ee
bybdh
text2.pdf
809
Communication Technology and Friendship
during the Transition from High School to College
Jonathon N. Cummings
Sloan School of Management
Massachusetts’s Institute of Technology
John B. Lee
Department of Sociology
Columbia University
Robert Kraut
Human-Computer Interaction Institute
Carnegie Mellon University
This research was supported by the Digital Society and Technologies program of
the National Science Foundation (IIS-0208900).
810
Abstract
Within social networks, relationships are “enacted.” They grow or decline through
communication and the exchange of social resources. Although geographic
distance can inhibit enactment, telecommunication technologies (e.g., phone, e-
mail, IM) are increasingly being used to maintain relationships over long
distances. In this paper, we examine the role these technologies play in
maintaining friendships. Taking advantage of a natural situation in which
friendships are at risk of ending, we follow students in their transition from high
school to college. When students move away from home, they reduce both their
communication with their old high school friends and their sense of psychological
connection to them. Longitudinal analyses show that communication slows the
decline in psychological closeness, but psychological closeness does not slow the
decline in communication. E-mail and IM are telecommunication technologies
that are especially useful among these students for maintaining friendships. The
usefulness of these technologies may stem from arbitrary pricing decisions, which
allow students to use them frequently, rather than from their intrinsic features,
such as media richness. Unlike the phone, pricing of e-mail and IM does not
depend on either message length or the distance the message must travel.
811
Introduction
People maintain only a limited number of personal relationships. Researchers
estimate that people typically keep ten to twenty important relationships, out of
the approximately 1,000 individuals whom they interact with or can identify (e.g.,
Fisher, 1982; Wellman, 1992). Friendships, in contrast to family relationships, are
especially fragile, and require active maintenance or they die (Canary & Stafford,
1994). While family ties exist because of the accident of birth and are often
maintained through obligation, friendship and romantic relationships are
voluntary. They grow, decline, and end through concrete actions (Allan, 1979).
In this paper we examine how young adults maintain friendships when faced with
life events that threaten them, such as moving from high school to college. In
particular, we examine the role that phone and computer communications play in
maintaining these friendships as the parties move geographically apart.
In Duck’s analogy (1988), friendships need a regular investment of effort;
otherwise, normal centripetal forces cause the friendship to come apart. In this
view, people develop and maintain particular relationships by enacting them, i.e.,
by carrying them out through regular exchanges of communication or social
support (Duck, 1988). Initial factors that bring people together, such as common
interests, shared work goals, beauty, or charm, lose power with time (Berg &
Clark, 1986). These factors must be supplemented with behavioral exchanges that
812
affect whether the relationship will be valued and retained, or devalued and
dropped (Berg & Clark, 1986). Regular contact is at the heart of friendship (Allan,
1979). We typically grow to like others with whom we communicate and spend
time, and this liking drives further companionship (Newcomb, 1961). It is in this
sense that friendships are enacted. They are maintained through communication
and other behavioral exchanges.
Since at least the 1930s, we have known that physical proximity increases the
likelihood of friendships and romantic relationships (e.g., Bossard, 1932;
Festinger, Schachter, & Back, 1950). However, when people are separated by
geographic distance, it becomes more difficult to enact relationships through
communication and the exchange of social support. As a result, when people
change residences and move away, personal ties often fade and dissolve. Rose
(1984), for example, found geographic separation to be the factor most often
associated with the disintegration of friendships.
Physical proximity is conducive to the growth and maintenance of personal
relationships, whereas physical distance leads to their dissolution. In part, this is
because proximity decreases the behavioral costs of communication between
people and hence increases its frequency, while distance increases the costs and
decreases the frequency. Proximity not only increases the frequency of
communication, but also shifts the types of interactions between individuals. For
instance, even if distant friends communicate frequently by phone and e-mail, the
813
distance itself make it difficult for them to spend leisure time together, to share
common activities, to be physically intimate, or to exchange certain types of
social support.
When people move apart, two factors contribute to the decline of the original
relationships. First, it becomes difficult to communicate with the people in the
original location. Second, building relationships in the new location consumes
some of the time and attention the person would need to maintain the old
relationships. As a result, friendships change after a move: one may drop a friend
altogether or shift that person from the active list to the list of those whom one
sends holiday greetings.
Communication technology and social relationships
Telecommunications technologies—literally communications at a distance—can
change the amount and type of communication between people who are located
remotely from each other, and thus can allow them to maintain friendships at a
distance. If friendships are enacted and maintained through communication, as
Duck proposed, then friendships among those who live far from each other are
less likely to decline the more they communicate. This hypothesis—that
communication regardless of the modality leads to the maintenance of friendships
at a distance— is over simplistic, and needs elaboration. In the section below we
814
discuss the evidence for how telecommunications affects friendship and features
of the relationships and the technology that moderate the link between
communication and friendship.
Differences among communication modalities. Phone, e-mail, and IM are the
three main telecommunication technologies used by American consumers. In
1876 the phone was invented, and in 1915 the first U.S. transcontinental phone
call was placed. According to AT&T, in 1945, a 10 minute evening phone call
between New York and Los Angeles cost $56.80 in 1995 dollars. By 1995, the
cost of this same call dropped to $1.50. At such a price, pre-World War II
conversations required good reasons to call. With today’s pricing, however,
people call anytime, just to talk. This drop in costs influences the extent to which
friendships can be sustained at a distance (Fischer, 1992). The first e-mail
message was sent in 1971 (Leiner et al., 2002) and the first of the IM programs,
ICQ (I seek you), was released in 1996 by Mirabilis (History of IM).
