Oxford Internet Institute, Research Report No.

8, November 2005

A study of adolescents’ online and

offline social relationships

Gustavo Mesch
Department of Sociology and Anthropology
University of Haifa, Har Hacarmel 31905, Israel
Email: [email protected]
Website: http://soc.haifa.ac.il/~gustavo

© The Oxford Internet Institute 2006

A study of adolescents’ online and offline social relationships

Abstract

Recent studies have shown that adolescents use the Internet not only to maintain
social relationships with distant relatives and friends but also to create new
relationships online; some of these friendships become integrated into their social
circle. Past studies focused mainly on the effect of the Internet on existing
relationships or the nature of online-only ties, so studies comparing the quality of
online and face-to-face relationships are missing. The goal of the current study is
to bridge this gap. In keeping with previous studies on social association, I argue
that the quality of social relationships is dependent on duration and diversity of
topics and activities carried out together. Time is important as it facilitates the
development of a collective shared history and identity. Intimacy develops through
the participation in shared activities and discussion of diverse issues of personal
concern. Using a representative sample of the adolescent population in Israel, it
was found that closeness to a friend is a function of social similarity, content and
activity multiplexity and duration of the relationships. Friendships originated in the
Internet are perceived as less close and supportive because they are relatively
new and online friends are involved in fewer joint activities and fewer topics of
discussion. The implications of the findings are discussed.
As the proportion of households in the population of Western countries gaining
access to the Internet is increasing, empirical evidence is accumulating that the
Internet is becoming more and more integrated in individuals’ everyday life,
including the formation and maintenance of intimate and non-intimate social
relationships (Wellman and Giulia, 1999; Haythornthwaite and Wellman, 2002).
Early studies reflected a concern with decreasing social involvement and
compared Internet users and non-users in the extent of involvement with existing
social relationships. The results on the impact of the Internet on existing
relationships were mixed. Some found a decrease in the involvement with
previous ties (Kraut et al., 1998; Nie et al., 2002) but others have shown that
Internet use does not affect involvement in close relationships and the community
(Katz and Rice, 2002; Hampton and Wellman, 2003; Mesch and Levanon, 2003)
and even supports and maintains relationships with friends and family after
moving to a new location (Cummings et al., 2004). Other studies restricted
themselves to the study of online social relationships only, documenting the
existence of supportive, intimate and personal relationships online (McKenna and
Bargh 1998; Walther and Boyd, 2002).
Lately, empirical evidence has shown that individuals use the Internet not only for
existing close ties but also to create new relationships in which companionship,
social support and information exchange take place. In some cases these online
relationships become incorporated into the Internet users’ face-to-face social
circle (Parks and Floyd, 1996; Hampton and Wellman, 2002; Mesch and
Levanon, 2003; Wolak et al., 2003; Mesch and Talmud, 2004). The effect of the
Internet on existing relationships has been extensively studied, but the literature is
wanting in the comparative study of the quality of personal relationships created
online and those created in face-to-face settings. The goal of the current study is

2
 Gustavo Mesch

to fill this gap. Using recently collected data of a representative sample of the
adolescent population in Israel, the differential quality of personal relationships
created online and face-to-face was investigated.
Keywords: computer mediated communication, online and offline social networks,
Internet use, strength of ties

Study overview

The conceptual model guiding the current study assumes that association with
others, and the quality of this association, is a socially structured process.
Individuals and families differ in their human, economic and social capital, and
according to the level of these resources we engage in daily activities (work,
study, leisure) in different contexts in which we associate with others. The
likelihood of association with others is higher when others are similar in social
characteristics because we belong to the same social contexts, are subjected to a
similar socialization process and therefore develop similar interests and concerns.
Thus, social similarity is a salient factor, as individuals that share a similar social
status such as age, gender and place of residence, are more likely to share
interests and concerns that facilitate the formation of common ground for social
exchange. Once a relationship has been established the strength of that
relationship is dependent on the ability of individuals to be involved in common
activities and intimate conversations. In part, this ability is a function of
relationship duration, as intimacy and participation in common activities requires
time. In that sense, the model departs from communication models that
emphasize the role of channels of communication. Rather, the choice of channel
is seen as partially dependent on social characteristics of the relationship, such
as similarity and intimacy.
In order to test the current model, a survey of a representative sample of Israeli
adolescents (n=996) was conducted. Adolescents aged 12-18 were asked to
provide information on the extent to which they have access to the Internet, and
their patterns of use. In addition, respondents were asked to provide information
regarding their friends. Using an ego network technique, adolescents provided
information on whether the friend was met face-to-face (school, neighborhood) or
online (chat rooms, email or Instant messenger) and whether they are similar in
terms of age, gender and place of residence. Friends met online and in face-to-
face settings were compared in terms of social similarity, relationship duration and
relationship content. The findings show that online ties are more distant than
offline ties because they are less developed than face-to-face ties. The length of
association with online friends is shorter than with friends who were met face-to-
face, indicating that fewer opportunities for intimate discussion and participation
and joint activities explain the relational quality differences.
The structure of the paper is as follows. First, a review of literature on adolescent
friendship is presented. This period in life is one in which the social circle of the
individual expands rapidly outside the family, thus presenting an opportunity for

3
A study of adolescents’ online and offline social relationships

the study of the sources of relationship formation and its quality. Second, central
findings of the literature on online and offline relationships are presented as a
basis to the conceptual model guiding the current study. Then, the study
methodology and findings are presented followed by a discussion of the
implications of the theoretical implications of the findings.

