. While this debate provided a public airing of many of the relevant
issues, it ultimately settled very little; the decline in network size had been
called into question, but no “smoking gun” had been identified.
CUTTING-EDGE RESEARCH
Recently several attempts have been made to resolve the uncertainty over
network size. While McPherson et al. speculated that the rise of the internet and other communications technologies might have led to a reduction
in strong network size, accumulating evidence suggests that this is not the
case. Tufecki (2010) finds that while online interaction does not consistently
increase sociability, it also does not consistently decrease it. Hampton, Sessions, and Her (2011) used a nationally representative phone survey conducted in 2008 to show that internet and communications technology usage
either has no effect, or a slight positive effect, on both network size and sociability. It thus is unlikely that the reduction in strong network size, if real, is
the result of the increasing usage of communications technologies.

Close Friendships among Contemporary People

7

In addition to evaluating the effect of internet usage on sociability, Hampton et al.’s (2011) study also included the same “important matters” name
generator that appeared on the 1985 and 2004 GSS administrations. Examination of these data reveals that only 12% of respondents reported no discussions of important matters in the preceding 6 months, a substantial decline
from the 23% detected by McPherson et al. (2008) over a period of only 4
years. These same data, however, indicated an average strong network size of
approximately two, with 1.19 of these being kin. Thus, while Hampton et al.
found a different level of social isolation, the estimated average size of the
network in 2008 was very similar to that detected by McPherson et al. (2006,
2008). A second study by Hampton, Goulet, Marlow, and Rainie (2011) using
a similar methodology and data collected in 2011 also yielded an average
strong network size of approximately two.
Research by Brashears (2011) investigated the artifacts proposed by Fischer
(2009) directly. This study used a general population experiment to evaluate the potential for training and fatigue effects to have reduced apparent
network size. In addition, because the data were collected from a representative sample of the national population in 2010, it is possible to use these
data to construct national estimates. The results failed to support the training or fatigue effect explanations for the reduction in network size, and once
more found an average strong network size of approximately two. However,
this research also found that only a little more than 4% of the population was
“socially isolated,” which is two-thirds less than Hampton et al.’s (2011) findings from 2008 and roughly one-fifth the level of social isolation observed by
McPherson et al. (2008) in the 2004 GSS. This suggests that while the average
strong network size is relatively stable over time, the level of social isolation varies wildly and is quite likely unreliable. Brashears (2011) presented
additional evidence to this effect, finding that of those who reported no discussions of important matters, roughly two-thirds said that this was because
they had no “important matters” to talk about, rather than no one with whom
to discuss them.
Finally, research by Paik and Sanchagrin, currently 2013, takes a new and
innovative approach to the debate by suggesting that the decrease in strong
network size may have been an artifact of fatigue among the interviewers
rather than fatigue among the respondents. They find that high levels of
social isolation are not distributed evenly by interviewer, but rather are significantly more likely to be observed by specific interviewers. They suggest
that perhaps interviewers guided respondents into answering that they had
no discussions of important matters in order to shorten the interviews. This
novel explanation is consistent with existing research showing that respondent artifacts probably did not produce the shrinking networks finding. At
the same time, however, it is unclear why fatigue would suddenly be an

8

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

issue in 2004, but not in 1985 when interview procedures were more cumbersome overall. Likewise, it does not account for Hampton et al.’s (2011)
results using a much shorter instrument, or Brashears’ (2011) results using
a computer-assisted self-interview method that does not include an interviewer.
This recent research has thus advanced our understanding of strong network size, but has not completely answered the question: are strong networks smaller now than they were before?
KEY ISSUES FOR FUTURE RESEARCH
The size of the average strong network is a matter of profound concern for
social science, but it remains, for the moment, in considerable doubt. It is
therefore also uncertain whether strong networks have declined in size in
recent decades, although several independent data collection efforts have
given us reason to think that they have. The controversy over so fundamental
an issue serves as a pointed reminder that we may not know as much about
how network items map onto social reality as we thought, and that much
basic research remains to be done.
One of the most crucial questions is: how do respondents mentally represent their social networks? While multiple studies indicate that strong
networks are smaller now than in the past, convincing evidence indicates
that people with whom we have strong ties may not all provide us with
the same types of benefits (Wellman & Wortley, 1990). In other words, if the
way respondents classify their alters changes, it may give the appearance
of a change in network size, without the substance of the network structure
actually changing very much at all. This concern is underscored by the
fact that discussions of “important matters” are apparently often regarding
topics that many would consider trivial (Bearman & Parigi, 2004). This
implies that the name generator may, to some extent, not be capturing actual
discussions, but rather a subset of the network that the respondent views as
suitable for important discussions. Although new media technologies, such
as Facebook and Twitter, have not led to rampant social isolation, they may
be contributing to a transformation of how we categorize our associates,
including our friends and discussion partners. Research in social psychology
has indicated that the meaning of social categories can both change over
time, and differ between social groups, and technology may play a role in
these differences. Thus, natural changes in how relationships are understood
may result in respondents and researchers defining ties in disparate ways.
Without similar definitions, researchers may end up with biased estimates,
possibly explaining the confusion over network size and social isolation.

