School Connectedness and the Transition Into and Out of Health-Risk Behavior Among Adolescents
Supportive and caring relations within families promote
academic achievement and protect against involvement
in health-risk behaviors by adolescents.1-4 Similarly,
supportive and caring relationships within schools (henceforth,
school connectedness) promote academic motivation
among adolescents.5-11 Much less is known, however, about
the influence of school connectedness on adolescent healthrisk
behaviors. Previous research generally suffers from
two limitations. First, most research is cross-sectional.12-15
The longitudinal research that does exist does not distinguish
between initiation or escalation or reduction of
health-risk behaviors.16,17 Second, school connectedness has
generally been treated as a broad construct that combines
students? perceptions of safety, support, belonging and
engagement.12,13,16,17 Such a broad conceptualization does not
provide clear guidance to policy makers and practitioners
on how to increase school connectedness. This paper
addresses these limitations by exploring the association
between two dimensions of school connectedness ?
perceived teacher support and social belonging ? and the
initiation, escalation and reduction of participation in six
adolescent health-risk behaviors.
Cross-sectional studies show that school connectedness
is associated with mental health and lower rates of involvement
in multiple health-risk behaviors, including substance
use, sexual intercourse, violence, delinquency, and suicidality.
12-15 One quasi-experimental study, the Seattle Social
Development Study, evaluated the effects of increasing the
school social bond among elementary school students. The
intervention group had significantly higher levels of school
connectedness than the control group at ages 13 and 18,
and was less likely to engage in violence or substance
Three dimensions of school connectedness are emphasized
in educational research: social support, belonging and
engagement.11,18-22 When young people receive empathy,
praise, and attention in a clear and consistent fashion, they
experience social support. The experience of social support
generates a sense of belonging which, in turn, leads to
increased engagement and academic motivation. Although
this theoretical model, originally laid out by Connell and
Wellborn,11 has been empirically supported for academic
outcomes, it has not been tested for health outcomes. Most
previous studies linking school connectedness to health-risk
behaviors combine the different dimensions of school
connectedness into a single measure or explore the effect of
a single dimension. Drawing on the theoretical framework
of Connell and Wellborn, we hypothesize that teacher
support will lead to delayed initiation of health-risk behaviors,
less escalation of involvement once the behavior is
initiated, and increased cessation of health-risk behaviors,
and that the effect of teacher support will be mediated by
Data were drawn from the National Longitudinal Study
of Adolescent Health (Add Health), a nationally representative
sample of American adolescents in grades 7-12 in
1995. The primary sampling frame for Add Health was US
high schools. A stratified sample of 80 high schools was
selected with probability proportional to the school?s
enrollment. A single feeder school was selected for each
high school with probability of selection proportional to the
percentage of the high school?s entering class that came
from the feeder school. Add Health includes private, religious,
and public schools from communities located in
urban, suburban, and rural areas of the country.23
All students in the eligible grade range at the participating
schools were asked to complete in-school questionnaires
during the 1994-1995 academic year. Based on
rosters of students from each school and the in-school questionnaires,
a representative sample of students was selected
for wave 1 in-home data collection. The response rate was
78.9%, yielding a sample of 20,745 students completing an
in-home questionnaire. Of these, 1,821 cases were not
assigned sampling weights. A second interview was
conducted during the following academic year for all
students except the 12th graders and a few select subsamples.
The wave 2 response rate was 88.2% (n = 14,738).
The present analysis restricts the sample to those students
who responded to both wave 1 and wave 2 surveys and who
were assigned survey weights at wave 2 (n = 13,570).
Measures of School Connectedness
Add Health contains six questions that tap aspects of
connection to school. Three of the questions were developed
by Bollen and Hoyle to measure social belonging.19
Students were asked how much they agreed or disagreed
with the following statements: ?You feel close to people at
your school,? ?You feel like you are part of your school,?
and ?You are happy to be at your school.? If the survey was
administered during the summer, the questions were asked
in the past tense, for example, ?Last year, you felt part of
your school.? Responses were recorded on a five-item
Likert-type scale ranging from ?strongly agree? to
Another three items asked the adolescent about his or
her perceptions of their teachers. The first question asked
students to report how much they agreed or disagreed with
the statement, ?The teachers at your school treat students
284| Journal of School Health| September 2004, Vol. 74, No. 7
School Connectedness and the Transition Into and Out
of Health-Risk Behavior Among Adolescents:
A Comparison of Social Belonging and Teacher Support
Clea McNeely, Christina Falci
Clea McNeely, Dept. of Population and Family Health Sciences, Johns
Hopkins Bloomberg School of Public Health, 615 North Wolfe St.,
Baltimore, MD 21205-2179; (email@example.com); and Christina Falci,
Dept. of Sociology, University of Minnesota, 909 Social Sciences Bldg.,
267 19th Ave., South, Minneapolis, MN 55455. This research was
supported by a W.T. Grant Faculty Scholars Award.
