Escaping or connecting ? Characteristics of youth who form close online relationships

Journal of
Adolescence
Journal of Adolescence 26 (2003) 105–119
www.elsevier.com/locate/jado
Escaping or connecting? Characteristics of youth who form
close online relationships
Janis Wolak*, KimberlyJ. Mitchell, David Finkelhor
Crimes against Children Research Center, University of New Hampshire, 126 Horton Hall, Durham, NH 03824, USA
Abstract
We used data from a US national sample of Internet users, ages 10–17 (N ¼ 1501), to explore the
characteristics of youth who had formed close relationships with people they met on the Internet (n ¼ 210).
Girls who had high levels of con?ict with parents or were highlytroubled were more likelythan other girls
to have close online relationships, as were boys who had low levels of communication with parents or were
highlytroubled, compared to other boys. Age, race and aspects of Internet use were also related. We know
little about the nature or qualityof the close online relationships, but youth with these sorts of problems
maybe more vulnerable to online exploitation and to other possible ill effects of online relationships. At the
same time, these relationships mayhave helpful aspects.
Published byElsevier Science Ltd. on behalf of The Association for Professionals in Services for
Adolescents.
Keywords: Adolescent friendships; Internet; Cyberspace; Online relationships
1. Introduction
Online relationships are a new phenomenon, but theyhave alreadybecome part of adolescent
culture. Psychologists have theorized about the meaning of online relationships during
adolescence (Turkle, 1995; Freeman-Longo, 2000), and law enforcement agents have warned
about the dangers of sexuallyexploitative online relationships (Armagh, Battaglia, & Lanning,
1999). Internet safetyinformation aimed at teens tacitlyacknowledges the extent of these
relationships byoffering safetyrules for conducting them and for attending face-to-face meetings
with online friends (Magid, 1998; Aftab, 2000). A national surveybythese authors con?rmed the
frequencyof online relationships, ?nding that 25% of Internet users ages 10–17 had formed casual
online friendships in the year before they were interviewed, and 14% had formed close online
*Corresponding author. Tel.: +1-603-862-4691; fax: +1-603-862-1122.
E-mail address: janis.wolak@unh.edu (J. Wolak).
0140-1971/02/$ 30.00 Published byElsevier Science Ltd. on behalf of The Association for Professionals in Services for
Adolescents.
PII: S 0 1 4 0 - 1 9 7 1 ( 0 2 ) 0 0 1 1 4 - 8

---

106
J. Wolak et al. / Journal of Adolescence 26 (2003) 105–119
friendships or online romances (Wolak, Mitchell, & Finkelhor, 2002). In fact, adolescents maybe
especiallydrawn to online relationships because of their intense interest in forming relationships,
and because the expansiveness of cyberspace frees them from some of the constraints of
adolescence bygiving them easyaccess to a world beyond that of their families, schools and
communities.
Much of the popular commentaryabout teens online suggests that the Internet is universally
interesting to youth (i.e. Tapscott, 1998). If this is so, it would seem that forming online
relationships might be one of the most generallyappealing aspects of Internet use among young
people, given that forming relationships is a developmental imperative of adolescence. We were
interested in whether online relationships such as close friendships and romances were spread
evenlythroughout the population of youth online or whether theywere more common among
some segments of youth Internet users. In fact, knowing if some youth are more likely than others
to be involved in online relationships could be seen as a ?rst step in learning about the impact
online relationships maybe having among adolescents. There has been little empirical data about
the characteristics of adolescents who form online relationships. We used data from a national
sample of youth Internet users to explore the associations between forming close online
relationships and a number of problems common among adolescents, including being highly
troubled, reporting high levels of parent–child con?ict, low levels of communication with parents,
and high levels of delinquency, along with demographic characteristics and aspects of Internet
use. We conducted separate analyses for girls and for boys because of theories that suggest the
qualities of close relationships differ between girls and boys and meet different social needs
(Gilligan, 1982; Buhrmester, 1996).
2. Method
The Youth Internet SafetySurveyused telephone interviews to gather information from a
national sample of 1501 Internet-using young people, ages 10–17. ‘‘Internet use’’ was
operationalized as going online at least once a month for the past 6 months on a computer at
home, a school, a library, someone else’s home, or some other place. Telephone numbers of
households with children in the target age group were identi?ed through another large national
surveywith which these researchers were involved. This was the Second National Incidence Study
of Missing, Abducted, Runawayand ThrownawayChildren, NISMART 2, a surveyof over
16,000 households with children, which was conducted between Februaryand December 1999.
Further information about the methodologies of both NISMART 2 and the Youth Internet
SafetySurveycan be found in a previous report (Finkelhor, Mitchell, & Wolak, 2000).
The interviews for the Youth Internet SafetySurveywere conducted between August 1999 and
February2000, byexperienced professional interviewers. Upon reaching a household, an
interviewer speaking with an adult screened for the requisite level of Internet use bya 10–17-year-
old youth in the household. When an eligible youth was identi?ed, the interviewer conducted a
short interview with the parent or caretaker who knew the most about the youth’s Internet use
and then asked for permission to speak with the youth. In families where more than one youth
had the requisite level of Internet use, the youth with the most use was chosen. In cases of equal
Internet use, the youth with the most recent birthday was chosen. When parental consent was