Close Friendships among Contemporary People

9

Social isolation, or the lack of any contacts, has received considerable attention, particularly given its extraordinarily high level in the 2004 GSS data.
However, subsequent data collections have failed to replicate this level of
isolation even while finding very comparable average network sizes. This
suggests that what may determine the level of social isolation is less the availability of associates with whom to discuss important matters, but rather the
availability of topics to have discussions about. When topics are relatively
hard to come by, social isolation may appear to increase. Researchers therefore need to work harder to distinguish measurements of the existence of a
relationship from measurements of particular uses of that relationship.
The instability in the estimates of social isolation emphasizes an additional
issue: that the concern over network size in general, and isolation in particular, may be somewhat misguided. The real question is not whether an
individual has any associates, but whether they have enough to provide the
support that they need. Our attention should be on social poverty, or an inadequate level of social support, rather than on social isolation, or no social
support whatsoever. It is likely the case that few individuals are truly isolated, but many may suffer under some degree of social poverty.
As the debate over the shrinking of American strong networks has made
painfully clear, we have a poor understanding at present of how networks
vary naturally in size over time. Certainly networks change over time with
individuals forming new friendships and losing old ones. At the population
level, there may be natural trends to the gain and loss of associates, and
these trends may vary over long periods of time, possibly in cyclical patterns. While many network studies have been carried out over the decades,
they often involve different populations, different name generators, or both,
and thus can tell us little about these natural variations. As such, even if the
decline in strong network size observed in much of the above research is real,
that does not mean that there is cause for alarm: over a span of decades, average strong network size may vary naturally by a considerable amount. The
simple reality is that, at the present time, we do not know if such natural variation exists and, if so, how extreme it may be, and thus cannot know if any
particular increase or decrease in network size is a virtue or a vice.
A promising medium for collecting such data is the utilization of passive data collection using social media and other electronic traces. These
approaches have real advantages in that they avoid many of the potential
biases introduced by question interpretation that have plagued network
research to date, and thus represent a valuable approach to measuring
network size. At the same time, however, it is unclear how relationships
recorded in social media (e.g., Facebook or Twitter) connect to the important
relationships of individuals. Some important associates may be contacted
rarely via social media, while more trivial contacts are interacted with

10

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

frequently using these technologies. It is thus critical that researchers
develop both expertise with these new sources of data, as well as explore the
connections, or differences, between them and more traditional measures.
CONCLUSION
A considerable amount of research effort has been expended on measuring
the size of modern strong tie networks and determining whether they have
grown smaller over the previous several decades. Unfortunately, the best
that we can say with the existing research is that they may have shrunk, but
that we cannot be sure. The early research on long-term change in network
size suggests that networks are smaller, and several independent data collection efforts since have also shown smaller networks, but the distinctive
social isolation finding has failed to replicate, and there are growing signs
that methodological artifacts may be partly to blame. All we can say with
certainty at this point is that more research, both fundamental and applied,
is needed in order to resolve this question. It is the privilege of authors,
however, to offer their own opinions and ours is this: strong tie networks
are different now than they were nearly three decades ago, but they are not
necessarily smaller. As our society has changed, so too have our definitions
of friend, confidant, and associate transformed to fit the new demands, and
constraints, of our lives. It may indeed be the case that we have fewer of
those persons with whom we might have once discussed “important matters,” be they kin or friends, but it is our view that modern Americans remain
intensely social, even if our relationships have been evolving into endless
forms most wonderful.
REFERENCES
Bearman, P., & Parigi, P. (2004). Cloning headless frogs and other important matters:
conversation topics and network structure. Social Forces, 83, 535–557.
Bernard, H. R., Johnsen, E. C., Killworth, P. D., McCarty, C., Shelley, G., & Robinson, S. (1990). Comparing four different methods for measuring personal social
networks. Social Networks, 12, 179–215.
Bernard, H. R., Shelley, G. A., & Killworth, P. (1987). How much of a network does
the GSS and RSW dredge up? Social Networks, 9, 49–61.
Brashears, M. E. (2011). Small networks and high isolation? A reexamination of
American discussion networks. Social Networks, 33, 331–341.
De Sola Pool, I., & Kochen, M. (1978). Contacts and influence. Social Networks, 1, 5–51.
Fischer, C. S. (1982). To dwell among friends: Personal networks in town and city. Chicago,
IL: University of Chicago Press.
Fischer, C. S. (2009). The 2004 GSS finding of shrunken social networks: An artifact?
American Sociological Review, 74, 657–669.