fairly.? Response categories ranged from ?strongly agree?
to ?strongly disagree.? A second question asked, ?Since
school started this year, how often have you had trouble
getting along with your teachers?? The five response categories
were ?never,? ?just a few times,? ?about once a
week,? ?almost every day,? and ?every day.? Responses to
this question were reverse-coded. The third question about
teachers appeared in a different section of the survey that
asked about how much different people in the young
person?s life care about him or her. The question was, ?How
much do you feel that your teachers care about you?? The
five response categories were ?not at all,? ?very little,?
?somewhat,? ?quite a bit,? and ?very much.?
Principal components and confirmatory factor analysis
were conducted to determine whether the social belonging
items and the items regarding teacher support comprised
two separate factors or a single construct of school connectedness.
The three social belonging measures loaded on one
principal component, whereas the three teacher support
items loaded on a second factor.19,24 This two-factor model
was tested using confirmatory factor analysis, and found to
have good model fit.24 The social belonging measure had
excellent reliability for a three-item scale (a = .78). The
teacher support scale had modest reliability (a = .63), probably
because two scale items addressed students? individual
relationship with their teachers whereas the third item
asked how teachers treat all students in the school. The
correlation between the two measures of school connectedness
was moderate (r = 0.43).
Measures of Health-Related Outcomes
The six health-related outcomes comprising a broad
array of adolescent health behaviors were measured at both
wave 1 and wave 2 to model the initiation, escalation and
cessation of behaviors.
Cigarette smoking was defined as a three-category variable
based on the number of days students reported smoking
cigarettes during the previous 30 days. No cigarette use
was defined as not having smoked in the past 30 days.
Occasional smoking was defined as having smoked on 1-19
days, and regular smoking was defined as having smoked
on 20-30 days of the previous 30 days.
Alcohol use was also defined as a nominal variable with
three categories indicating the frequency with which the
student reported getting ?drunk or very, very high on alcohol?
during the previous 12 months. No alcohol use was
defined as never having gotten drunk, occasional use was
defined as having gotten drunk up to once a month or less,
and regular alcohol use was defined as having gotten drunk
2-3 days a month or more.
Marijuana use was defined as a nominal variable with
three categories ? no use, occasional use and regular use ?
based on the number of times a student reported smoking
marijuana during the last 30 days. No marijuana use was
defined as not having smoked marijuana, occasional use
was defined as having smoked marijuana four or less times,
and regular use was defined as having smoked marijuana
more than four times in the previous 30 days.
Suicidality is a three-category variable indicating
whether or not a student had seriously considered suicide in
the past year and, if so, whether or not they had attempted
suicide. The three categories are: no suicidal thoughts,
suicidal thoughts in the past year, and suicide attempt in the
Transition to first sexual intercourse was defined as a
three-category variable indicating whether adolescents who
never had sex at wave 1 had sex by the wave 2 interview,
and, among those who had sexual intercourse by wave 2,
whether a condom was used the first time the adolescent
had sex. The three categories were labeled never had sex,
first sex with condom and first sex without condom.
Initiation of sexual intercourse was determined by the question,
?Have you ever had sexual intercourse? When we say
sexual intercourse, we mean when a male inserts his penis
into a female?s vagina.? Condom use was measured with a
series of questions asking respondents what form of birth
control they used, if any, when they had sex the first time.
Respondents could report up to three different methods of
Weapon-related violence was defined as a dichotomous
variable indicating whether the adolescent committed at
least one of the following acts in the previous year: threatened
to use a weapon to get something from someone,
pulled a knife or gun on someone, shot or stabbed someone,
used a weapon in a fight, or hurt someone badly enough to
need bandages or medical care.