Personal relationships during adolescence

The current study focuses on the nature of social relationships during

adolescence. During this period, social relationships outside the family expand
and their quality has been linked to various behavioral outcomes (Giordano,
2003). Social interaction with peers provides a forum for learning and refining the
socio-emotional skills needed for enduring relationships. Through interactions
with peers, adolescents learn how to cooperate, to take different perspectives,
and to satisfy growing needs for intimacy (Rubin et al., 1998; Crosnoe, 2000).
Youths who report having friends are more confident, more altruistic, and less
aggressive, and demonstrate greater school involvement and work orientation
(Hartup and Stevens, 1997).
Youniss and Smollar (1985) have argued that adolescents’ friends are intimate
and more accepting than parents, who are necessarily more oriented toward the
future and more concerned with the potentially negative consequences of their
child’s behavior. This greater level of acceptance helps explain the high levels of
self-disclosure and mutual trust that often develop for the first time at this age,
and that are characteristics of close friendship ties (Giordano, 2003). In that
sense, personal relationships are for adolescents a type of social support. Those
with more supportive friendships were shown to have higher self-esteem, to suffer
depression or other emotional disorders less often, and to be better adjusted to
school than youths with less supportive friendships (Berndt et al., 1989; Hartup
and Stevens, 1997; Collins et al., 1999; Beraman and Moody, 2004). The rapid
growth of friendship at adolescence provides a unique setting for the comparative
analysis of friendships created online and offline that is addressed in this study.

The quality of face-to-face and online social ties

The literature on personal relations has long been concerned with the quality of
the ties that bind individuals. One way to measure this quality is by the strength of
these ties (Marsden and Campbell, 1984). A tie’s strength is usually assessed by
means of a combination of factors such as perceived closeness, intimacy and
trust. Weaker ties are evinced in more casual relationships and in sparser
exchanges; they typify relationships of those who enjoy fewer kinds of support.
Strong ties exist in relationships on a high level of intimacy, involving more self-
disclosure, shared activities, emotional as well as instrumental exchanges, and

4
 Gustavo Mesch

long-term interaction (Marsden and Campbell, 1984; Haythornthwaite and

Wellman, 2002).
Studies on the quality of online relationships are divided in their conclusions
regarding the qualities of social ties that are created and maintained through the
Internet. Positions on the social impact of the Internet derive from arguments
regarding ‘social affordability’, dominated by two contrasting views,
communication channel determinism and social constructivism. Early
conceptualizations, assuming the technological qualities of the Internet, described
the weakness of electronic media, as a communication channel, in supporting
social ties. The ‘reduced social cues perspective’ is based on the observation that
computer-mediated communication (CMC) allows the exchange of fewer cues
than face-to-face environments and suggests that CMC is less appropriate for the
support of emotional exchanges or the conveyance of complex information and a
sense of social presence. This early perspective is quite skeptical of the ability of
CMC to support strong ties. Moreover, precisely because CMC provides access
to a wider audience of individuals who may share interests and hobbies, it has
been suggested that the reduced social cues environment on which CMC is
based is more suited for supporting weak ties by reducing the risks associated
with contacting unknown others (Sproul and Kiesler, 1986; Rice and Love, 1987).
While recognizing that computer-mediated communication is often impersonal,
Walther has argued that in many cases it becomes hyper-personal leading to
social ties that are intimate, supportive and emotionally meaningful. According to
this approach, through time online ties develop a shared system of clues, using
expressions and emoticons that overcome the lack of social presence (Walther,
1996).
Social constructivists, by contrast, argue that some features of online
communication, such as anonymity, isolation, lack of ‘gating features’, and ease
of finding others with the same interests, make it easier for individuals to form
strong ties (Joinson, 2001; McKenna et al., 2002). The formation of close
interpersonal relationships requires the establishment of trust, that is, a sense that
intimate information disclosed in interpersonal exchanges is not widely
disseminated and is not used to ridicule friends. The relative anonymity of the
Internet reduces the risks of such disclosure, especially intimate information,
because such intimate information can be shared with no fear for embarrassment
resulting from disclosing intimate information to members of the close-knit, often
transitive, face-to-face social circle (McKenna et al., 2002).
Empirical evidence for these perspectives is mixed. A few studies report that the
quality of online social interactions and relationships is lower than that of face-to-
face interactions. Employees of a multinational bank reported that email
communication was less reliable than face-to-face. In another study, college
students evaluated email communication as inferior to communication in person
as a means of maintaining personal relationships (Cummings et al., 2002). In
other words, offline friends are perceived as closer because the frequency of
communication with face-to-face friends is higher than with online friends.
Other studies, however, have shown that people often disclose intimate
information about themselves online (Joinson, 2001; McKenna et al., 2002). The

5
A study of adolescents’ online and offline social relationships

high levels of self-disclosure in CMC interactions proved to be related to

anonymity (Joinson, 2001). Individuals who disclosed personal and intimate
information over the Internet reported greater closeness to their online friends
(McKenna et al., 2002). Yet, none of these directly compared the quality of online
and face-to-face relationships.
It is the argument of this paper that the quality of social ties is heavily dependent
not only on the place where friends meet, but also on social similarity, which were
not studied in previous studies. While I do not underestimate the relevance of
communication channels in shaping some aspects of social relationships, the
selection of communication channels is at least partially shaped by social factors
that determine the very formation of the relationship and its quality. In the next
section, the conceptual model is presented.

Social similarity and the nature of social ties

Studies on the formation, development, maintenance, and dissolution of close

social relationships have emphasized the importance of social similarity (Hartup
and Stevens, 1997; Maccoby, 1998; McPherson et al., 2002). This notion holds
that ‘contact and friendship formation between similar individuals occurs at a
higher rate than among dissimilar individuals’ (McPherson et al., 2002). Social
similarity is the result of opportunities for interaction emerging from the social
structuring of activities in society that expose individuals to each other.
Social similarity is an exogenous variable that reflects both opportunities for
mutual exposure and friendship selection, and as such, shapes the content and
the quality of the relationship being created. In that sense, social similarity among
friends is frequent because it provides important rewards. Similar individuals are
likely to participate in enjoyable joint activities with others who have similar
interests, hence to receive validation of their attitudes and beliefs. Participation in
the same activities increases the frequency and duration of social interaction.
Furthermore, similarity has been associated with stable and strong ties (Hallinan
and Kubitschek, 1988). When social dissimilarity exists at the beginning of
relationships, or a mismatch occurs in ascribed social statuses, relationships tend
to be unstable and are more likely to terminate as individuals move on to other
relationships in which there is greater similarity (Hallinan and Kubitscheck, 1988).