Close Friendships among Contemporary People

11

Hampton, K. N., Goulet, L. S., Rainie, L., & Purcell, K. (2011). Social networking sites and our lives. Pew Internet and American Life Project, June 16,
2011. Retrieved from http://www.pewinternet.org/Reports/2011/Technologyand-socialnetworks.aspx.
Hampton, K. N., Sessions, L. F., & Her, E. J. (2011). Core networks, social isolation,
and new media: Internet and mobile phone use, network size, and diversity. Information, Communication & Society, 14, 130–155.
Marin, A. (2004). Are respondents more likely to list alters with certain characteristics? Implications for name generator data. Social Networks, 26, 289–307.
Marsden, P. V. (1987). Core discussion networks of Americans. American Sociological
Review, 52, 122–131.
McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2006). Social isolation in America: Changes in core discussion networks over two decades. American Sociological
Review, 71, 353–375.
McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2008). Erratum: Social isolation in America: Changes in core discussion networks over two decades. American
Sociological Review, 73, 1022.
McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2009). Models and marginals:
Using survey evidence to study social networks. American Sociological Review, 74,
670–681.
Paik, A., & Sanchagrin, K. (2013). Social isolation in America: An artifact. American
Sociological Review, 78, 339–360.
Roberts, S. G. B., Dunbar, R. I. M., Pollet, T. V., & Kuppens, T. (2009). Exploring variation in active network size: Constraints and ego characteristics. Social Networks,
31, 138–146.
Tufecki, Z. (2010). Who acquires friends through social media and why? ‘Rich Get
Richer’ versus ‘Seek and Ye Shall Find’. Proceedings of the 4th International Conference on Weblogs and Social Media (ICWSM). AAAI Press.
Wellman, B., & Wortley, S. (1990). Different strokes from different folks: Community
ties and social support. The American Journal of Sociology, 96, 558–588.

FURTHER READING
Fischer, C. S. (2011). Still connected: Family and friends in America since 1970. New York,
NY: Russell Sage Foundation.
Kadushin, C. (2012). Understanding social networks: Theories, concepts and findings.
Oxford, England: Oxford University Press.
Putnam, R. D. (2001). Bowling alone: The collapse and revival of American community.
New York, NY: Simon and Schuster.
Rainie, L., & Wellman, B. (2012). Networked: The new social operating system. Cambridge, MA: MIT Press.

MATTHEW E. BRASHEARS SHORT BIOGRAPHY
Matthew E. Brashears is an Assistant Professor of Sociology at Cornell University. In addition to his work on ego networks and average

12

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

network size, he is interested in the connections between social networks and social psychology. His current funded research includes an
NSF-sponsored study linking cognition and memory to social network
structure and a Department of Defense-sponsored effort to develop new
ways of detecting terrorist groups preparing chemical, biological, radiological, and nuclear attacks. His newest research is on the mutation of
information as it diffuses through social networks. Personal webpage:
http://www.soc.cornell.edu/faculty/brashears.html
LAURA AUFDERHEIDE BRASHEARS SHORT BIOGRAPHY
Laura Aufderheide Brashears is a Visiting Scholar at Cornell University.
A sociological social psychologist, her research sits at the intersection of
inequality and the self. She is currently examining how the strength of
one’s racial and ethnic identities influences prejudice toward racial/ethnic
out-groups and preference for one’s racial/ethnic in-group. Her past research
includes substantial study into the effects of self-esteem on academic identities and academic achievement in African American, Mexican American,
and White youth, as well as the influence of nonparental significant others
on adolescent self-esteem.
RELATED ESSAYS
Emotion and Intergroup Relations (Psychology), Diane M. Mackie et al.
Emergence of Stratification in Small Groups (Sociology), Noah Askin et al.
Problems Attract Problems: A Network Perspective on Mental Disorders
(Psychology), Angélique Cramer and Denny Borsboom
The Development of Social Trust (Psychology), Vikram K. Jaswal and Marissa
B. Drell
Herd Behavior (Psychology), Tatsuya Kameda and Reid Hastie
Cultural Neuroscience: Connecting Culture, Brain, and Genes (Psychology),
Shinobu Kitayama and Sarah Huff
Social Classification (Sociology), Elizabeth G. Pontikes
Social Relationships and Health in Older Adulthood (Psychology), Theodore
F. Robles and Josephine A. Menkin
Emerging Trends: Shaping Age by Technology and Social Bonds (Sociology),
Annette Spellerberg and Lynn Schelisch
Social Neuroendocrine Approaches to Relationships (Anthropology), Sari M.
van Anders and Peter B. Gray


Close Friendships among
Contemporary People
MATTHEW E. BRASHEARS and LAURA AUFDERHEIDE BRASHEARS

Abstract
Do contemporary people have fewer friends than they used to? In this entry, we
examine the research on strong tie networks, or networks containing people especially important to us, in order to provide a partial answer. We begin with a review
of the methods employed to collect data on social network connections. We then summarize the literature on the size of Americans’ social networks as well as change in
this size over time. Finally, we conclude with an analysis of where the study of strong
networks needs to go and some suggestions for getting it there. Overall, the research
suggests that while strong networks have changed over the past 30–40 years, and
may be smaller overall, people are no more isolated than they were in the past.