Measures of Background Characteristics
The models included five potential confounding
sociodemographic characteristics: race/ethnicity, age,
gender, family structure, and household income. They also
included depressed mood, parental attachment, and grade
point average, three factors known to predict both school
connectedness and the health-risk behaviors. Depressed
mood was measured by an index comprised of the following
four items: felt depressed, felt lonely, felt sad, and
could not shake the blues. The index ranged between 1 and
13, and indicated good reliability (a =.83). Parental attachment
was measured by six items assessing attachment to
both mothers and fathers. Adolescents answered a set of
three questions separately for each parent. One item asked,
?How close do you feel to your mother/father?? with five
response choices ranging from not at all to very much. The
other two items were also statements with five response
choices, from strongly agree to strongly disagree: ?Most of
the time, your mom/dad is warm and loving to you,? and
?Overall, you are satisfied with your relationship with your
mother/father.? If respondents only reported on one
parental-figure, then the parental attachment measure is
based on one parent. The index ranged between 1 and 18,
and had good alpha reliability of .80 (.76 for mothers only
and .79 for fathers only). All regression models in the study
controlled for these background characteristics; however,
for simplicity and clarity we did not report the coefficients
for the control variables. In the suicidality models, history
of family suicide was also included as a control variable.
We used conditional multinomial logistic and conditional
logistic regression to model the probability of transitions
both into and out of six health-risk behaviors. For
every behavioral status at wave 1, we modeled the transition
to every other possible behavioral status at wave 2.
Specifically, for adolescents who have no or occasional
involvement in health-risk behaviors at wave 1, we modeled
the probability of increased involvement by wave 2. For
Journal of School Health| September 2004, Vol. 74, No. 7| 285
adolescents who had occasional or regular involvement in
health-risk behaviors at wave 1, we modeled the probability
of decreased involvement over time. All analyses were done
in Stata 8.0 using sampling weights and adjusting for the
complex sampling design.25
The description of the sample is presented in Table 1.
On average, most students feel a sense of belonging and
that their teachers respect and care about them. The value
of both scales is just over nine out of a possible eleven
points. Nonetheless, there is good variability in the
measures, and the responses span the full possible range of
one to eleven. Taken together, the reported prevalence of
behaviors reveals that most American middle and high
school students do not engage in health-risk behaviors. Just
over a quarter (27%) of students reported using cigarettes in
the previous month. Of those who did report smoking, onehalf
were experimental smokers (smoked less than 20 days
in the previous month) and one-half were regular smokers.
Seventeen percent of students said they occasionally had
gotten ?drunk or very, very high? on alcohol (up to once a
286| Journal of School Health| September 2004, Vol. 74, No. 7
month or less) and 9% reporting getting drunk regularly,
defined as two or three times a month or more. Thirteen
percent used marijuana in the previous 30 days, and about a
quarter (23%) had engaged in weapon-related violence at
least once in the previous year. A third of the sample had
first sexual intercourse prior to the wave 1 interview. Sadly,
9% of the respondents seriously considered suicide in the
previous year and 4% attempted suicide.
Table 2 shows the patterns of transition into and out of
levels of involvement in health-risk behaviors between
waves 1 and 2. The cells report the proportion of adolescents
in a given behavioral category at wave 1 that are in
each behavior category at wave 2. For example, the first
row in the table shows that 81% of the respondents who
had never smoked at wave 1 were also nonsmokers at wave
2. In addition, 15% of nonsmokers at wave 1 transitioned
into smoking occasionally, and 5% transitioned into smoking
regularly at the wave 2 interview.
Overall, Table 2 shows great stability over time among
those adolescents who reported no involvement in a risk
behavior. Of those who did not engage in a risk behavior at
wave 1, over 80% reported not engaging in that behavior
one year later. The most change over time is seen for
adolescents who reported occasional or experimental
participation in a behavior at wave 1. Among occasional
substance users in wave 1, just 30% still smoked occasionally,
40% still got drunk occasionally, and 25% still used
marijuana occasionally at wave 2. A third of adolescents
who occasionally used alcohol and cigarettes at wave 1
reported no use at wave 2. Fully half of adolescents who
occasionally used marijuana in the 30 days prior to the
wave 1 interview reported no use at wave 2. The cessation
of occasional use reflects the experimental nature of healthrisk
behaviors in adolescence. Regular substance use is
more stable over time. Nearly 80% of adolescents who
smoked regularly at wave 1 also smoked regularly at wave
2, and half of the adolescents who regularly got drunk or
used marijuana also did so a year later.
The respondents, as a group, reported less violence at
wave 2 than at wave 1 (13% compared to 23%, respectively).