The nature of social interaction and strength of ties

Friendship is distinguished from other types of social relationships because

contact with friends is more intense. Intensity is usually a feature that describes
the history of the relationship and refers to its duration (Lee and Campbell, 1992).
A central characteristic of friendships is the development of a history of shared
experiences that define a feeling of belonging and shared identity. In addition, the

6
 Gustavo Mesch

development of central characteristics of friendship such as trust and reciprocity

are at least partially temporal processes. Trust develops through a process of
mutual disclosure of personal information, and this requires time.
Another important dimension is the content of a relationship. Different from formal
relationships, in which social interaction is partial and based on social status,
friendships are more holistic. A friend differs from a co-worker or a relative in that
friends are not restricted to a few topics of conversation or a few shared activities.
To be friends is to be together and to talk about anything. In that sense, an
important concept is multiplexity, a notion that describes the content of
relationships. Multiplexity suggests that a relationship is stronger when a tie
between two people encompasses multiple activities or topics of conversation
rather than a single activity or shared topic. Studies have shown that higher
multiplexity is reported among friends who report having a similar social
background such as age, gender, and ethnicity (Stoller et al., 2001). In other
words, background similarity or homophily increases the likelihood of multiplexity.
Individuals who share status characteristics are more likely to have a broad
spectrum of topics to talk about and activities to get involved in. While in some
studies multiplexity has been used as a proxy for tie strength (Stoller et al., 2001),
in the seminal work of Mardsen and Campbell (1984) it was not found to be a
central component of tie strength. These authors showed that emotional intensity,
indicated by measures of closeness and trust, is the best measure of the strength
of a tie.

The Israeli context

In Israel, Internet use is expanding rapidly. In 1998, only 11% of Israeli

households reported having access to the Internet; the figure had risen to 30% by
2002 (CBS, 2002). As elsewhere, in Israel there is a digital divide. Internet use is
proving higher among males than females, and socioeconomic differences are
reflected in Internet use. Most Internet users reported earning an average or
above-average income, and being of Western origin (CBS, 2002).
Adolescents’ use of the Internet has expanded even faster. While in 2001 only
35% of the adolescent population had access to the Internet, by 2004 65% had
access at home. As to purpose, the overwhelming majority of adolescent Internet
users reported that it was mainly for social purposes. Almost 74% of these
respondents said that they liked to meet new people through the net (Minerva
Center of Youth Studies, 2004).
In Israel, as elsewhere, youth represent a significant proportion of Internet users
and in this sense they call for special attention. Furthermore, most current
research focuses on English-speaking countries. Little is known about the
connection between Internet use and social relationships in non-English-speaking
countries.
Studies have started to examine the relationship of Internet use and social
involvement, and 14% of Israeli adolescents reported having friends whom they

7
A study of adolescents’ online and offline social relationships

met online (Mesch and Talmud, 2004). These adolescents were found to have a
more dispersed and heterogeneous network in terms of gender and age than
those who did not have online friends (Mesch and Talmud, 2004). The goal of this
paper is to investigate the differential quality of social relationships created online
and face-to-face among adolescents in Israel. In this, and in keeping with the
literature reviewed, we focus on the effect of social similarity, duration of the
relationship, and multiplexity on the strength of ties among adolescents in Israel.

Methods

This study was part of the annual national youth survey conducted by the Minerva
Center for Youth Studies at the University of Haifa. The data were collected
between June and October 2004. The annual survey covers a representative
sample of 1000 households in Israel. The sampling procedure begins with a
random sample of 60 localities with a population of 2000 or more. Then,
according to the size of the adolescent population in each settlement,
neighborhoods are selected randomly. The number of neighborhoods in each
settlement is determined by the juvenile population size (13-18 years old) in the
locality. At least one neighborhood is randomly selected in settlements with a low
proportion of adolescents, and more than one in the larger urban areas. In each
neighborhood, 15 households are randomly selected. The selected
neighborhoods represent all geographic areas of Israel, and also different sizes of
settlements, from big cities to small towns and villages. The survey includes items
on social and demographic characteristics of the youth, socio-demographic
characteristics of their closest friends, types of resources exchanged, and degree
of perceived closeness to each friend.
In the survey, each adolescent was asked for the names of six close friends. The
respondent provided information on each friend’s age, gender, and place of
residence; and whether he/she met him/her for the first time at school or through
extracurricular activities, in the neighborhood or online. The adolescent was also
asked to indicate the length of time that he/she had known him/her, and the
extent to which the respondent felt closeness and trust, and would ask for help
from each of the friends named.
The interviews were conducted face-to-face in the respondent’s house by trained
interviewers. Certain items on the questionnaire measured the socio-demographic
characteristics of the adolescent and of ego-networks (up to six friends). Here we
focus on the degree of similarity in age, gender, and place of residence between
the respondent and the first friend who was named.

8
 Gustavo Mesch

Measures

Dependent variables

Intensity of friendship was determined by means of a single measure. Following

the work of Lee and Campbell (1992) respondents were asked to state how long
they had known each friend.
To measure multiplexity, content and activity multiplexity were distinguished and
two different scales were built. Adolescents were presented with a list of nine
items and were asked to indicate for each one if it was a frequent topic of
conversation between them and their friends. The topics were school, parents,
family, friends, sports, personal problems, music and TV programs, romantic
relationships, and dress and fashion. The measure of content multiplexity was
built by summation of all the topics. The scale had an acceptable reliability of
α=0.674 (means=4.57 sd=2.11). The second measure, activity multiplexity, was
constructed of five items, these being activities; adolescents were asked to
indicate which activities they did with the first friend they named. Responses were
meeting at parties, meeting at homes, meeting at school, going out together, and
participating in the same extracurricular activities. The scale showed an
acceptable reliability of α=0.607. The final scale was built as a sum across all the
items.
Strength of ties was measured by a number of survey items. Referring to the first
friend named, respondents were asked to indicate how close they felt to him/her,
how important he/she was for them, how far they would ask him/her for help, and
how far they trusted him/her. Responses were given on a five-point Likert scale.
The items were subjected to a factor analysis using varimax rotation. One factor
was found and a scale was built with reliability α=0.811. Next the scale was built
by a simple summation of the responses over all the items.