INTRODUCTION
Human life takes place within a web, or network, of interpersonal relationships. Some of these relationships (i.e., network ties) are relatively weak,
ranging from the mere awareness of another’s existence to acquaintanceship. These relatively weak relationships typically do not provide us with
emotional support or other resources, but can provide us with knowledge
that our closer associates lack. Other relationships are strong, including close
friends, family members, and spouses. These relatively strong relationships
provide us with major support when we are in need and with companionship when we are lonely, but rarely give us access to knowledge that we could
not obtain elsewhere. Our relationships, both strong and weak, also impose
costs directly through the time and mental effort required to maintain contact
with others, and indirectly via the loans of goods and services we provide to
our associates. Finally, our social networks expose us to risks that would not
exist without them (e.g., we cannot be betrayed until we have an associate
whom we trust). This mixture of costs, benefits, and risks makes social networks a defining element of human life, and it is crucial to social science that
we understand them.
Emerging Trends in the Social and Behavioral Sciences. Edited by Robert Scott and Stephen Kosslyn.
© 2015 John Wiley & Sons, Inc. ISBN 978-1-118-90077-2.

1

2

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Much effort has been expended trying to understand one of the most basic
characteristics of strong tie networks: their size. How large is the average
individual’s social network of important people, and has this number been
increasing, decreasing, or holding steady over time? In other words, do contemporary people have fewer friends, or fewer important people in their
lives, than they used to? By understanding the size of strong tie networks,
and the tendency for these networks to change, we gain a better understanding of how human society functions and how individuals experience the
world. Moreover, the explosion of technologies connecting us to one another
makes such knowledge all the more critical.
This entry begins by reviewing the methods used to obtain estimates of the
size of strong tie networks. We then summarize the foundational research on
typical network size as well as longitudinal research exploring how network
size has changed over the past 30–40 years; on the whole, this research provides consistent signs that portions of our strong tie networks may be smaller
than they used to be, but these declines are probably not as severe as was once
feared, and are far from unambiguous. Lastly, we cover current cutting edge
research in the area, discuss key issues where further investigation is needed,
and conclude with our thoughts on the current state of the research.
FOUNDATIONAL RESEARCH
MEASURING FRIENDSHIP NETWORKS
How does one measure the size of a social network? In general three
approaches have been employed: ego network methods, sociometric methods, and the network scale-up approach. In an ego network data collection,
the researcher begins with a probability sample of some population. While
there are many ways of drawing such a sample, the end result is a set of
persons that reflect the group under study; thus a national sample will
reflect, or be representative of, the national population as a whole. Individuals selected by the sampling procedure (i.e., “egos”) are then invited to
participate in a survey that includes a “name generator,” or a question meant
to elicit a set of names that meet a set of criteria (e.g., “Who have you spoken
to in the last 8 h?”). Once a list of names has been obtained, a second set of
questions known as “name interpreters” are then often asked (e.g., “What is
this person’s sex?”) about each person named (i.e., “alter”), or a subset of the
people named. Because the egos are a representative sample, the resulting
data allow researchers to develop a representative estimate of certain
characteristics of personal networks (e.g., average network size and average
diversity) at the level of the population. At the same time, since the egos
are drawn from a larger population, few, if any, of the individuals sampled

Close Friendships among Contemporary People

3

will be directly connected to each other. Thus, we cannot understand the
complete structure of any one network using ego network data collection.
Research also shows that egos have elaborate systems of categorization for
their associates, viewing some as appropriate for recreational contact, some
as appropriate for emotional support, and some as appropriate for loans of
material assistance (e.g., Wellman & Wortley, 1990). As a result, the number
of names obtained is exquisitely sensitive to the specific name generator
asked (Bernard, Shelley, & Killworth, 1987; Marin, 2004). Put differently, a
name generator asking about whom you spend free time with may produce
a very different list than one asking about whom you would borrow a large
sum of money from, even though all persons named may be strongly tied
to you. Thus, size estimates using different types of name generators must
be compared only cautiously and the most appropriate comparisons are
between studies using the same name generator(s).
In a sociometric data collection, all individuals in a specific group (e.g., a
neighborhood or a school) are sampled, with the goal of illuminating the
entire network structure. This method often involves a roster of names,
allowing respondents to simply check off those others with whom they are
connected in particular ways, although such a roster is not required. This
approach is superior to an ego network sample in revealing the overall
structure of a social network, but there are several limitations. First, a sociometric approach requires the researchers to identify all the members of a
particular group before starting data collection. Second, because this method
is not based on sampling, a much higher proportion of the study population
must participate in order for the data to be valid. In both cases, these goals
may not be practical or even possible to achieve. Third, sociometric data
are of limited use in estimating overall population characteristics because
the characteristics of a subset do not necessarily match the characteristics
of a superset (e.g., the average network size in a neighborhood is not
necessarily the same as for a nation). This final difficulty may be declining
in significance, however, as social media makes national-level network data
available to researchers.
Finally, network size has been measured with what is often known as
a “scale-up” method. This approach begins much like an ego network with
a probability sample, but the researcher asks a name generator focused on
a population segment whose prevalence is already known. For example,
the researcher could ask, “How many people do you know who are named
“Jonathon”?” Since it is possible to determine the population prevalence
of people with this name (e.g., voter registration and phone books), the
researcher can use the answer to estimate the total size of the respondent’s
network assuming random mixing. This approach to estimating network
size imposes a relatively low burden on the respondent, but the assumption