Just 6% of adolescents who did not report violence
within the past year at wave 1 reported a violent incident at
wave 2. In contrast, 65% of adolescents who reported
violence at wave 1 reported no violence in the past year at
Eighty-six percent of the adolescents who did not have
sexual intercourse prior to wave 1 did not initiate sexual
intercourse between wave 1 and wave 2. Among adolescents
who did initiate sexual intercourse between wave 1
and 2 (not shown in table), 67% used a condom. Finally, a
Journal of School Health| September 2004, Vol. 74, No. 7| 287
startling 30% of students who attempted suicide in the year
prior to wave 1 made at least one more attempt by wave 2.
The results from the multivariate models examining the
relationship between school connectedness and the transi-
288| Journal of School Health| September 2004, Vol. 74, No. 7
tion into and out of the six health-risk behaviors are shown
in Tables 3 and 4. Three models are presented for each risk
behavior. Model 1 contains the teacher support measure and
the control variables. Model 2 contains the social belonging
measure and the control variables. Model 3 contains both
school connectedness variables along with the control variables.
Because results from Model 3 are presented in two
separate columns, the columns are labeled Model 3a and
Each panel of Table 3 contains the results of three separate
multinomial logit regressions that model the transition
from a given status at wave 1 (eg, nonsmoking) into two
alternative statuses at wave 2 (eg, occasional smoking or
regular smoking). The table reports relative risk ratios,
which represent the risk of transitioning to an alternative
status relative to not transitioning for each one-unit change
in the school connectedness measure, holding all other variables
constant. For example, the first relative risk ratio in
Model 1 of Table 3 is the risk of transitioning from no cigarette
use at wave 1 to occasional use at wave 2 for each
one-unit change in teacher support, relative to remaining a
nonsmoker. These relative risk ratios are the association
between the school connectedness variables above and
beyond the independent effect of all the background characteristics,
including parent attachment and academic
Since multinomial logistic regression models simultaneously
estimate two coefficients for each independent variable
within the model, we report the statistical significance
of both the individual relative risk ratios and of the joint
hypothesis that both relative risk ratios are equal to one
(Wald F-test), as recommended by Hosmer and
Lemeshow.26 All Wald tests were adjusted for the complex
We hypothesized that both teacher support and social
belonging would be associated with a decreased probability
of initiating health-risk behaviors and an increased probability
of reduction or cessation, and that social belonging
mediates the effect of teacher support on health-risk behavior.
Comparing the relative risk ratios in model 1 to those in
model 3a shows how the inclusion of social belonging in
the model changes the effect of teacher support on adolescent
health-risk behaviors. Similarly, comparing model 2 to
model 3b shows how the inclusion of teacher support in the
model changes the effect of social belonging on adolescent
health-risk behaviors. If social belonging mediates teacher
support, we would expect to see the association between
teacher support and the health-risk behaviors to diminish
once social belonging is included in the model. We would
also expect the association of social belonging and the
outcomes to remain unchanged once teacher support is
added to the model. However, as shown in Model 2, social
belonging has no effect on initiation or cessation of the
health-risk behaviors, with the exception of marijuana use.
In some cases, social belonging is a significant risk factor
for initiation of health-risk behavior after teacher support is
also included in the model.
Cigarette, Alcohol, and Marijuana Use. In both Model
1 and 3a of Table 3, teacher support is a protective factor
for the initiation of cigarette smoking. Teacher support is
also protective against the escalation from occasional to
regular smoking. This suggests that not only might teacher
support protect against experimentation with cigarettes, but
it also might protect against an addictive habit among those
who have experimented with cigarettes. In model 2, social
belonging is not associated with the transition from being a
nonsmoker to smoking occasionally, and we find no
support in models 3a and 3b that social belonging mediates
the protective effect of teacher support. However, in model
3b, which includes both school connectedness variables,
social belonging is a risk factor for the transition from
nonsmoking to occasional smoking. Students who feel that
they are part of school, who feel close to people at school,
and who like going to school are more likely to start smoking
occasionally, once support from teachers is held
The pattern is the same for alcohol use. Teacher support
at wave 1 is associated with a lower probability of transitioning
from never getting drunk to both occasional and
regular episodes of getting drunk by wave 2. When both
connectedness measures are included in the model (Models
3a and 3b), social belonging becomes a risk factor for the
initiation of occasional and regular smoking among
nonsmokers. Neither school connectedness measure
predicts the transition from occasional use to regular use.
Likewise, they do not predict a reduction in alcohol use,
whether the transition be a decrease in use or quitting altogether.