Scholars have argued that these media are not equally useful for building and
maintaining social relationships. In this section, we briefly review the empirical
evidence that suggests that these technologies differ in their usefulness for starting
and maintaining social relationship. We also identify three reasons why they
differ: (1) intrinsic properties of the media that influence both social presence and
how much information is exchanged during a communication session, (2) cost
structures that change how frequently people communicate over them, and (3)
815
features of the technologies and communication genres associated with them that
change the content of the communication.
A growing empirical literature examines the hypothesis that communication
modalities differ in their ability to support social relationships. Most of this
research compares text-based, computer-mediated communication with face-to-
face communication (e.g., Walther, 1995), although sometimes comparisons are
made to the phone as well. Cross-sectional research by Parks and Roberts (1998)
and by Cummings, Butler and Kraut (2002) have suggested that ties created or
primarily maintained online are of lower quality than those sustained through
other means. For example, Parks and Roberts (1998) surveyed respondents in an
electronic group about their relationships with another member of the group, as
well as with a matched person in their social network outside of the electronic
group. The authors found that respondents spent less time, were less
interdependent with, and were less committed to their on-line partner compared to
the off-line one. Cummings, Butler, and Kraut (2002) reported similar results.
They asked respondents to indicate how close they felt toward two different
people outside of the household. The first was the individual with whom they
communicated most often by e-mail, and the second was the person with whom
they reported communicating most using other modalities, including personal
visits and the phone. Respondents reported feeling significantly less close to the
e-mail partner.
816
In contrast to these survey studies, experimental studies suggest that social
relationships through computer-mediated communication can be as strong as
those that develop face-to-face, if participants are allowed to communicate for
enough time (Walther, 2002). McKenna (this volume) reports research showing
that cross-sex undergraduate pairs who met each another for the first time in an
Internet chat room tended to like one another more compared to those who met
face-to-face.
As these conflicting results demonstrate, social relationships that are maintained
using different communication media do not necessarily grow at the same rate or
to the same depth. Research currently presents no clear consensus about whether
one communication modality is better than another for the maintenance of social
relationships. The reason for these conflicting results may be that the
communication media themselves may inconsistently influence fundamental
communication variables; these variables may then mediate their influence on the
social relationships. In particular, the communication media may influence both
the frequency of communication sessions and the quality of communication
during a communication session.
Cost structures and the frequency of communication. Different communication
modalities impose both financial and behavioral costs, which are likely to affect
how frequently people use one or the other modality to communicate. For
example, telecommunication providers typically charge for each long distance
817
phone call based on distance and duration, while Internet service providers’ rates
for e-mail or instant message sessions are independent of distance and duration.
Previous research has shown that communication volume is highly sensitive to
these costs, and that people communicate substantially more when these costs are
reduced (e.g., see Mayer, 1977, Figure 7). The medium’s design affects the
behavioral cost of communication as well; for example, phones require
communicators to be simultaneously available before they can converse. To
overcome these limitations, people play “phone tag,” use answering machines to
convert synchronous to asynchronous communication, or they may restrict their
calling to known times of availability (Lacohee & Anderson, 2001). In contrast, e-
mail is asynchronous and does not require simultaneous availability. Yet another
medium, IM, requires simultaneous availability. Many IM applications (IM)
provide awareness services, which inform users when a potential partner is online
and available. In contrast to phone calls, however, the awareness services in IM
software help to synchronize the simultaneous availability of partners, and are
therefore likely to increase the frequency of communication.
Quality of communication. Communication media differ on the amount of
information they transmit and their interactivity, among other features (Clark &
Brennan, 1991). These features have implications for relationships that are
maintained using these media (Sproull, 1991). Social presence theory suggests
that media differ on the social presence that they afford; for example, face-to-face
communication provides more social presence than the phone, which in turn
818
provides more than text-based communication. The thesis is that media with more
social presence should be better at supporting social relationships (Short, 1976).
Although many scholars contrast computer-mediated communication to face-to-
face communication without differentiating among the varieties of text-based
communication, the degree of interactivity is likely to be especially important
both for language understanding in general (Clark & Brennan, 1991) and for
companionship (Rafaeli & Sudweeks, 1997). In contrast to the relatively
contemplative style of composing and reading e-mail messages, the
conversational style of an IM session makes the event more engaging and
analogous to “being with” a communication partner; i.e., IM offers more social
presence than e-mail. This interactivity encourages users to tailor messages to
particular recipients (Kraut, Lewis, & Swezey, 1982) and to use an informal of
communication style, making it easier for communicators to quickly repair
mistakes.
Media richness theory suggests that media can be differentiated on the number of
social cues that they convey, and their level of interactivity (Daft, 1984; Dennis &
Kinney, 1998). IM is richer than e-mail because it is more interactive. Among the
interactive media, face-to-face communication is richer than the phone, which in
turn is richer than IM, because both face-to-face and phone communications offer
more affective and interpretive cues such as tone of voice. These richer media are
better at reducing ‘uncertainty’ and ‘equivocality’ than leaner media such as IM
and e-mail; hence richer media are better suited to maintain relationships. Walther
819
(Walther & Parks, 2002) argues that because writing is slower than talking, less
information per unit time is conveyed in computer-mediated communication than
in face-to-face and phone conversations. As a result, people need more time to
develop relationships conducted over the Internet, although he argues that given
enough time, people can develop and maintain strong social relationships online.
In contrast to research suggesting that more social cues, social presence, and
interactivity are better for supporting social relationships, other researchers have
identified conditions under which fewer social cues, reduced social presence, and
less interactivity may be better. For example, Postmes and his colleague (Postmes,
Spears, & Lea, 1998, 2000) propose that the individuating information available
in rich communication media interferes with identification with a group as a
source of social influence. McKenna and colleagues (this volume) argue that the
lack of social presence and superficial social cues that are available in text-based,
computer-mediated communication allows individuals a better opportunity to
display and learn about each others’ true selves, and may be especially useful for
introverts or socially awkward individuals to slowly form social relationships,
without the pressures of face-to-face meetings.