4

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

of random mixing is not necessarily valid. For example, individuals tend
to associate with those similar to themselves (i.e., “Birds of a feather flock
together”) in a process known as homophily, thus preventing random
mixing and introducing bias into this estimation method.
All three of these approaches can be implemented actively or passively.
Active data collection involves an explicit request to individuals for the
needed information (e.g., a name generator or a roster). This method
is frequently used, both because it is practical and because it gives the
researcher more control over which ties are measured, but is vulnerable
to respondent forgetting, recall biases, and simple fatigue. In passive data
collection, the researcher either directly observes individuals having contact
with others (e.g., observes interactions between students in a classroom), or
obtains technological traces of past interactions (e.g., email, Facebook, and
phone records). In either case, individuals reveal their network ties through
their behavior rather than by answering a set of explicit questions. To the
extent that an entire social network is available in an electronic form (e.g.,
Facebook), it may be possible to capture even a massive network passively,
providing a wealth of data to the researcher. However, it is often difficult
to reliably distinguish different types of relationships using passive data
collection (e.g., not all “friends” on Facebook represent an individual’s
strong ties, even though the title “friend” implies that they might), and the
resulting data are often limited to one method of contact (e.g., email).
CLASSICAL ESTIMATES OF NETWORK SIZE
Since the 1970s, considerable attention has been devoted to estimating network size. De Sola Pool and Kochen (1978) considered the problem in the first
issue of the flagship journal for social networks (i.e., Social Networks), estimating that the average network contains several hundred ties, both strong and
weak. This conclusion was based on an informal meta-analysis of existing
studies, as well as a compilation of data sources with very different characteristics, leaving considerable room for error, and distinctions were not made
between various types of associates. A more comprehensive effort to measure network size was made by Claude Fischer (1982), who collected data
from respondents in and around an urban area in California using name generators to reveal the types of benefits people could obtain through strong
ties, including mutual favors, socializing, and emotional support. While this
effort was far more rigorous than its predecessors, it was geographically limited, and thus unable to provide estimates of network size among the larger
national population.
Building on Fischer’s work, the 1985 General Social Survey (GSS), a probability sample of the US noninstitutionalized adult population conducted

Close Friendships among Contemporary People

5

every year or every other year, included a single name generator that has
since become a standard in network research examining strong ties. This
name generator, known as the “important matters” item, asks respondents to
name those persons with whom they have discussed important matters during the preceding 6 months. No definition is given for an “important matter,”
because respondents might differ on what they consider to be “important,”
but they will most likely discuss important things with their strong ties. Thus,
the important matters item provides a way to measure strong tie networks.
The GSS name generator for the first time permitted an estimation of typical
strong network size nationwide. Using these data, Marsden (1987) found that
in 1985 Americans discussed “important matters” with an average of three
others, approximately half of whom were kin and half of whom were nonkin
(including friends). Obviously, these results do not mean that the average
American only has 1.5 friends, but the “important matters” item captures
those relationships that are relatively strong, and thus these results suggested
that in 1985 our strong network of important confidantes was relatively small
and evenly split between kin and nonkin.
It should be noted that other lines of research have produced different, and
sometimes much larger, estimates of strong tie network size than the GSS,
ranging from 3 to 150 (e.g., Bernard et al., 1990; Roberts, Dunbar, Pollet, &
Kuppens, 2009). However, because of the sensitivity of estimates to differences in name generator phrasing, it is difficult at best to compare many of
these results to each other. As such, we focus on the estimates deriving from
the GSS, which are both nationally representative, and have been repeated
longitudinally in several independent studies.
STUDIES OF NETWORK CHANGE
While the 1985 GSS provided a crucial look at the social environment of the
average American, and the first nationwide estimates of strong network
size, it was unable to provide any sense of how stable these environments
were. While a partial, and only semi-comparable repetition, of the network
module was included in the 1987 GSS, it was not until 2004 that the GSS
included a full repeat of the network module. Analysis of these data by
McPherson, Smith-Lovin, and Brashears (2006) proved to be quite surprising: whereas in 1985 Americans had an average of three others with
whom they discussed important matters, 1.4 of which were kin, in 2004
this number had declined to two, 0.88 of which were kin. Moreover, while
in 1985 only 10% of respondents gave no names in response to the name
generator, by 2004 approximately 25% of all respondents appeared to be
“socially isolated.” These results suggested that American strong networks
had declined in size by roughly one-third, and that social isolation had