---

J. Wolak et al. / Journal of Adolescence 26 (2003) 105–119
107
given, the interviewer spoke with the youth, con?rmed his or her level of Internet use, described
the surveyand obtained the youth’s consent. Youth interviews lasted from about 15 to 30 minutes.
Theywere scheduled at the convenience of youth participants and arranged for times when they
could talk freelyand con?dentially. Youth respondents received brochures about Internet safety
and $10.
2.1. Participation rate
Seventy-?ve per cent of the households approached completed the screening necessary to
determine their eligibilityfor participation in the survey. The completion rate among households
with eligible respondents was 82%. Five per cent of parents in eligible households refused the
adult interview. Another 11% of parents completed the adult interview but refused permission for
their child to participate in the youth interview. In 2% of eligible households, parents consented to
the youth interview, but youth refused to participate.
2.2. Sample
The ?nal sample consisted of 1501 youth (boys=790, girls=708). The mean age was 14.14
years (s.d=1.96). Table 1 further describes the demographic characteristics of the sample.
2.3. Instrumentation
The data used in this paper come from the youth interview, except for the demographic data,
which come from the parent interview, with the exception of information about race. The purpose
of the Youth Internet SafetySurveywas to assess how often young people encounter unwanted
sexual solicitations, pornographyand harassment online. The youth interview included questions
about the existence of close online relationships because some adolescent Internet users have been
sexuallysolicited in the context of these relationships. The y
outh interview also covered
delinquent behaviour, drug and alcohol use, depression, parent–child relationships, conventional
victimization, and other similar questions. The questions referred to events that happened ‘‘in the
past year,’’ except for the questions about symptoms of depression, which referred to the past
month.
Dependent variable: Youth were asked two questions about relationships initiated online, that
were used to de?ne close online relationships. These questions were formulated after we held a
series of focus groups with youth Internet users and learned that youth distinguished between
casual online friendships that involved liking a person theyhad met online, and closer friendships
or romances which involved more intimate exchanges. The ?rst question asked, ‘‘Have you had a
close friendship with someone you met on the Internet who you didn’t know in person? I mean
someone you could talk online with about things that were real important to you?’’ The second
one asked, ‘‘Have you had a romantic online relationship with someone you met on the Internet? I
mean someone who felt like a boyfriend or girlfriend.’’ Youth who answered ‘‘yes’’ to either of
these questions were considered to have established close online relationships. (Detailed
descriptive information about the online relationships, including data about face-to-face meetings,
is reported in Wolak, et al., 2002.)

---

108
J. Wolak et al. / Journal of Adolescence 26 (2003) 105–119
Table 1
Youth and household characteristics
Characteristic
All youth
Youth with close online
(N ¼ 1501)
relationships (n ¼ 210)
Gender
Male
53%
46%
Female
47%
53%
Age
10
4%
2%
11
8%
3%
12
11%
5%
13
15%
12%
14
16%
16%
15
18%
25%
16
17%
19%
17
13%
18%
Mean age
14.14
14.78
Lives with both biological parents
64%
62%
Highest educational level of an adult in household
Not HS graduate
2%
2%
HS graduate
21%
22%
Some college
22%
27%
College graduate
31%
32%
Post-college degree
22%
16%
Household income
o$20,000
8%
9%
$20,000 to $50,000
38%
39%
>$50,000 to $75,000
23%
25%
>$75,000
23%
21%
Type of community
Large city14%
14%
Suburb of large city21%
17%
Large town
15%
16%
Small town
28%
35%
Rural area
20%
17%
Race
Non-hispanic white
73%
81%
African American
10%
7%
American Indian or Alaskan native
3%
1%
Asian
3%
1%
Hispanic white
2%
3%
Other
7%
6%
Do not know/refused
2%
1%
Note: Some categories do not add to 100% due to rounding and/or missing data.