Method
Overview
To examine the role of communication technologies in sustaining friendships, we
tracked high school students as they moved to college, a situation that places
820
existing social relationships at risk. As they move from high school to college,
students go from a secure world populated with high school friends who often
attend the same school and lived in the same town, to a world where these
relationships disperse as both the student and friends relocate. Relationships that
were once supported by geographic proximity are threatened. On the other hand,
new relationships form at the beginning of college, which are supported by
physical proximity; these new relationships may squeeze out the old ones. By
following the course of old high school and new college friendships, we evaluate
how communication in general, and different media in particular, facilitate the
development and maintenance of relationships.
This study had two samples, representing two different high school graduation
years. In both samples, we asked respondents at the end of high school to identify
up to 20 friends and acquaintances from high school; at end of their first semester
in college, those same respondents again identified up to 20 friendships that they
formed in college. Out of these, in the first group, we sampled 4 high school
friends and 4 college friends in the first sample; in the second group we sampled 3
high school friends and 3 college friends. Respondents reported on the frequency
of their communication with each relationship partner in-person as well as by
phone, e-mail, and IM, and their psychological closeness to that partner. We
followed these relationships for up to three years. We use hierarchical linear
growth models to examine the influence of time and communication frequency on
changes in psychological closeness.
821
Sampling respondents
We collected data from two samples of high school students. Sample 1 included
500 high school students who were admitted to Carnegie Mellon University
(Pittsburgh, PA) in the Spring of 2000, stratified by distance from home: 100
were students randomly selected within 15 miles of Pittsburgh, 100 were foreign
residents, and 300 were students randomly selected by U.S. zip code. Sample 2
included 500 high school students who were admitted to Carnegie Mellon in the
Spring of 2001, also stratified by distance from home: 100 students randomly
selected within 15 miles of Pittsburgh, 100 students randomly selected between
100-200 miles from Pittsburgh, 100 students randomly selected between 400-800
miles from Pittsburgh, 100 students randomly selected between 1700-5000 miles
from Pittsburgh, and 100 randomly selected international students. Participants
were sent a $2 bill before each survey was administered, and were entered into a
lottery for prizes after they completed the survey.
In each sample, survey data were collected during the spring of the students’
senior year in high school (June), at the end of their first freshman semester in
college (December), at the end of their freshman year (May), at the end of their
sophomore year (May), and for sample 1, the end of their Junior year (May). Of
the 1000 students initially invited to participate, 62.9% completed the first survey,
48.2% completed the second survey, 39.6% completed the third survey, 31.3%
822
completed the fourth survey, and 22.8% completed the fifth survey (sample 1
only).
Sampling relationships
We used name generators to sample the respondents’ high school and college
relationships. The purpose of the name generators was to elicit a wide variety of
social ties, from which we randomly selected individual relationships to follow.
We used this procedure rather than allowing respondents to nominate
relationships on their own, because the self-nomination would have restricted
variance on the outcomes of interest (i.e., psychological closeness). People tend to
select individuals who are emotionally close and currently provide support in their
lives (Burt, 1986). This selection on the dependent variable would either have
lead to regression towards the mean or made differences in changes in closeness
difficult to observe.
The name generators were phrased, “Think about relationships with specific
people in your (high school | college) social circle…(a) who provide you with
practical assistance, (b) with whom you discuss hobbies, sports, movies, and other
spare-time activities, (c) with whom you socialize, (d) who give you advice about
important issues, and (e) who are in the same organizations as you.” For each of
the 5 types of relationships, respondents entered up to 4 names along with the
gender and age of the tie.
823
Respondents used a web-based survey to complete the name generators; the
software then randomly selected a subset of the relationships that students
identified, stratified by gender. In Sample 1 were 4 high school relationships (two
males and two females) and 4 college relationships (two males and two females).
Sample 2 included 3 high school relationships (at least one male and at least one
female) and 3 college relationships (at least one male and at least one female). We
wanted to balance the gender of partners in order to broaden the kinds of
relationships studied, and again, to create variance on the outcome measures since
people, in general, feel closer to women than to men (e.g., Duck, Rutt, Hurst, &
Strejc, 1991; Wheeler & Nezlek, 1977). For the high school and college
relationships, respondents were asked whether each tie was a relative, romantic
partner, acquaintance, friend, close friend, or other. In the analyses reported
below, only acquaintances, friends, and close friends are included as “friends” to
avoid idiosyncrasies associated with relatives and romantic partner.
Measures
Psychological closeness. Though both samples used the same questions, response
options in Sample 1 during the first three time periods are slightly different than
those used in Sample 2 and the final two time periods in Sample 1. Closeness was
measured on a 5-pt scale with the question “How close do you feel to…”. In the
first sample (during Spring 2000, Fall 2000, and Spring 2001), the response
options only included (1) not very and (5) very, while in the second sample (and
during Spring 2002 and Spring 2003 in the first sample) the response options
824
included (1) Not at all, (2) Not too much, (3) Neutral, (4) Somewhat, and (5)
Very.
Time. The purpose of this paper is to examine how the relationship between
respondents and their partners change over time. We code time in months as the
interval between questionnaires: six months between the first three questionnaires
and 12 months thereafter. Time is 0 when a partner first appears in the data set,
i.e., at the initial questionnaire for high school friends and at the second
questionnaire for college friends.