6

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

more than doubled, in approximately 20-years. These results implied that
American social life was growing more fragile but a major puzzle remained:
what could be causing such a dramatic change in association? McPherson,
Smith-Lovin, and Brashears were unable to definitively identify a cause,
but speculated that the change was linked to the growth of the internet and
mobile phones, which might permit Americans to rely more exclusively
on a smaller number of associates. Nevertheless, the existing data were
insufficient to test their speculations and the decline remained unexplained.
One possible explanation for the decline, advanced by Claude Fischer
(2009), was that the data were simply wrong. He argued that the apparent
decline in network size in 2004 was attributable to one or a combination of
three factors: a training effect (i.e., respondents learn how to avoid lengthy
follow-up questions), a fatigue effect (i.e., respondents give incomplete
answers because they are tired), or a computer error. Using a set of similar,
but not identical, items on network size in the GSS data, Fischer argued
that the GSS “important matters” results were almost certainly an artifact of
these potential flaws. Thanks to Fischer’s concerns, a number of miscoded
responses in the original data were identified, but reanalysis of the GSS
data (McPherson, Smith-Lovin, & Brashears, 2008) using updated weighting
procedures continued to show that strong networks in 2004 consisted of
roughly two people, 1.39 of whom were kin, and that approximately 23%
of the respondents were socially isolated. McPherson, Smith-Lovin, and
Brashears (2009) likewise defended their position, showing that artifacts
almost certainly existed in both the 1985 and 2004 datasets, but that the
downward trend over time persisted even when controlling for these
artifacts. While this debate provided a public airing of many of the relevant
issues, it ultimately settled very little; the decline in network size had been
called into question, but no “smoking gun” had been identified.
CUTTING-EDGE RESEARCH
Recently several attempts have been made to resolve the uncertainty over
network size. While McPherson et al. speculated that the rise of the internet and other communications technologies might have led to a reduction
in strong network size, accumulating evidence suggests that this is not the
case. Tufecki (2010) finds that while online interaction does not consistently
increase sociability, it also does not consistently decrease it. Hampton, Sessions, and Her (2011) used a nationally representative phone survey conducted in 2008 to show that internet and communications technology usage
either has no effect, or a slight positive effect, on both network size and sociability. It thus is unlikely that the reduction in strong network size, if real, is
the result of the increasing usage of communications technologies.

Close Friendships among Contemporary People

7

In addition to evaluating the effect of internet usage on sociability, Hampton et al.’s (2011) study also included the same “important matters” name
generator that appeared on the 1985 and 2004 GSS administrations. Examination of these data reveals that only 12% of respondents reported no discussions of important matters in the preceding 6 months, a substantial decline
from the 23% detected by McPherson et al. (2008) over a period of only 4
years. These same data, however, indicated an average strong network size of
approximately two, with 1.19 of these being kin. Thus, while Hampton et al.
found a different level of social isolation, the estimated average size of the
network in 2008 was very similar to that detected by McPherson et al. (2006,
2008). A second study by Hampton, Goulet, Marlow, and Rainie (2011) using
a similar methodology and data collected in 2011 also yielded an average
strong network size of approximately two.
Research by Brashears (2011) investigated the artifacts proposed by Fischer
(2009) directly. This study used a general population experiment to evaluate the potential for training and fatigue effects to have reduced apparent
network size. In addition, because the data were collected from a representative sample of the national population in 2010, it is possible to use these
data to construct national estimates. The results failed to support the training or fatigue effect explanations for the reduction in network size, and once
more found an average strong network size of approximately two. However,
this research also found that only a little more than 4% of the population was
“socially isolated,” which is two-thirds less than Hampton et al.’s (2011) findings from 2008 and roughly one-fifth the level of social isolation observed by
McPherson et al. (2008) in the 2004 GSS. This suggests that while the average
strong network size is relatively stable over time, the level of social isolation varies wildly and is quite likely unreliable. Brashears (2011) presented
additional evidence to this effect, finding that of those who reported no discussions of important matters, roughly two-thirds said that this was because
they had no “important matters” to talk about, rather than no one with whom
to discuss them.
Finally, research by Paik and Sanchagrin, currently 2013, takes a new and
innovative approach to the debate by suggesting that the decrease in strong
network size may have been an artifact of fatigue among the interviewers
rather than fatigue among the respondents. They find that high levels of
social isolation are not distributed evenly by interviewer, but rather are significantly more likely to be observed by specific interviewers. They suggest
that perhaps interviewers guided respondents into answering that they had
no discussions of important matters in order to shorten the interviews. This
novel explanation is consistent with existing research showing that respondent artifacts probably did not produce the shrinking networks finding. At
the same time, however, it is unclear why fatigue would suddenly be an