---

J. Wolak et al. / Journal of Adolescence 26 (2003) 105–119
109
Independent variables: Aside from demographic characteristics, most of the independent variables
are composites derived from factor analysis loadings (see Table 2). These composites identifyyouth
who were highly troubled, reported high parent–child con?ict or low communication with parents, or
who engaged in a high degree of delinquent behaviour. Two variables described Internet use, having
home Internet access and a composite indicating high Internet use. We dichotomized the composite
variables, for two reasons. First, dichotomous variables create clear categories — troubled girls, for
example — as opposed to a scale with a range of degrees of a characteristic. Second, theyallow the
use of logistic regression odds ratios that can be discussed in terms of relative risk.
2.4. Statistical analysis
Bivariate analyses: A series of Pearson chi-square tests and relative risk estimates were used in
two series of comparisons. First, we compared the characteristics of youth Internet users who had
Table 2
Construction of composite independent variables
Composite variable
Factor loading
Eigenvalue
Variance
High parent–child con?icta
Parent:
Yells
0.81
Takes awayprivileges
0.62
Nags
0.75
1.6
0.53
Low communication with parentb
How often parent knows:
Where youth is
0.88
Who youth is with
0.88
1.5
0.77
High degree of delinquent behaviourc
Above average use of alcohol or drugs (4+times/yr)
0.80
At least one delinquent behaviord
0.80
1.3
0.64
Highly troubled c
High depression (5+symptoms)
0.69
Physical or sexual assault, past year
0.68
At least one negative life evente
0.55
1.2
0.41
High Internet usea
Above average or expert user
0.71
Internet veryor extremelyimportant
0.57
Online 4+days/week
0.68
Online 2+hours/day0.49
1.5
0.38
a Youth with scores more than 1 s.d. above the mean are coded as having this characteristic; others are coded as zero.
b Youth with scores more than 1 s.d. below the mean are coded as having this characteristic; others are coded as zero.
c Youth with scores more than 2 s.d. above the mean are coded as having this characteristic; others are coded as zero.
d Delinquent behaviours include being picked up bythe police, assault, vandalism and theft.
e Negative life events include death of a familymember, moving, divorce or separation, parent job loss.

---

110
J. Wolak et al. / Journal of Adolescence 26 (2003) 105–119
established close online relationships to those who had not. Then we compared the characteristics
of girls who formed close online relationships with girls who did not and made a similar
comparison for boys.
Multivariate analyses: Finally, we conducted two logistic regressions, one for girls and one for
boys, to further test the associations of adolescent problems, characteristics of Internet use, and
age and race to close online relationships among each gender. In both the bivariate analyses and
logistic regressions, odds ratios were adjusted to derive estimates of associations that closely
represent the relative risk (Zhang, 1998).
3. Results
Fourteen per cent of youth (n ¼ 205) reported close online friendships and 2% (n ¼ 30)
reported online romances. Overall, 14% of youth (n ¼ 210) reported close online relationships,
because some youth (n ¼ 25) reported both friendships and romances. Girls were slightlymore
likelythan boy
s to have close online relationships (16% of girls vs. 12% of boys, p ¼ 0:05;
O.R.=1.3, C.I=1.0–1.8).
3.1. Bivariate associations of characteristics with close online relationships
In bivariate analyses, a disproportionate number of adolescents with close online relationships
were highlytroubled, reported high amounts of con?ict with their parents, low communication
with parents and engaged in high levels of delinquency(see Table 3). Youth with these
relationships also were more likelyto be high school age (14–17), non-Hispanic white, report high
levels of Internet use and have home Internet access.
Because theories about adolescent relationships suggest girls and boys pursue close relation-
ships in different manners and for different reasons, we looked at the associations between the
characteristics described above and close online relationships separately, for both genders, and
found that variations emerged (see Table 4). Because of this, separate logistic regressions were
done for girls and boys. Table 5 shows the correlations among the variables used in the regression
equations for girls and for boys.
3.2. Logistic regression for girls
The initial regression model included all of the variables listed in Table 4. The ?nal model
indicates that ?ve variables were related to forming close online relationships among girls (see
Table 6). Age was the onlyrelated demographic characteristic. Girls who were aged 14–17 were
about twice as likelyas girls who were 10–13 to form these kinds of relationships. Because close
relationships are an important developmental aspect of older adolescence, the high school age
girls probablyhad more interest in close online relationships and were less supervised and more
independent, with more freedom to pursue their online interests.
Two problem characteristics were associated with close online relationships, high parent–child
con?ict and being highly troubled. The highlytroubled girls had levels of depression, victimization
(mostlyphysical assaults bypeers), and troubling life events at least two standard deviations