Communication frequency. For each time period, respondents reported the
frequency with which they communicated with each partner (a) in-person, (b) by
phone, (c) by e-mail, and (d) by IM. They answered on 7-point Likert scales,
ranging from never to multiple times per day. For ease of interpretation, we
transformed the Likert scales to days per months of communication1. Overall
communication was the sum of communication across the four modalities. Since
our goal is to predict changes in psychological closeness among respondents and
their partners based on communication frequency, we use lagged communication
frequency to reduce ambiguity in making causal claims. That is, in the analyses
below, communication in the preceding time period predicts communication at
the subsequent time period, and changes in psychological closeness between the
two periods.
825
Analysis
In order to examine how communication using different technologies influences
changes in social relationships, we used hierarchical linear growth modeling, also
known as multi-level modeling (Bryk & Raudenbush, 1992; Singer & Willett,
2003). Multi-level modeling takes into account the non-independence of the data,
with each respondent describing multiple partners during multiple questionnaire
administrations. For the data described above, a 3-level model is required. In this
model, level 1 represents the respondent (e.g., the respondent’s sex, age, and
race), level 2 represents the relationship (e.g., partner’s sex and age,
communication frequency, and closeness with that partner), and level 3 represents
time (e.g., repeated observations).
In these analyses, the coefficients for time-invariant level 1 variables (e.g.,
respondents’ gender, age at the first time period, and race) and time-invariant
level 2 variables (e.g., partner’s gender and age at the first time period) represent
cross-sectional associations. The coefficients indicate whether, for example,
women report closer relationships with their partners on average than men, or
whether respondents report closer relationships on average with women partners
than males. The coefficients for time-varying level 2 variables (e.g., prior
frequency of communication with a partner) test whether communication predicts
subsequent psychological closeness. We can do this because, as shown below,
both the communication variables and the closeness variable are moderately
stable over time. The association of communication frequency with subsequent
826
closeness primarily represents the cross-sectional association of communication
and closeness.
The main effects and interactions with time represent the variables that predict
change in psychological closeness. The coefficient for the main effect of time
indicates the degree to which a participant’s closeness with a particular partner
changes over twelve months. A negative coefficient indicates a decline in
psychological closeness. The time X communication interactions indicate how
frequency of communication moderates the effects of time on changes in the
relationship. A positive coefficient for the interaction of time with communication
indicates that partners who communicate more have a slower decline in closeness.
We test the effects of communication frequency both for overall communication
and for the four modes of communication.
Results
Preliminary statistics. Combined descriptive statistics for respondents from
Sample 1 and Sample 2 are shown in Table 1. Approximately 51% of the
respondents were males; 30% were Caucasian, 30% were Asian, and they were 18
years old at the time of the first questionnaire. Of the friends and acquaintances
they described, 65% were from the high school years and 35% were added during
the first semester in college. All matched the respondents in gender and age.
827
At the time of the initial questionnaire, when friends and acquaintances were
initially described, respondents were communicating with them approximately 60
times per month or roughly twice a day. Participants communicated most with
those they felt closest to. The contemporaneous Pearson correlation between
overall communication with a partner and closeness to that partner, measured on
the same questionnaire was .41, p < .001.
[Insert Table 1 about here]
About half of this communication was conducted in-person (28 times per month),
with phone (15 times per month) and IM (15 times per month) occurring more
frequently than e-mail communication (9 episodes per month). Surprisingly, face-
to-face communication predicted closeness less well than phone, e-mail, or IM
communication: the contemporaneous correlation between closeness and
communication frequency was .10 for in-person communication, .41 for phone
communication, .35 for electronic mail, and .39 for IM. Presumably, this is
because phone, e-mail, and IM communication is primarily volitional (i.e., at least
one party intended the communication to occur), while in-person communication
is to a degree involuntary. Whether they wanted to or not, participants talked to
each other when they were in the same place (possibly to avoid being perceived as
rude).
828
All communication dropped over time, with the largest declines for face-to-face
communication. This supports our hypothesis that geographic distance affects the
patterns of communication in relationships (see Figure 1).
[Insert Figure 1 about here]
Table 1 also shows the inter-temporal stability of the time-varying measures:
communication frequency and psychological closeness. These were computed by
taking the Pearson correlation between the same variable measured on adjacent
questionnaires. The inter-temporal stability is moderate. Correlations for the
communication measures range from .44 to .49. Since the response rate to the
questionnaires dropped over time, this stability measure is most heavily
influenced by questionnaires early in the study. Because one might expect most
change in communication and relationships to occur during this period, this was
also the period that we sampled at 6-month intervals rather than yearly intervals.
The closeness that respondents expressed towards their partners also declined
across time (see Figure 2). Table 2 shows moderate stability in the measure of
psychological closeness, with the test/retest correlations being .65.
[Insert Figure 2 and Table 2 about here]
829
Predicting psychological closeness. Table 2 describes hierarchical linear growth
models, predicting psychological closeness with a partner from respondent
characteristics, partner characteristics, time, and overall communication
frequency. In these models, we standardized the psychological closeness variable,
with a mean of zero and a standard deviation of one. Thus we interpret the
coefficients as the extent to which a unit increase in an independent variable is
associated with psychological closeness, as measured in standard deviation units.
Model 1 examines the associations of psychological closeness with stable
characteristics of respondents and partners, and time. Older respondents reported
feeling less close to their partners than younger respondents felt to their partners
(and older respondents felt less close to older partners). In addition, respondents
reported feeling moderately closer to their high school friends than the ones they
met in college (by .3 standard deviation units). The highly reliable negative
coefficient for time confirms the pattern seen in Figure 2; respondents’ closeness
to their partners declines with time. On average, closeness to these friends and
acquaintances declines a fifth of a standard deviation per year. The non-
significant coefficient for the Time X Partner that was added in college indicates
that closeness diminishes approximately equally with time both for high school
and college friends, even though participants felt moderately closer to their high
school than college friends.