Independent variables

A unique feature of the current survey is that it included a measure of the place
where the first friend was met for the first time. For each friend, respondents were
asked to indicate whether he/she was first met on the Internet, at school, in
extracurricular activities, or in the neighborhood. From this question we computed
a measure distinguishing the setting in which the first friend was met. A dummy
variable was created indicating the place in which the friend was met for the first
time; the relevant categories were face-to-face (neighborhood, at school, in
extracurricular activities) and online (through chat rooms, bulletin boards, or email
use).
A number of measures of Internet use were used. Adolescents were asked to
report the number of hours per day that they used the Internet. The variable was
introduced as a continuous measure. Secondly, adolescents were asked to
indicate for how long they had access to the Internet from home, and the variable
was introduced as a continuous measure.

9
A study of adolescents’ online and offline social relationships

To measure friends’ similarity, three measures were created. Adolescents were

asked for the place of residence of the first friend. Possible responses were: in
the same neighborhood, in the same city, in another city in Israel, in another
country. A conservative approach to the measurement of neighborhood was
taken and a dummy variable that was coded 1 when the first friend was reported
to live in the same neighborhood or the same city was created. When the friend
was reported as living in another city or another country the variable was coded 0.
This conservative approach was taken because in some central and northern
areas of the country the density of the population is such that having a friend in a
nearby neighborhood may mean having a friend in another city.
Adolescents were asked the age (in years) of the first friend that they named.
Similarity in age was measured by taking the age of alter and subtracting it from
the age of ego. Then a dummy variable was calculated, and was coded 1 when
the ego was the same age as, or one year younger or older than, the alter. In
other words, 1 indicated age similarity and 0 indicated age dissimilarity. The
definition of age similarity used in this study is consistent with previous studies
that defined same-age friendship when youngsters were within 12 months of each
other’s age (Hartup and Stevens, 1997).
Gender similarity was defined likewise. Adolescents were asked the gender of the
first friend they named. Then the gender of the ego and that of alter were
compared and a dummy variable measuring gender similarity was created. The
variable was coded 1 when the genders of ego and of alter were the same and 0
when they were not.
In addition, in the multivariate analysis, adolescent’s age, gender, number of
siblings, and nationality (1=Jew) and for mother’s education were introduced.

Sample description

Of the 1000 adolescents contacted, 987 agreed to participate in the study.

Respondents’ average age was 15.52 years (sd 1.66); girls and boys were almost
equally represented (52% were boys). In terms of religious denomination, 79%
were Israeli Jews. In socioeconomic status, average father’s education was 12.63
years (sd 3.50) and average mother’s education was 12.52 years (sd 3.37).
Regarding family status, 86.8% reported that their parents were married and
13.2% of parents were separated or divorced.
Access to the Internet was reported by 66.7% of the adolescents. Respondents
were asked where the first friend was met: 60% first met the friend at school, 28%
in the neighborhood, and 12% first met online. In our sample, the majority of the
adolescents met their closest friend at school, but a significant percentage (40%)
met their closest friend in other social settings such as the neighborhood and
online. The descriptive analysis showed that for the whole sample 53.4% of the
friends first named lived in the same neighborhood as the respondent, 85% were
of the same gender as the respondent, and 87% were the same age as the
respondent.

10
 Gustavo Mesch

Adolescents reporting having online friends did not differ in age from those
reporting not having online friends (15.51 years and 15.65 years; p=n.s.). Gender
differences existed as a higher percentage of boys than girls reported having
online friends. Of those reporting having a friend who was met online, two-thirds
were boys and only one-third were girls.
Regarding socio-demographic, adolescents who reported that their friend was
met at school or in the neighborhood showed on average a higher percentage of
age similarity. While 89% of the adolescents who did not have an online friend
reported that their friends were about their age, only 77% of the respondents who
had an online friend reported this. A similar situation emerged regarding gender
similarity. Of the adolescents without an online friend, 88% had friends of the
same gender; for the ones with an online friend the percentage was 69%. These
differences are important as they indicate greater dissimilarity in dyadic
characteristics; this should be controlled, as in previous studies social similarity
has proved a predictor of stable relationships and strong ties. Furthermore,
multivariate analysis that controls for age is needed as social similarity in age,
gender, and residence diminishes as adolescents grow older.
Differences in the mean duration, multiplexity and strength of the association with
the first friend named were found. When the adolescent reported that the friend
was met online, the average strength of the tie turned out to be lower (12.10;
sd=2.52) than when the friend was met face-to-face (13.92; sd=1.79). The
heterogeneity of tie strength was higher for online friends, as indicated by the
standard deviations. Duration of the friendship was also higher for face-to-face
friends; on average they reported a duration of 3.81 years (sd=0.55) while for
those reporting an online friend, duration was 3.07 years on average (sd=1.21).
As to multiplexity, statistically significant differences were found for respondents
who reported meeting an adolescent online and face-to-face. Adolescents whose
friend was met online reported fewer topics of conversation (mean=3.78, sd=2.36)
than adolescents who met their friend face-to-face (mean=4.57, sd=2.17), a
difference that was statistically significant (p<0.05); they also reported fewer
shared activities (for online friend, mean=2.77, sd=1.49); for face-to-face friend
mean=3.61, sd=1.77). Having established a significant difference in the number of
topics discussed and shared activities, it was interesting to know if the topics and
activities differed not only in number but also in type.
Adolescents were asked, after they had indicated whether the first friend they
named was met in a face-to-face setting or online, to state the activities they
engaged in with this friend. Table 1 presents these activities, as engaged in
proportionately by adolescents reporting meeting the first friend face-to-face and
online. Distinct differences are evident in activities undertaken with face-to-face
friends and with online friends. Certain activities were reported more by
adolescents with a face-to-face friend than by adolescents with an online friend.
Face-to-face relationships yielded a higher proportion of phone conversations,
meetings at school, meetings at friends’ houses and of hanging out together.
Regarding going to parties together, no differences were found, and as regards to
extracurricular activities in the evenings, online friends were more likely to
participate together. Overall, face-to-face friends apparently engaged in different
activities from online friends. Yet as seen from the table, online relationships were

11
A study of adolescents’ online and offline social relationships

not wholly virtual. Friends who met online did engage in face-to-face activities, but
it is important to keep in mind that they were just fewer, not non-existent.