---

J. Wolak et al. / Journal of Adolescence 26 (2003) 105–119
111
Table 3
Characteristics of youth Internet users with close online relationships compared to those who do not have close online
relationships
Characteristic
Close online
Na
Adjusted odds
95% con?dence
relationships
ratio
interval
(N ¼ 210)
Demographic
Age
Ages 10–13
8%
558
Ages 14–17
17%
942
2.3nnn
1.6–3.2
Race
9%
373
Minority15%
1128
White
1.8nn
1.2–2.6
Gender
Female
16%
708
1.3n
1.0–1.8
Male
12%
790
Problems
High parent–child
con?ict
No
12%
1204
Yes
25%
276
2.5nnn
1.8–3.4
Low communication
with parent
No
13%
1333
Yes
26%
163
2.5nnn
1.7–3.7
High degree of
delinquency
No
13%
1418
Yes
28%
83
2.5nnn
1.5–4.2
Highlytroubled
No
12%
1270
Yes
23%
231
2.2nnn
1.5–3.1
Internet use
High level of Internet
use
No
9%
1091
Yes
29%
410
2.4nnn,b
2.1–2.8
Home Internet access
No
7%
392
Yes
16%
1109
1.2nnn,b
1.1–1.3
npp0.05, nnpp0.01, nnnpp0.001.
a N=1501, but data is missing for some variables.
b Adjusted to correct for over-estimation of risk, including C.I.

---

112
J. Wolak et al. / Journal of Adolescence 26 (2003) 105–119
other
.8
.7
.1
.8
.0
.6
.3
rval
n?dence
some
1.1–2
1.2–3
1.5–3
1.5–5
1.1–3
1.8–2
1.1–1
95%
co
inte
—
causes
data
ratio
,b
,b
,b
n
nn
nnn
nnn
n
nnn
nnn
Odds
1.7
2.1
n.s.
2.3
3.0
1.8
2.2
1.2
Missing
boys
cases.
other
three
a
to
51
N
276
513
206
584
624
149
674
113
739
664
126
559
231
199
591
in
se
missing
compared
)
is
clo
tionships
=97
boys
9%
7%
7%
5%
Boys
online
rela
(
N
14%
14%
11%
17%
10%
24%
11%
27%
11%
18%
24%
15%
gender
and
e
girls,
becaus
.5
.5
.3
.1
.3
.3
rval
—
—
other
n?dence
boys
95%
co
inte
1.2–1
2.0–3
1.5–5
1.6–3
2.3–3
1.1–1
to
for
C.I.
ratio
,a
,b
,b
,b
,b
=790
n
nnn
nnn
***
nnn
nnn
nnn
compared
Odds
1.4
n.s.
2.7
2.8
n.s.
2.3
2.8
1.2
and
including
girls
risk,
for
of
relationships
a
49
31
N
280
428
167
541
578
126
657
677
604
104
529
179
191
517
=708
n
online
but
0.001
close
se
p
ships
over-estimation
p
clo
cases.
with
tion
nnn
sample,
for
line
=112)
%
of
8
9%
8%
Girls
on
rela
(
N
21%
17%
12%
34%
15%
33%
15%
26%
13%
30%
35%
19%
girls
0.01,
entire
of
p
correct
p
f
d
numbers
to
nn
for
3
7
o
t–child
in
t
e
of
rnet
ristic
MS
4
10–1
14–1
use
nication
ency
use
0.05,
s
s
inte
paren
degre
level
p
=1501
inority12%
hite
paren
p
N
Adjusted
Age
Age
M
W
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
n
a
b
Table
Characteristics
Characte
Demographic
Age
Race
PROBLE
High
con?ict
Low
commu
with
High
delinqu
Highlytrouble
Internet
High
internet
Home
access
variations