830
Model 2 in Table 2 adds the overall frequency of communication to the model.
The coefficient for lagged communication indicates that respondents felt closer to
partners with whom they had more communication in the previous time period.
Because communication frequency is measured on a log scale, the coefficient
means that respondents report feeling approximately 7% of a standard deviation
unit closer to partners with whom they communicate twice as frequently. The
positive Time X communication frequency interaction indicates that the decline in
psychological closeness is less for partners with whom respondents communicate
with more. Doubling communication with a partner reduces the decline in
closeness to that person by about 13% (ratio of the coefficients .026/.206).
[Insert Table 3 about here]
To predict psychological closeness, Table 3 decomposes the overall frequency of
communication into the modalities through which it occurs. We use the same
analysis framework as in Table 2 (Model 2), except that we include the
communication frequency for the four communication modalities in place of
overall communication frequency. The coefficients for the main effects of the
communication are in Table 3, Model 3; all are positive and significantly greater
than zero. This pattern indicates that respondents feel closer to friends with whom
they communicate using each of the four modalities. Differences in the size of the
coefficients are instructive. The coefficient for phone communication is twice is
high as the coefficients for e-mail and IM, the two computer-mediated
831
communication modalities. These results suggest that respondents are more likely
to talk by phone to those with whom they feel closer, than to communicate to
those same partners with either e-mail or IM. The two computer-mediated
communication coefficients in turn are twice as high as that for in-person
communication, again reflecting the frequent non-volitional nature of in-person
communication.
Model 4 in Table 3 adds the Time X Communication interactions to examine the
association of communication with changes in closeness. The positive coefficients
for both e-mail and IM suggest that communication using these modalities
reduces the decline in closeness. Doubling e-mail communication with a partner is
associated with a 34% reduction in the decline in closeness with that partner
(.036/.104). Similarly, doubling IM communication with a partner is associated
with an 18% reduction in the decline in closeness with that partner (.020/.104).
Predicting communication. The longitudinal results in Tables 2 and 3 are
consistent with the claim that communication prevents declines in psychological
closeness. However, to examine the causal direction in more detail, we conducted
supplementary analyses predicting changes in communication from closeness.
Table 4 presents results from hierarchical linear growth models that predict
frequency of communication (in the log scale) based on time, psychological
closeness at the prior time period, and their interactions. Stable characteristics of
the respondents and their partners were included as control variables. Overall,
832
during their college years, respondents communicated 175% more with partners
whom they added during college than with their high school friends. This
difference depended on the media that respondents selected. The gap was largest
for in-person communication and phone communication than for e-mail and IM.
[Insert Table 4 about here]
Consistent with Figure 2, the analyses show that overall communication drops
with time by about 43%, and that this effect is larger for in-person than for other
types of communication. Consistent with the correlations between closeness and
communication frequency that were reported previously, the multivariate analyses
show that students communicate more across all modalities with partners with
whom they felt closer in the previous time period.
The interactions between time and psychological closeness are the most
interesting results to assess the direction of the causal link between
communication and closeness. The non-significant interaction between time and
psychological closeness for overall communication suggests that respondents’
preexisting closeness with a partner does not mitigate the drop in communication.
The time and psychological closeness interactions were non-significant for both e-
mail and IM, again suggesting that respondents’ preexisting closeness with a
partner does not mitigate the drop in computer-mediated communication. The
positive time and psychological closeness interaction for in-person
communication suggests that prior closeness mitigates the drop in face-to-face
833
communication. We speculate that this result occurs because students make an
effort to see only their closest friends when they return from school for breaks or
summer vacation. In contrast, the negative interaction for phone communication
suggests that the drop in calls is especially steep for partners with whom the
respondent had previously felt very close. We speculate that financial costs force
students to refrain from calling once-close friends; people whom they called
frequently when they lived in the same town and attended the same school.
Discussion
Our findings indicate that when students move from home to college, they reduce
both their communication with their high school friends and their closeness, that
sense of psychological connection. This same effect occurs with new friends that
students make during their first semester in college as they get further into their
college career. The purpose of this paper was to see how these factors were
causally related, and to assess whether modern telecommunication technologies
change the risk to relationships that distance and time can introduce.
To summarize the main results, longitudinal analyses show that although
psychological closeness to high school and college friends declines with time, this
decline is less steep among pairs who communicate more. This pattern is
consistent with Duck’s hypothesis (1998) that communication can mitigate the
centripetal forces causing social relationships to split apart. Surprisingly, when we
looked at the influence of different communication modalities, communication by
834
e-mail and IM seem to retard the drop in closeness, but communication in-person
and by phone do not. This pattern of interactions occurs even though phone
communication is the technology that best predicts psychological closeness at any
given time.
Why do computer-mediated communications seem to guard against the disruption
of social relationships more than in-person and phone communication for these
students? This finding is inconsistent with some of the literature reviewed in the
introduction which focused on the media’s respective intrinsic properties. Both its
media richness and social presence suggest that communicating by phone would
be most useful for guarding against threats to relationships.
Why does closeness drop least among partners who regularly use e-mail and IM?
One possibility is that frequency of communication for computer-mediated media
is less affected by distance than communication by phone. The technological
advantages of computer-mediated communications are irrelevant in mitigating
distance as a factor; rather, this advantage is simply the result of government
regulations and corporate marketing decisions. Unlike traditional long distance
phone providers, Internet service providers do not charge more to send data
packets across town or across the country. Moreover, pricing schemes used by
long distance carriers are not fixed. This may change, however, if Internet phone
technologies become more popular since the Internet is regulated less by the
835
Federal Communications Commission and state agencies than are conventional
phone services (E.g., Federal Communications Commission, 2003).