Table 1. Proportion of adolescent and friend engaging in shared activities according to

origin of the relationship

Things we do together Friend was met face-to-face Friend was met online

Phone conversations 0.741 0.583**

Going to parties 0.364 0.305

Meeting at school 0.65 0.331*

Meeting at friends’ houses 0.684 0.194*

Hanging out 0.669 0.361*

Going to school extracurricular 0.09 0.11*

activities

**p<0.01, *p<0.05

Given that Internet friends met online met face-to-face less often, it was
reasonable to inquire into the nature of this relationship and to want to know the
resources they exchanged in order to be considered friends. To address this
question the proportion of adolescents with face-to-face and online friends
according to topics of discussion was compared.

Table 2. Proportion of adolescent and friend discussing diverse topics according to origin
of the relationship

Topics we discuss together Friend was met face-to-face Friend was met online

School 0.63 0.60

Parents 0.50 0.39

Friends 0.75 0.71

Hobbies 0.421 0.421

Personal problems 0.593 0.368*

TV shows and movies 0.618 0.526

Romantic relationships 0.499 0.342**

Fashion and dieting programs 0.546 0.421

**p<0.01, *p<0.05

For most of the non-personal topics (school, parents, friends, hobbies, TV shows,
movies and fashion) the difference in the proportion of face-to-face and online
friends who talked about them was not statistically significant. With several topics,
such as school and friends, the proportion was quite high (more than 60% of
adolescents with a face-to-face friend and adolescents with an online friend). Yet

12
 Gustavo Mesch

two topics did show a significant difference: personal problems and romantic
relationships. A higher proportion of adolescents who met their friend face-to-face
than of adolescents who met their friend online discussed intimate issues. Thus,
intimacy was higher among face-to-face friends than among online friends.
But how were these differences in certain activities and topics of conversation
associated with the quality of relationships? I start exploring this issue by
presenting the bivariate correlation matrix. Table 3 present the bivariate
correlations, means and standard deviations of the variables included in the
analysis.
The purpose of this exploratory analysis was to examine the pattern and size of
associations among its variables. The bivariate correlations between the
dependent variables were significant and of low magnitude, not threatening
multicollinearity. As can be expected, the strength of the relationship was
positively correlated with content multiplexity (r=0.316, p<0.01), activity
multiplexity (r=0.169, p<0.01), and duration of the relationship (r=0.175, p<0.01).
As required for testing a mediating effect, the strength of the relationship was
negatively related to the origin of the relationship (r=–0.123, p<0.01), indicating
that without controls, relationships originating in face-to-face settings, such as
school or neighborhood, were perceived as closer than relationships originating
online. The importance of social similarity can be appreciated as well. All the
measures of social similarity were negatively related to age, indicating that with
age, gender similarity and similarity in residence diminish. The direction of the
bivariate correlation was the same for residential similarity but was not statistically
significant. The correlation indicates, as in many past studies, that homophily
decreases with age. The measures of multiplexity and duration of the relationship
were also related to the quality of the relationship. Age similarity was positively
related to content and activity multiplexity, indicating that similar individuals tend
to conduct more diverse activities together and to talk about more topics. Gender
similarity was only related to activity multiplexity, indicating that individuals of the
same sex are likely to spend more time together in more diverse activities.
Interesting as they are, bivariate results are limited, as they do not control for
different variables. To conduct this test, a multivariate analysis was needed.

13
Table 3. Correlations and descriptive statistics for the sample

Tie strength 1.0

Content 0.322** 1.0

multiplexity

Activity 0.189** 0.380** 1.0

multiplexity

Duration 0.148** 0.093* 0.089* 1.0

Online friend – – – – 1.0

0.128** 0.082* 0.106** 0.292**

Age 0.063 0.067 –0.031 0.023 0.035 1.0

Gender – – –0.070 –0.011 0.061 0.032 1.0

0.118** 0.219**
(1=Male)

Nationality 0.051 0.039 0.092 0.039 –0.027 0.021 0.08 1.0

(1=Jew)

Parental –0.038 –0.052 0.045 0.008 –0.018 – 0.025 0.046 1.0

status 0.005

Parental –0.007 –0.003 0.038 0.064 –0.054 – 0.075 0.233** 0.047 1.0
education 0.036

Number of –0.020 –0.031 –0.073 –0.034 –0.004 0.027 0.027 – –0.012 – 1.0
siblings 0.298** 0.353**

Gender 0.042 0.030 0.063 0.203** – – 0.029 –0.062 0.011 0.014 0.030 1.0
similarity 0.137** 0.079*

Age 0.062 0.093* 0.106** 0.046 –0.050 – – 0.042 – 0.015 – 0.257** 1.0
similarity 0.047 0.014 0.120** 0.004
 Residential 0.177** 0.022 –0.011 0.127** – – 0.015 –0.033 0.052 –0.050 – 0.145** 0.059 1.0
similarity 0.161** 0.069 0.015

Duration of 0.107** 0.093* 0.135** 0.014 –0.012 0.037 0.008 0.294** 0.020 0.314** –/274 –0.003 0.070 –0.008 1.0
Internet use

Daily –0.010 – 0.062 –0.019 .049 0.021 0.055 0.058 0.022 –0.035 – –0.071 – –0.004 0.013 1.0
frequency of 0.085* 0.101 0.057
use

Means (sd) 13.9 4.49 3.55 3.76 0.12 15.53 0.51 0.80 0.78 13.219 2.60 0.85 0.87 0.53 3.4 3.9
(1.5)
(1.85) (2.21) (1.78) (0.64) (0.22) (1.67) (0.50) (0.39) (0.25) (3.28) (1.48) (0.34) (0.33) (0.49) (4.42)

*p<0.01, **p<0.05
A comparative study of youth online and offline social relationships