---

J. Wolak et al. / Journal of Adolescence 26 (2003) 105–119
113
Table 5
Pearson correlations among variables associated with close online relationships for girls and for boys
1
2
3
4
5
6
7
8
Girls
1. Close online relationship
1.000
2. Older than 13
0.169nn
1.000
3. White
0.059
0.068
1.000
4. Con?ict with parents
0.233nn
0.102nn
À0.118nn
1.000
5. Highlytroubled
0.159nn
À0.007
À0.023
0.108nn
1.000
6. Low communication
0.125nn
0.095n
À0.006
0.222nn
0.108nn
1.000
7. Highlydelinquent
0.059
0.145**
À0.011
0.153nn
0.165nn
0.349nn
1.000
8. High Internet use
0.300nn
0.151nn
0.017
0.144nn
0.062
0.020
0.098nn
1.000
9. Home Internet access
0.124nn
0.075n
0.180nn
À0.011
À0.089n
À0.009
À0.025
0.185nn
Boys
1.
2.
3.
4.
5.
6.
7.
8.
1. Close online relationship
1.000
2. Older than 13
0.080n
1.000
3. White
0.090n
0.072n
1.000
4. Con?ict with parents
0.065
0.036
À0.096nn
1.000
5. Highlytroubled
0.079n
0.015
À0.056
0.117nn
1.000
6. Low communication
0.146nn
0.129nn
À0.047
0.141nn
0.010
1.000
7. Highlydelinquent
0.121nn
0.139nn
0.027
0.121nn
0.153nn
0.172nn
1.000
8. High Internet use
0.226nn
0.139nn
0.033
0.117nn
0.039
0.040
0.080n
1.000
9. Home Internet access
0.119nn
0.054
0.127nn
À0.026
À0.098nn
À0.020
0.058
0.213nn
npp0:05; nnpp0:01:
higher than the other girls in the sample. Girls with high levels of parent–child con?ict reported
yelling, nagging and taking away privileges by parents at a level at least one standard deviation
higher than other girls. Girls in either of these categories were more than twice as likelyas other
girls to have formed close online relationships.
Two variables, low communication with parents and being highly delinquent, were not
signi?cantlyassociated with close online relationships when other variables were controlled for.
The delinquencyvariable was not signi?cant at the bivariate level among girls. The variable
denoting low communication was signi?cant in bivariate analyses, but a relatively small number
of girls reported this characteristic, which was signi?cantlyintercorrelated with parent–child
con?ict and being highlytroubled. Shared variance with signi?cantlyrelated variables probably
accounts for this characteristic’s lack of signi?cance in the regression.
Finally, and not surprisingly, Internet use and access were strongly associated with close online
relationships. Girls who reported high levels of Internet use were more likelythan other girls to
report these relationships, as were girls with home Internet access, even controlling for high levels
of Internet use.
3.3. Logistic regression for boys
In the ?nal regression model for boys, ?ve variables were signi?cantly associated with close
online relationships, being non-Hispanic white, low communication with parents, being highly

---

114
J. Wolak et al. / Journal of Adolescence 26 (2003) 105–119
nnn
—
0.3–0.9
—
1.1–3.3
1.6–3.3
—
1.7–2.5
1.1–4.3
0.084
0.160
68.698
95%
con?dence
interval
514.768
ratio
=786)
(
n
a
a
Odds
—
0.5
—
1.9
2.3
—
2.1
2.1
boys
among
and
—
Sig.
0.304
0.017
0.025
0.000
0.000
0.030
=702)
(
n
girls
0.273
0.745
—
0.636
1.064
1.210
0.763
Boys
B
À
among
nnn
interval.
1.3–3.9
—
1.7–3.3
1.5–3.3
—
———
1.9–3.0
1.1–3.9
0.146
0.250
relationships
95%
con?dence
interval
506.069
110.859
con?dence
online
a
a
a
close
Odds
ratio
2.3
—
2.5
2.3
—
—
2.4
2.1
including
with
risk,
of
—
associated
Sig.
0.002
0.116
0.000
0.000
0.000
0.022
over-estimation
characteristics
for
0.830
0.470
1.179
1.044
—
——
1.325
0.732
of
Girls
B
À
correct
to
13
Snell)
regressions
use
&
0.001.
6
than
Internet
chi-
parent–
p
con?ict
parents
Internet
p
log
(Cox
(Nagelkerke)
Adjusted
2
2
2
nnn
a
Table
Logistic
Characteristic
Demographic
Older
Minority
Problems
High
child
Highlytroubled
Low
communication
with
Highly
delinquent
Internet
High
use
Home
access
À
likelihood
Model
square
R
R

---

Document Outline

• Escaping or connecting? Characteristics of youth who form close online relationships
  • Introduction
  • Method
    • Participation rate
    • Sample
    • Instrumentation
    • Statistical analysis
  • Results
    • Bivariate associations of characteristics with close online relationships
    • Logistic regression for girls
    • Logistic regression for boys
    • Differences in close online relationships between youth with problems and youth without
  • Discussion
    • Limitations of the survey
    • Recommendations
  • Acknowledgements
  • References

---