One strength of our research is the use of longitudinal data, which enabled us to
follow relationships over time. As a result, we were not constrained by the
ambiguities inherent in cross-sectional analyses of how strong ties were at a single
point in time. For example, the cross-sectional results suggest that communication
by phone is the strongest predictor of psychological closeness with a
communication partner. This result is consistent with predictions from media
richness and social presence theories; i.e., that richer and more interactive media
will better support social relationships. However, the cross-sectional results are
also consistent with an alternative explanation: students primarily call people with
whom they feel close, while they are more careless with their computer-mediated
communication. The longitudinal results tell a different story. They show that
communication over a computer is associated with less erosion in social
relationships, while communication by phone is not. Our conclusion from this
pattern of results is that communication frequency, not communication quality, is
the important element that sustains relationships; we also conclude that for
economic reasons, the frequency of communication by phone is especially
sensitive to distance, while communication by computer is not.
One weakness in this research is the lack of an identifiable mechanism for why
different communication technologies had varied effects on relationship
836
maintenance. It is likely that norms of communication technology use contribute
to how a change in that communication will influence the relationship. For
example, the norm among some high school students is to congregate on IM at
specified times in the evening to “hang out” with friends, whereas college
students tend to make themselves available whenever their computers are
connected to a network. These norms of availability can shape how just one
computer-mediated communication, IM, influences relationship maintenance.
Future research would benefit from measuring communication norms. Students
may believe that regular use of a particular media, regardless of content, can
sustain a relationship at the desired level of closeness and support.
Equally important, the communication content that is exchanged over different
media will undoubted influence how a communication change influences
relationships. As Lacohee and Anderson (2001) note, phone conversations are
often used to exchange social support, especially among women (see also Boneva
& Kraut, 2003). In contrast, as Boneva and colleagues (this volume) note, much
of IM communication is used for exchanging chitchat, and social support is rare.
An obvious issue raised by the discussion of maintaining personal relationships
during the transition from high school to college is whether or not there are
academic consequences. Are students more likely to stay in college when
relationships are maintained successfully? Do students perform better in their
courses? Is college more satisfying for them than high school? We do not have
837
answers to these questions, but would speculate that personal relationships
contribute to a positive quality of life for students in college.
Summary. For students who move off to college, the ties with their high school
friends and the friends that they make their first semester in college are fragile.
On average, these relationships decline with time. Consistent with an enactment
model of relationship, however, communicating with these friends prevents the
relationships from declining as swiftly as they otherwise would. Communication
seems to inject energy into a relationship and prevents it from going dormant. In
contrast, simply feeling close to these friends does not prevent the communication
from declining.
E-mail and IM seem to be the telecommunication technologies that are especially
useful for maintaining friendships among young adults. The utility of these
technologies may not stem from their intrinsic features, for example, their
interactivity, media richness, the effort needed to type messages, or their ability to
convey social presence. Rather, we suspect that arbitrary economic decisions may
be more important. Unlike the phone, which costs more when talking longer to
someone farther away, pricing for the computer-mediated communication
technologies does not depend upon either the length of a message or the distance
it must travel.
838
Chapter 18 Table 1
Variable Mean Std N Test-retest
reliability
(Pearson r)
Respondent gender1 ( % male) 51.1 49.5 585 N/A
Respondent age1 18.8 .79 585 N/A
Respondent race1 (% white) 53.8 50.0 585 N/A
Respondent race1 (% Asian) 31.4 45.3 585 N/A
Partner age2 19.2 14.5 2533 N/A
Partner gender2 (% male) 51.2 42.9 2533 N/A
Added friend in college2 (%) 35.2 47.9 2533
N/A
Frequency of communication overall2,3 60.7 47.3
2526 .49
Frequency of communication in-person2,3 28.1 21.6 2526 .47
Frequency of communication by phone2,3 15.2 19.1 2526 .46
Frequency of communication by e-mail2,3 9.1 14.8 2526 .46
Frequency of communication by IM 2,3 15.4 21.0
2526 .44
Psychological closeness2 3.99 1.00 2526 .65
839
1 Measured at the initial questionnaire
2 Measured when a partner first enters the dataset (questionnaire 1 for high school
partners and questionnaire 2 for college partners)
3 Communication episodes per month
4 For time-varying variables, communication and closeness
840
Chapter 18 Table 2
Effect Estimate Stderr DF Pr > |t| Estimate Stderr DF Pr > |t| Intercept -.058 .099 435 -.506 .120 435 ***