Table 4 presents the results of regressing socio-demographic variables,

propinquity, similarity variables, and the origin of the friendship on the length of
time friends had known each other. The results show that length of acquaintance
was related to measures of propinquity and similarity. As may be expected,
friends living in the same neighborhood were acquainted for longer, reflecting the
effect of propinquity and probably length of residence on length of time a friend
had been known. Furthermore, friends of the same sex reported a longer duration
of friendship.
The second model in Table 4 presents the results of regressing the same
variables on our measure of content multiplexity. In this table I find some different
effects; age and gender were related to multiplexity, probably reflecting
developmental processes. Older adolescents were more likely to report
discussing more topics and day-to-day issues with their friends. Previous
literature has shown that as adolescents become older they are more likely to
confide in their friends about their grievances. In addition, as adolescents grow
older the issues that generate mutual interest become more diversified and broad.
On the other hand, it is noticeable that girls reported more topics of conversation
than boys. The literature on gender differences in friendships reports that for girls,
friendship means talking and intimacy on different topics; for boys, friendship is
more doing things together. This may explain why boys’ interests are more
focused and narrow.
Propinquity and similarity were also associated with the diversity of topics that
adolescent friends discuss. Adolescents whose friend resided in the same
neighborhood reported a wider diversity of topics of conversation than
adolescents whose friends lived in another neighborhood or city. Proximity is
certainly an important component of opportunity, as easy and casual access to a
friend probably means more informal opportunities for conversation in which more
wide-ranging topics of conversation are likely to arise. Gender homogeneity is
important as well. Apparently cross-gender friendships are more restricted in their
topics of conversation.
The third model in Table 4 present the results regressing the independent
variables on activity multiplexity. Gender was again negatively related to activity
multiplexity, indicating that boys were less likely than girls to engage in a large
variety of activities with their friends. Nationality was also found to have a
significant effect: Israeli Jews were more likely to share more activities together.
In this analysis, similarity of gender, age and propinquity were found not to be
related to the diversity of activities that adolescents undertake together. The origin
of the friendship was notably related to the degree of activity multiplexity.
Adolescents who met their friends in face-to-face settings such as the school or
the neighborhood reported, as expected, a more diverse number of activities
together. Table 5 presents a three-stage OLS in which the independent variables
are regressed on the strength of ties.

16
 Gustavo Mesch

Table 4. OLS regression predicting duration of friendship, topic multiplexity, and activity
multiplexity

Duration of friendship Content multiplexity Activity multiplexity

Variable name Parameter Standard Parameter Standard Parameter Standard

estimate parameter estimate parameter estimate parameter
estimate estimate estimate
(se) (se) (se)

Age 0.031 0.081** 0.126 0.101* –0.009 –0.008

(0.015) (0.052) (0.045)

Gender 0.001 0.001 –0.815 –0.196* –0.128 –0.037

(1=male)
(0.051) (0.173) (0.149)

Nationality 0.084 0.054 –0.037 –0.007 0.144 0.033

(1=Jew)
(0.071) (0.242) (0.208)

Parents’ marital –0.004 –0.012 –0.060 –0.057 0.055 0.062

status
(0.013) (0.044) (0.038)
(1=married)

Mother’s 0.005 0.027 –0.054 –0.083 –0.017 –0.030

education
(0.009) (0.030) (0.026)

Number of 0.004 0.009 –0.146 –0.101* –0.049 –0.040

siblings
(0.020) (0.068) (0.059)

Gender 0.402 0.221* –0.138 –0.023 0.112 0.022

similarity
(0.077) (0.262) (0.224)

Age similarity –0.031 –0.016 0.377 0.059 0.574 0.106*

(0.082) (0.280) (0.242)

Propinquity 0.114 0.090** 0.013 0.003 –0.023 –0.007

(0.052) (0.178) (0.153)

Duration of use 0.003 0.008 0.209 0.154* 0.185 0.164*

(0.018) (0.061) (0.052)

Frequency of 0.002 0.011 –0.040 –0.081** 0.027 0.063

daily use
(0.006) (0.021) (0.018)

Online friend –0.806 –0.281* –0.707 –0.076* –0.879 –0.111*

(0.118) (0.307) (0.343)

Constant 2.794* 3.672* 2.743

(0.309) (1.049) (0.901)
2
Adjusted R 0.150 0.090 0.052

*p<0.01, **p<0.05

17
Table 5. OLS regression predicting strength of ties

Basic model Content multiplexity Activity multiplexity Friendship duration

Variable name Parameter Standard Parameter Standard Parameter Standard Parameter Standard
estimate (se) parameter estimate (se) parameter estimate (se) parameter estimate (se) parameter
estimate estimate estimate estimate

Age 0.095 0.089** 0.069 0.065 0.097 0.092** 0.079 0.074

(0.046) (0.045) (0.045) (0.045)

Gender –0.347 –0.097** –0.170 –0.048 –0.362 –0.103* –0.378 –0.108*

(1=male)
(0.152) (0.150) (0.149) (0.149)

Nationality 0.005 0.001 0.015 0.003 0.021 0.005 –0.014 –0.003

(1=Jew)
(0.213) (0.206) (0.209) (0.209)

Parents’ –0.051 –0.057 –0.036 –0.040 –0.063 –0.072 –0.053 –0.060

marital status
(0.039) (0.038) (0.038) (0.038)
(1=married)

Mother’s –0.028 –0.049 –0.015 –0.027 –0.023 –0.042 –0.029 –0.052

education
(0.027) (0.026) (0.026) (0.026)

Number of –0.013 –0.010 0.021 0.017 –0.004 –0.004 –0.023 –0.019

siblings
(0.059) (0.058) (0.059) (0.058)

Gender –0.294 –0.057 –0.268 –0.052 –0.343 –0.068 –0.503 –0.100**

similarity
(0.230) (0.222) (0.224) (0.231)

Age similarity 0.182 0.032 0.106 0.019 –0.012 –0.002 0.120 0.022
(0.250) (0.241) (0.247) (0.245)

Propinquity 0.579 0.162* 0.576 0.161* 0.574 0.163* 0.533 0.151*

(0.157) (0.151) (0.153) (0.154)
 Duration of 0.160 0.137* 0.111 0.096** 0.105 0.092** 0.137 0.120*
use
(0.054) (0.053) (0.053) (0.053)

Frequency of 0.006 0.014 0.015 0.035 0.016 0.038 0.020 0.046

daily use
(0.018) (0.018) (0.018) (0.018)

Online friend –0.959 –0.120* –0.792 –0.100 –0.584 –0.074 –0.378 –0.048
(0.346) (0.635) (0.344) (0.360)