8

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

issue in 2004, but not in 1985 when interview procedures were more cumbersome overall. Likewise, it does not account for Hampton et al.’s (2011)
results using a much shorter instrument, or Brashears’ (2011) results using
a computer-assisted self-interview method that does not include an interviewer.
This recent research has thus advanced our understanding of strong network size, but has not completely answered the question: are strong networks smaller now than they were before?
KEY ISSUES FOR FUTURE RESEARCH
The size of the average strong network is a matter of profound concern for
social science, but it remains, for the moment, in considerable doubt. It is
therefore also uncertain whether strong networks have declined in size in
recent decades, although several independent data collection efforts have
given us reason to think that they have. The controversy over so fundamental
an issue serves as a pointed reminder that we may not know as much about
how network items map onto social reality as we thought, and that much
basic research remains to be done.
One of the most crucial questions is: how do respondents mentally represent their social networks? While multiple studies indicate that strong
networks are smaller now than in the past, convincing evidence indicates
that people with whom we have strong ties may not all provide us with
the same types of benefits (Wellman & Wortley, 1990). In other words, if the
way respondents classify their alters changes, it may give the appearance
of a change in network size, without the substance of the network structure
actually changing very much at all. This concern is underscored by the
fact that discussions of “important matters” are apparently often regarding
topics that many would consider trivial (Bearman & Parigi, 2004). This
implies that the name generator may, to some extent, not be capturing actual
discussions, but rather a subset of the network that the respondent views as
suitable for important discussions. Although new media technologies, such
as Facebook and Twitter, have not led to rampant social isolation, they may
be contributing to a transformation of how we categorize our associates,
including our friends and discussion partners. Research in social psychology
has indicated that the meaning of social categories can both change over
time, and differ between social groups, and technology may play a role in
these differences. Thus, natural changes in how relationships are understood
may result in respondents and researchers defining ties in disparate ways.
Without similar definitions, researchers may end up with biased estimates,
possibly explaining the confusion over network size and social isolation.

Close Friendships among Contemporary People

9

Social isolation, or the lack of any contacts, has received considerable attention, particularly given its extraordinarily high level in the 2004 GSS data.
However, subsequent data collections have failed to replicate this level of
isolation even while finding very comparable average network sizes. This
suggests that what may determine the level of social isolation is less the availability of associates with whom to discuss important matters, but rather the
availability of topics to have discussions about. When topics are relatively
hard to come by, social isolation may appear to increase. Researchers therefore need to work harder to distinguish measurements of the existence of a
relationship from measurements of particular uses of that relationship.
The instability in the estimates of social isolation emphasizes an additional
issue: that the concern over network size in general, and isolation in particular, may be somewhat misguided. The real question is not whether an
individual has any associates, but whether they have enough to provide the
support that they need. Our attention should be on social poverty, or an inadequate level of social support, rather than on social isolation, or no social
support whatsoever. It is likely the case that few individuals are truly isolated, but many may suffer under some degree of social poverty.
As the debate over the shrinking of American strong networks has made
painfully clear, we have a poor understanding at present of how networks
vary naturally in size over time. Certainly networks change over time with
individuals forming new friendships and losing old ones. At the population
level, there may be natural trends to the gain and loss of associates, and
these trends may vary over long periods of time, possibly in cyclical patterns. While many network studies have been carried out over the decades,
they often involve different populations, different name generators, or both,
and thus can tell us little about these natural variations. As such, even if the
decline in strong network size observed in much of the above research is real,
that does not mean that there is cause for alarm: over a span of decades, average strong network size may vary naturally by a considerable amount. The
simple reality is that, at the present time, we do not know if such natural variation exists and, if so, how extreme it may be, and thus cannot know if any
particular increase or decrease in network size is a virtue or a vice.
A promising medium for collecting such data is the utilization of passive data collection using social media and other electronic traces. These
approaches have real advantages in that they avoid many of the potential
biases introduced by question interpretation that have plagued network
research to date, and thus represent a valuable approach to measuring
network size. At the same time, however, it is unclear how relationships
recorded in social media (e.g., Facebook or Twitter) connect to the important
relationships of individuals. Some important associates may be contacted
rarely via social media, while more trivial contacts are interacted with

10

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

frequently using these technologies. It is thus critical that researchers
develop both expertise with these new sources of data, as well as explore the
connections, or differences, between them and more traditional measures.
CONCLUSION
A considerable amount of research effort has been expended on measuring
the size of modern strong tie networks and determining whether they have
grown smaller over the previous several decades. Unfortunately, the best
that we can say with the existing research is that they may have shrunk, but
that we cannot be sure. The early research on long-term change in network
size suggests that networks are smaller, and several independent data collection efforts since have also shown smaller networks, but the distinctive
social isolation finding has failed to replicate, and there are growing signs
that methodological artifacts may be partly to blame. All we can say with
certainty at this point is that more research, both fundamental and applied,
is needed in order to resolve this question. It is the privilege of authors,
however, to offer their own opinions and ours is this: strong tie networks
are different now than they were nearly three decades ago, but they are not
necessarily smaller. As our society has changed, so too have our definitions
of friend, confidant, and associate transformed to fit the new demands, and
constraints, of our lives. It may indeed be the case that we have fewer of
those persons with whom we might have once discussed “important matters,” be they kin or friends, but it is our view that modern Americans remain
intensely social, even if our relationships have been evolving into endless
forms most wonderful.
REFERENCES
Bearman, P., & Parigi, P. (2004). Cloning headless frogs and other important matters:
conversation topics and network structure. Social Forces, 83, 535–557.
Bernard, H. R., Johnsen, E. C., Killworth, P. D., McCarty, C., Shelley, G., & Robinson, S. (1990). Comparing four different methods for measuring personal social
networks. Social Networks, 12, 179–215.
Bernard, H. R., Shelley, G. A., & Killworth, P. (1987). How much of a network does
the GSS and RSW dredge up? Social Networks, 9, 49–61.
Brashears, M. E. (2011). Small networks and high isolation? A reexamination of
American discussion networks. Social Networks, 33, 331–341.
De Sola Pool, I., & Kochen, M. (1978). Contacts and influence. Social Networks, 1, 5–51.
Fischer, C. S. (1982). To dwell among friends: Personal networks in town and city. Chicago,
IL: University of Chicago Press.
Fischer, C. S. (2009). The 2004 GSS finding of shrunken social networks: An artifact?
American Sociological Review, 74, 657–669.