Respondent gender (0=female; 1=male) -.050 .061 435 -.061 .059 435
Respondent age -.070 .032 435 * -.072 .031 435 *
Respondent white .057 .091 435 .049 .088 435
Respondent Asian .129 .097 435 .101 .094 435 Partner added in college (0=no; 1=year) -.365 .067 2672 *** -.354 .065 2662 ***
Partner gender (0=female; 1=male) -.058 .042 2672 -.040 .040 2662
Partner & respondent are same gender .416 .084 2672 *** .424 .080 2662 ***
Partner age -.008 .004 2672 t -.004 .004 2662
Time (years since partner entered analysis) -.223 .025 2672 *** -.205 .056 2662 ***
Time X Partner added in college .056 .056 2672 .046 .056 2662
Overall communication frequency, previous period (log) .079 .015 2662 ***
Time X Communication frequency (lagged) .027 .013 2662 *
Model 1 Model 2
841
Chapter 18 Table 3
Effect Estimate Stderr DF Pr > |t| Estimate Stderr DF Pr > Intercept -.465 .106 420 *** -.448 .109 420 ***
Respondent gender ( % male) -.030 .063 420 -.029 .063 420
Respondent age -.080 .033 420 * -.080 .033 420 *
Respondent white .078 .094 420 .075 .094 420
Respondent Asian .104 .101 420 .105 .101 420 Partner added in college (0=no; 1=year) -.446 .081 2109 *** -.475 .085 2105 ***
Partner gender (0=female; 1=male) -.028 .041 2109 -.025 .041 2105
Partner & respondent are same gender .443 .083 2109 *** .426 .083 2105 ***
Partner age .000 .004 2109 .000 .004 2105
Time (years since partner entered analysis) -.079 .031 2109 * -.107 .036 2105 **
Time x Partner added in college .091 .061 2109 t .134 .067 2105 *
In-person communication frequency (log) .017 .010 2109 *** .035 .017 2105 *
Phone communication frequency (log) .074 .010 2109 *** .082 .017 2105 ***
E-mail communication frequency (log) .038 .009 2109 *** .009 .016 2105
Instant messaging communication frequency (log) .039 .008 2109 *** .022 .013 2105 t
Time x In-person communication frequency (log) -.027 .016 2105
Time x Phone communication frequency (log) -.010 .018 2105
Time x E-mail communication frequency (log) .037 .015 2105 *
Time x Instant messaging communication frequency (log) .020 .011 2105 t
Model 1 Model 2
842
Chapter 18 Table 4
Effect Estimate Stderr DF P Estimate Stderr DF P Estimate Stderr DF P Estimate Stderr DF P Estimate Stderr DF P Intercept 1.705 .246 436 *** -.349 .248 435 -2.008 .249 436 *** -.657 .272 436 * -1.165 .329 435 ***
Respondent gender ( % male) .091 .093 436 .039 .091 435 -.043 .097 436 -.144 .112 436 .291 .128 435 *
Respondent age .056 .049 436 -.005 .048 435 .092 .051 436 t .195 .058 436 *** -.038 .067 435
Respondent white .276 .140 436 * .280 .137 435 * .073 .146 436 .165 .168 436 .446 .192 435 *
Respondent Asian .034 .149 436 .129 .147 435 -.099 .155 436 -.091 .179 436 .461 .206 435 * Partner added in college (0=no; 1=year) 1.756 .113 2969 *** 4.361 .124 2714 *** 1.176 .114 2954 *** .404 .121 2961 *** .547 .153 2842 ***
Partner gender (0=female; 1=male) -.037 .064 2969 .070 .067 2714 -.074 .066 2954 -.254 .069 2961 *** .013 .090 2842
Partner & respondent are same gender -.109 .129 2969 -.129 .134 2714 -.045 .133 2954 .025 .139 2961 -.180 .181 2842
Partner age -.021 .007 2969 ** -.002 .007 2714 -.006 .007 2954 -.006 .007 2961 -.060 .009 2842 ***
Time (years since partner entered analysis) -.616 .127 2969 *** -.435 .126 2714 *** .375 .125 2954 ** -.231 .134 2961 t -.509 .162 2842 **
Time x Partner added in college -1.115 .090 2969 *** -2.215 .094 2714 *** -.778 .090 2954 *** -.413 .096 2961 *** -.524 .118 2842 ***
Psychological closeness (lagged) .466 .050 2969 *** .188 .051 2714 *** .666 .050 2954 *** .469 .053 2961 *** .574 .066 2842 ***
Time x Psychological closeness .035 .033 2969 .065 .033 2714 * -.133 .033 2954 *** -.044 .035 2961 .002 .043 2842
Instant Messaging Communicaiton (log)Overall Communication (log) In-person Communicaiton (log) Phone Communicaiton (log) Email Communicaiton (log)
843
Chapter 18 Figure 1
0
1
2
3
4
5
6
0.00 6.00 12.00 24.00 36.00
M onths since first questionnaire
C om
m un
ic at
io n
fr eq
ue nc
y In person Phone Email Instant Messaging
Chapter 18 Figure 2
844
C lo
se ne
ss
1
2
3
4
5
0 6 12 24 36
Months since first questionnaire
845
References
Ackerman, M. S. (1990). Answer Garden: A tool for growing organizational
memory, Proceeding of the ACM Conference on Office Information Systems
(pp. 31-39). NY: Association of Computing Machinery.
Allan, G. (1979). A sociology of friendship and kinship. London: George Allen &
Unwin.
Attewell, P. (1986). Imperialism within complex organizations. Sociological
Theory, 4, 115-125.
Berg, J., & Clark, M. (1986). Differences in social exchange between intimate and
other relationships: Gradually evolving or quickly apparent? In V. Derlega, &
Winstead, B. (Ed.), Friendship and social interaction. New York: Springer-
Verlag.
Boneva, B., & Kraut, R. (2003). Email, gender and personal relationships. In B.
Wellman & Haythornthwaite (Eds.), The Internet in Everyday Life (pp. 372-
403). Malden, MA: Blackwell Publishers.
Brannick, M., Price, C., & Salas, E. (Eds.). (1997). Team performance assessment
and measurement: Theory, methods, and applications. Mahwah, NJ:
Lawrence Erlbaum Associates.
Clark, H., & Brennan, S. (1991). Grounding in communication. In L. Resnick & J.
Levine & S. Teasley (Eds.), Perspectives on socially shared cognition (pp.
127-149). Washington, DC: American Psychological Association.
Cummings, J., Butler, B., & Kraut, R. (2002). The quality of online social
relationships. Communications of the ACM, 45(7), 103-108.
846
Daft, R., & Lengel, R. (1984). Information richness: A new approach to
managerial behavior and organizational design. Research in Organizational
Behavior, 6, 191-233.