Content 0.227 0.264*

multiplexity
(0.037)

Activity 0.164 0.162*

multiplexity
(0.044)

Duration 0.413 0.143*

(0.135)

Constant 12.431* 11.548* 12.095 11.468*

(0.933) (0.913) (0.924)* (0.996)
2
Adjusted R 0.062 0.124 0.132 0.072

*p<0.01, **p<0.05, +p<0.10

A comparative study of youth online and offline social relationships

In the first step, demographic variables and origin of friends were regressed on
closeness to friends. The model shows that propinquity was positively related to
the strength of the ties in all the analyses. The results indicate that individuals
who lived in the same neighborhood reported more closeness and trust in their
friends. The same result was found in all the models even when other relevant
variables were controlled. Face-to-face friends were more likely to be reported as
close friends. In the next step I incorporated the measures of content multiplexity.
The results show that this variable was a suppressor of the effect of friend’s
origin, as it became statistically non-significant. The next model in Table 5 shows
a similar result for activity multiplexity. Again, the introduction of this variable
washed out the effect of origin of the friendship. The third model introduced the
measure of duration of the relationship, and it washed out the previous
statistically significant effect of the origin of the friend.
The results indicate that online friends were perceived as less close both because
of inadequate duration of the friendship and insufficient multiplexity of
communication.

Discussion

The goal of the current study was to study the differential quality of online and
offline social relationships. In studying this topic a conceptual model that assumes
that relationships are socially structured, based on broad social processes of
sorting and selection of individuals according to their resources in different social
contexts that shape the likelihood of association, was presented and tested.
Adolescents sharing social statuses are more likely to associate because these
social statuses shape their concerns and interests. Thus social similarity, and not
communication channels as suggested by theories of computer-mediated
communication, are the exogenous factors that cause individuals to associate.
Once they have done so, the intensity, content and duration of the relationship is
shaped at least partially by their shared social status, and these in turn shape the
quality of the association. This conceptual model was empirically tested in a
representative sample of adolescents that had access to and use of the Internet.
This data set was particularly suited to test the hypothesis because it allowed a
distinction between adolescents who made friends online and those who did not.
The findings provide partial support for the association of social similarity and
various measures of intensity and content of the relationship. As suggested by the
conceptual model, gender similarity and propinquity were positively related to the
duration of friendship. Adolescents of the same sex reported knowing each other
longer. Individuals residing in the same location reported the same. Regarding the
number of topics discussed, measures of similarity were not found to be related;
regarding activity, only age similarity was related. However, in all the models
measuring intensity and content of the relationships the origin of the relationship
was found to be significant. Adolescents with an online friend reported that this
friend was known for a shorter time than face-to-face friends, they discussed
fewer topics, and they participated in fewer shared activities. The finding seems to

20
 Gustavo Mesch

indicate that online friends play a reduced and probably more specialized role in
the lives of adolescents than face-to-face friends at extracurricular activities and
parties. But they are met less at school, and respondents hang out with them
less. As to the content of the topics discussed, not only was there less discussion
of topics but the topics discussed tended not to be of a personal nature, such as
romantic relationships and personal problems.
The multivariate analysis revealed that without controlling for the intensity, content
and activities of the relationship, online friends tended to be perceived as less
close than face-to-face friends. The model that included measures of similarity
showed that even after controlling for similarity measures, in particular
propinquity, online ties were still weaker. This finding indicates that the reason
that online ties are perceived as distant is not their geographical distance. When
measures of the intensity, content and shared activities were introduced, the
effect of origin of the relationship washed out. This statistical result provides some
explanation of why relationships created online are perceived to be weak ties.
First, the time dimension in any association appears to be important, probably
because duration of the relationship is a proxy for shared events and
circumstances in which a history of the relationship is developed, and it is in the
context of these shared events that mutual trust and reciprocity develops.
Second, independently of the duration, the number of topics discussed and the
number of shared activities washed out the effect of friendship origin. This result
indicates that independently of time, close relationships tend to be holistic, not
restricted to particular activities and topics. Online relationships at this point
appear to be restricted to non-personal topics and not everyday activities, and in
that sense they are perceived as less integrated in the daily life of the individuals
and as more distant.

Conclusion

The findings of the study provide partial support for the expectation of
communication perspectives in that ties that originated online appear to be
weaker than the ones that originated offline. In that sense, the central expectation
of the social constructivists that the anonymity, isolation and lack of gating
features of computer-mediated communication make it easier to form strong ties
was rejected in this study. One plausible explanation for the discrepancy of our
results with those of some previous studies is the different nature of the research
design. Social constructivists relied on experimental designs, in which a small
sample of a highly self-selected population is used. Our study relies on a national
representative sample of adolescents, and sources of variation not accounted in
experimental designs are included in a larger study.
Although the limited clues of computer-mediated communication were not directly
tested in this study, the empirical findings do not support the assumption that the
channel of communication is responsible for the lack of intimacy with online ties.
Taken together, the results provide partial support for the hypothesis and expand
the hyper-personal model of online relationships. Walther (1996) has argued that

21
A comparative study of youth online and offline social relationships

although computer-mediated communication is a limited channel, time is critical

as it allows the development of shared clues and symbols of communication. The
effect of relationship duration found in this study supports this argument.
Relationship duration is critical as it facilitates not only the development of shared
language and understandings but also provides opportunities for interaction.
At the same time, our findings expand these arguments as it is shown that not
only duration explains the lack of intimacy in online social ties but multiplexity also
has a central role. Online social relationships then, appear to become personal
when intimacy in topics of conversation and shared activities develop. It is
through expansion of topics to conversation that relationship boundaries are
created and intimacy exercised. Furthermore, multiple shared activities provide
opportunities for the formation of shared memories conducive to a shared identity
of friends. In that sense, and in contradistinction to McKenna et al. (2002) who
attribute this intimacy to Internet anonymity, it was found that social similarity of
the partners is the force behind this intimacy.
The current study adds to the expanding literature on Internet communication and
relationship formation. The study highlights the centrality of social characteristics
of friends and their social interaction, factors that have not been emphasized in
previous studies that were based on communication theories only. The field will
benefit from studies that uncover the mechanism by which individuals share their
personal concerns and find space in the Internet and their daily life for shared
activities.