Close Friendships among Contemporary People

11

Hampton, K. N., Goulet, L. S., Rainie, L., & Purcell, K. (2011). Social networking sites and our lives. Pew Internet and American Life Project, June 16,
2011. Retrieved from http://www.pewinternet.org/Reports/2011/Technologyand-socialnetworks.aspx.
Hampton, K. N., Sessions, L. F., & Her, E. J. (2011). Core networks, social isolation,
and new media: Internet and mobile phone use, network size, and diversity. Information, Communication & Society, 14, 130–155.
Marin, A. (2004). Are respondents more likely to list alters with certain characteristics? Implications for name generator data. Social Networks, 26, 289–307.
Marsden, P. V. (1987). Core discussion networks of Americans. American Sociological
Review, 52, 122–131.
McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2006). Social isolation in America: Changes in core discussion networks over two decades. American Sociological
Review, 71, 353–375.
McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2008). Erratum: Social isolation in America: Changes in core discussion networks over two decades. American
Sociological Review, 73, 1022.
McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2009). Models and marginals:
Using survey evidence to study social networks. American Sociological Review, 74,
670–681.
Paik, A., & Sanchagrin, K. (2013). Social isolation in America: An artifact. American
Sociological Review, 78, 339–360.
Roberts, S. G. B., Dunbar, R. I. M., Pollet, T. V., & Kuppens, T. (2009). Exploring variation in active network size: Constraints and ego characteristics. Social Networks,
31, 138–146.
Tufecki, Z. (2010). Who acquires friends through social media and why? ‘Rich Get
Richer’ versus ‘Seek and Ye Shall Find’. Proceedings of the 4th International Conference on Weblogs and Social Media (ICWSM). AAAI Press.
Wellman, B., & Wortley, S. (1990). Different strokes from different folks: Community
ties and social support. The American Journal of Sociology, 96, 558–588.

FURTHER READING
Fischer, C. S. (2011). Still connected: Family and friends in America since 1970. New York,
NY: Russell Sage Foundation.
Kadushin, C. (2012). Understanding social networks: Theories, concepts and findings.
Oxford, England: Oxford University Press.
Putnam, R. D. (2001). Bowling alone: The collapse and revival of American community.
New York, NY: Simon and Schuster.
Rainie, L., & Wellman, B. (2012). Networked: The new social operating system. Cambridge, MA: MIT Press.

MATTHEW E. BRASHEARS SHORT BIOGRAPHY
Matthew E. Brashears is an Assistant Professor of Sociology at Cornell University. In addition to his work on ego networks and average

12

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

network size, he is interested in the connections between social networks and social psychology. His current funded research includes an
NSF-sponsored study linking cognition and memory to social network
structure and a Department of Defense-sponsored effort to develop new
ways of detecting terrorist groups preparing chemical, biological, radiological, and nuclear attacks. His newest research is on the mutation of
information as it diffuses through social networks. Personal webpage:
http://www.soc.cornell.edu/faculty/brashears.html
LAURA AUFDERHEIDE BRASHEARS SHORT BIOGRAPHY
Laura Aufderheide Brashears is a Visiting Scholar at Cornell University.
A sociological social psychologist, her research sits at the intersection of
inequality and the self. She is currently examining how the strength of
one’s racial and ethnic identities influences prejudice toward racial/ethnic
out-groups and preference for one’s racial/ethnic in-group. Her past research
includes substantial study into the effects of self-esteem on academic identities and academic achievement in African American, Mexican American,
and White youth, as well as the influence of nonparental significant others
on adolescent self-esteem.
RELATED ESSAYS
Emotion and Intergroup Relations (Psychology), Diane M. Mackie et al.
Emergence of Stratification in Small Groups (Sociology), Noah Askin et al.
Problems Attract Problems: A Network Perspective on Mental Disorders
(Psychology), Angélique Cramer and Denny Borsboom
The Development of Social Trust (Psychology), Vikram K. Jaswal and Marissa
B. Drell
Herd Behavior (Psychology), Tatsuya Kameda and Reid Hastie
Cultural Neuroscience: Connecting Culture, Brain, and Genes (Psychology),
Shinobu Kitayama and Sarah Huff
Social Classification (Sociology), Elizabeth G. Pontikes
Social Relationships and Health in Older Adulthood (Psychology), Theodore
F. Robles and Josephine A. Menkin
Emerging Trends: Shaping Age by Technology and Social Bonds (Sociology),
Annette Spellerberg and Lynn Schelisch
Social Neuroendocrine Approaches to Relationships (Anthropology), Sari M.
van Anders and Peter B. Gray



Media

Close Friendships among Contemporary People