Dennis, A. R., & Kinney, S. T. (1998). Testing media richness theory in the new
media: The effects of cues, feedback, and task equivocality. Information
Systems Research, 9(3), 256-274.
Duck, S. (1998). Human relationships (3rd ed.). Thousand Oaks, CA: Sage
Publications.
Duck, S., Rutt, D., Hurst, M., & Strejc, H. (1991). Some evident truths about
conversations in everyday relationships: All communications are not created
equal. Human Communication Research, 18(2l), 228-267.
Federal Communications Commission. (2003). FCC to Begin Internet Telephony
Proceedings (Vol. 2004, pp. Press release). Washington, DC.
Fischer, C. S. (1992). America Calling: A Social History of the Phone to 1940.
Berkeley: University of California Press.
Floyd, R. W. (1978). The paradigms of programming. Communications of the
ACM, 22(8), 355-460.
Fulk, J., Schmitz, J., & Steinfield, C. (1990). Organizations and communication
technology
A social influence model of technology use. In J. Fulk, & Steinfield, C. (Ed.) (pp.
117-140). Newbury Park, CA: Sage.
Giuliano, V. (1981). Teleworking: A prospectus. Telephony, 200(67-75).
847
History of IM. Retrieved Feb 22, 2004, from the World Wide Web:
http://www.almark.net/Internet/html/slide6.html
Jacobson, D. (1999). Impression formation in cyberspace: Online expectations
and offline experiences in text-based virtual communities. Journal of
Computer-Mediated Communication Special Issue, 5(1), NP.
Jarvenpaa, S., Knoll, K., & Leidner, D. (1998). Is anybody out there? Antecedents
of trust in global virtual teams. Journal of Management Information Systems,
14(4), 29-64.
Katz, D., & Kahn, R. (1978). The social psychology of organizations. New York:
Wiley.
Katz, E. L., P. (1955). Personal influence: The part played by people in the flow
of mass communication. New York: The Free Press.
Katz, J. E. (1988). Telecommunications privacy policy in the U.S.A.: Socio-
political responses to technological advances. Telecommunications Policy, 12,
353-368.
Katz, M. L., & Shapiro, C. (1994). Systems competition and network effects.
Journal of Economic Perspectives, 8(2), 93-115.
Katz, R., & Tushman, M. (1979). Communication patterns, project performance,
and task characteristics: An empirical evaluation and integration in an R&D
setting. Organizational Behavior and Human Decision Processes, 23, 139-
162.
848
Kraut, R. E., Lewis, S. H., & Swezey, L. W. (1982). Listener responsiveness and
the coordination of conversation. Journal of Personality & Social Psychology,
43(4), 718-731.
Lacohee, H., & Anderson, B. (2001). Interacting with the phone. International
Journal of Human-Computer Studies, 54(5), 665-699.
Leiner, B. M., Cerf, V. G., Clark, D. D., Kahn, R. E., Kleinrock, L., Lynch, D. C.,
Postel, J., Roberts, L. G., & Wolff, S. (2002, December 10, 2003). A Brief
History of the Internet (Version 3.32). The Internet Society. Retrieved Feb 10,
2004, from the World Wide Web:
http://www.isoc.org/internet/history/brief.shtm
Mayer, M. (1977). The phone and the uses of time. In I. d. S. Pool (Ed.), The
social impact of the phone (pp. 225-246). Cambridge, MA: MIT Press.
Parks, M., & Roberts, L. (1998). Making MOOsic: The development of personal
relationships on line and a comparison to their off-line counterparts, Journal
of Social and Personal Relationships (Vol. 15, pp. 517-537).
Postmes, T., Spears, R., & Lea, M. (1998). Breaching or building social
boundaries? SIDE-effects of computer-mediated communication.
Communication Research, 25(6), 689-715.
Postmes, T., Spears, R., & Lea, M. (2000). The formation of group norms in
computer-mediated communication. Human Communication Research, 26(3),
341-371.
Short, J., Williams, E., & Christie, B. (1976). The social psychology of
telecommunications. New York: Wiley.
849
Sproull, L., & Kiesler, S. (1991). Connections: New ways of working in the
networked organization: MIT Press.
Walther, J. B. (1995). Relational aspects of computer-mediated communication:
Experimental observations over time. Organization Science, 6(2), 186-203.
Walther, J. B. (2002). Time effects in computer-mediated groups: Past, present,
and future. In P. Hinds & S. Kiesler (Eds.), Distributed work (pp. 235-257).
Cambridge, MA: MIT Press.
Walther, J. B., & Parks, M. R. (2002). Cues filtered out, cues filtered in:
Computer-mediated communication and relationships. In I. M. L. Knapp & J.
A. Daly (Eds.), Handbook of interpersonal communication (3rd ed., pp. 529-
563). Thousand Oaks, CA: Sage Publications.
850
Endnotes
1 Response labels changed slightly across questionnaire administrations. In the
first two administrations of Sample 1, the responses options were (0) never, (1)
less often, (2) monthly, (3) bi-weekly, (4) weekly, (5) daily, (6) many times per
day, while in all other administrations they were (0) never, (1) every few months,
(2) every few weeks, (3) 1-2 days a week, (4) 3-5 days a week, (5) about once a
day, (6) several times a day.
851
Captions
Chapter 18 Table 1: Variable means at initial time period
Chapter 18 Table 2: Predicting psychological closeness from respondent and
partner characteristics, overall communication, and time
Chapter 18 Table 3: Predicting psychological closeness from respondent and
partner characteristics, communication over different communication modalities,
and time
Chapter 18 Table 4: Predicting communication frequency from respondent and
partner characteristics, psychological closeness, and time
Chapter 18 Figure 1: Decline in communication over time by medium
Chapter 18 Figure 2: Decline in psychological closeness over time