References

Beraman, P.S. and Moody, J. (2004) Adolescents’ Suicidability. American Journal

of Public Health 94:89–95.
Berndt, T., Miller, K. and Par, K. (1989) Adolescents’ perceptions of friends’ and
parents’ influence in other aspects of their school attachment. Journal
of Early Adolescence 9:419–435.
Central Bureau of Statistics (CBS) (2002) The Social Survey (Jerusalem: Israel).
Collins, A.W., Laursen, B.P. and Hartup, W.W. (1999) Relationships as
Developmental Contexts: The Minnesota Symposia on Child
Psychology, Vol. 30 (Erlbaum: Mahwah, NJ).
Crosnoe, R. (2000) Friendships in Childhood and Adolescence: The Life Course
and New Directions. Social Psychology Quarterly 4:377–391.
Cummings, J., Butler, B. and Kraut, R. (2002) The quality of online social
relationships. Communications of the ACM 45:103–108.
Cummings, J., Lee, J.B. and Kraut, R. (2004) Communication Technology and
Friendship during Transition from High School to College. In: R.Kraut,
M.Brynin and S.Kiesler (Eds) Domesticating information technology
(Oxford University Press: Oxford).

22
 Gustavo Mesch

Giordano, P. (2003) Relationships in Adolescence. Annual Review of Sociology

29:257–281.
Hallinan, M. and Kubitschek, W.N. (1988) The Effects of Individual and Structural
Characteristics on Intransitivity in Social Networks. Social Psychology
Quarterly 51:81–92.
Hampton, K. and Wellman, B. (2002) The not so global village of Netville. In:
B.Wellman and C.Haythornthwaite (Eds) The Internet in Everyday Life
(Blackwell: Oxford), pp. 345–372.
Hartup, W. and Stevens, N. (1997) Friendships and adaptation in the life course.
Psychological Bulletin 121:355–370.
Haythornthwaite, C. and Wellman, B. (2002) The Internet in Everyday Life.
(Blackwell: Oxford).
Joinson, A. (2001) Self disclosure in CMC: The role of self awareness and visual
anonymity. European Journal of Social Psychology 31:177–192.
Katz, J. and Rice, R. (2002) Social Consequences of Internet Use: Access,
Involvement and Interaction, chap. 10 (MIT Press: Boston, MA).
Kraut, R., Patterson, M., Lundmark, V., Kiesler, S., Mukiphadhyay, T. and
Scherlis, W. (1998) Internet paradox: a social technology that reduces
social involvement and psychological well-being? American
Psychologist 53:1011–1031.
Lee, B. and Campbell, K.E. (1992) Sources of Personal Neighbor Networks:
Social Integration, Need, or Time? Social Forces 70:1077–1100.
Maccoby, E. (1998) The Two Sexes: Growing up Apart, Coming Together
(Belknap/Harvard University Press: Cambridge, MA).
Marsden, P.V. and Campbell, K.E. (1984) Measuring tie strength. Social Forces
63:482–501.
McKenna, K.Y.A. and Bargh, J.A. (1998) Coming out in the age of the Internet.
Journal of Personality and Social Psychology 75:681–694.
McKenna, K., Green, A.S. and Gleason, M.E.G. (2002) Relationship formation on
the Internet: What is the big attraction. Journal of Social Issues 58:9–
31.
McPherson, M., Smith-Lovin, L. and Cook, J.M. (2002) Birds of a Feather:
Homophily in Social Networks. Annual Review of Sociology 27:415–
444.
Mesch, G. and Levanon, Y. (2003) Community networking and locally-based
social ties in two suburban localities. City and Community 2:335–351.
Mesch, G. and Talmud, I. (2004) Homophily and Quality of Online and Offline
Social Relationships among Adolescents (Minerva Center of Youth
Studies: Haifa) (unpublished manuscript).
Nie, N.H., Hillyguss, S.S. and Erbring, L. (2002) Internet user, interpersonal
relations and sociability: A time diary study. In: B.Wellman and

23
A comparative study of youth online and offline social relationships

C.Haythornthwaite (Eds) The Internet in Everyday Life (Blackwell:

Oxford), pp. 215–244.
Parks, M.R. and Floyd, R. (1996) Making friends in cyberspace. Journal of
Communication 46:80–97.
Rice, R.E. and Love, G. (1987) Electronic emotion: Socio-emotional content in a
computer mediated communication network. Communication Research
14:85–108.
Rubin, K.H., Bukowski, W. and Parker, J.G. (1998) Peer interactions,
relationships, and groups. In: W.Damon and N.Eisenberg (Eds)
Handbook of child psychology, Vol. 3 (5th ed.) (Wiley: New York), pp.
619–700.
Sproul, L. and Kiesler, S. (1986) Reducing Social Context Cues: Electronic Mail in
Organizational Communication. Management Science 32:1492–1512.
Stoller, E.P., Miller, B., and Guo, S. (2001) Shared Ethnicity and Relationship
Multiplexity within Informal Networks of Retired European American
Sunbelt Migrants. Research in Aging 23:304–325.
Walther, J.B. (1996) Computer-mediated communication: Impersonal,
Interpersonal and Hyperpersonal Interaction. Communication
Research 23:3–43.
Walther, J.B. and Boyd, S. (2002) Attraction to computer-mediated social support.
In: C.A.Lin and D.Atkin (Eds) Communication technology and society:
Audience adoption and uses (Hampton Press: Cresskill, NJ), pp. 153–
188.
Wellman, B. and Giulia, M. (1998) Net surfers don’t ride alone: Virtual
communities as communities. In: P.Kollok and M.Smith (Eds)
Communities in Cyberspace (Routledge: London) [Revised version in:
B.Wellman (Ed.) (1999) Networks in the Global Village (Westview
Press: Boulder, CO), pp. 331–367].
Wolak, J., Mitchell, K.J. and Finkelhor, D. (2003) Escaping or connecting?
Characteristics of youth who form close online relationships. Journal of
Adolescence 26:105–119.
Youniss, J. and Smollar, J. (1985) Adolescents’ Relations with Mothers, Fathers,
and Friends (Chicago University Press: Chicago).

24