For marijuana use, teacher support and social bonding
are protective against transitioning into either
occasional or regular use from no marijuana use. However,
social belonging is not related to initiating marijuana use
once teacher support is included in the model. Teacher
support is also protective of transitioning into regular use
from no marijuana use.
Suicidality, Sexual Intercourse, and Violence. Table 4
presents the results of models predicting the effect of
school connectedness on suicide, first sexual intercourse,
and violence. Teacher support protects against suicidal
attempts for those students who do not report experiencing
suicidal thoughts at wave 1. Teacher support is also protective
against the transition to first sexual intercourse,
whether protected by condom use or not.
Panel C in Table 4 shows the results of a logistic regression
analysis and therefore presents the risk ratios (rather
than relative risk ratios) for transitioning into and out of
violence between wave 1 and wave 2. Violence is the only
outcome for which teacher support is not only protective
against initiation of a health-risk behavior but is also associated
with cessation of the behavior. When students feel
supported by their teachers, they are less likely to engage in
weapon-related violence and are also more likely to desist
if they have been violent in the past.
Predicted Probabilities. Because relative risk ratios can
be difficult to interpret, we follow the recommendation of
Hosmer and Lemeshow26 and calculate predicted probabilities
from the models containing both school connectedness
variables and the full set of control variables. Table 5
presents the predicted percent of the sample that would
transition from no participation in the health-risk behavior
to various levels of participation for three values of teacher
support: the mean, one standard deviation below the mean,
and one standard deviation above the mean.
The effect of teacher support on adolescent health-risk
behavior is quite large. For example, if all respondents in
the sample had a teacher support score that was one standard
deviation above the mean, the percent of students who
Journal of School Health| September 2004, Vol. 74, No. 7| 289
transitioned from not smoking to occasional smoking
would decrease by 15% ? from 14% of the sample to 12%
? compared to if all students had the mean score of teacher
support. For cigarette and alcohol use, it appears that
increasing teacher support is more protective against the
initiation of regular use than occasional use.
The association between teacher support and the
outcomes may be somewhat overstated by these predicted
probabilities because there may be unmeasured factors that
cause both teacher support and the health-risk behavior
outcomes. Moreover, it is unlikely that teacher support
could be changed a full standard deviation without the
adolescents? scores on other protective factors being
changed as well. Nonetheless, even if these predicted probabilities
are an upper bound estimate of the association
between teacher support and the initiation of health-risk
behaviors, they suggest a substantial protective effect.
The results of this study show that different dimensions
of school connectedness have different effects on the initiation
of six health-risk behaviors: cigarette smoking, drinking
to the point of getting drunk, marijuana use, suicidal
ideation or attempt, first sexual intercourse, and weaponrelated
violence. Adolescents who perceive that their teachers
are fair and care about them ? referred to as teacher
support ? are less likely to initiate any of these six healthrisk
behaviors. This finding is consistent with previous
research showing that when students think their teachers
care about them personally and care about their learning,
they are more likely to be engaged in school, to do better
academically, and to participate in fewer health-risk behaviors.
12-18 However, our study shows that adolescents who
feel part of school and enjoy going to school ? referred to
as social belonging ? are not protected from initiation of
any of these health-risk behaviors. Rather, controlling for
teacher support reveals a suppressed risk effect of social
belonging on two health-risk behaviors: the initiation of
occasional smoking and drinking to the point of getting
drunk. These findings are not consistent with our hypothesis
that teacher support generates a sense of belonging which,
in turn, reduces involvement in health-risk behaviors.
290| Journal of School Health| September 2004, Vol. 74, No. 7
The degree to which school connectedness protects
against or promotes the initiation of health-risk behavior
might depend on the type of connection adolescents have to
school. Adolescents can develop conventional or unconventional
connection to school.3 Conventional connectedness
involves connections to individuals who engage in prosocial
behaviors and who regulate prosocial behavior in
others. Unconventional connectedness, in contrast, involves
connection to individuals who engage in behaviors that do
not conform to prosocial norms. Thus, an adolescent?s
school connectedness will be conventional or unconventional
depending on to whom an adolescent develops a
connection. The type of connectedness will determine the
direction of influence of school connectedness on healthrisk
behaviors.3 Conventional connectedness protects
against the initiation of health-risk behaviors whereas
unconventional connectedness is likely to promote the initiation
of health-risk behaviors.22
Adolescents develop connections at school to both peers
and adults, such as teachers. Connectedness to teachers is
presumed to be conventional because teachers reinforce
participation in behaviors that are sanctioned by the school.
Connection to peers, on the other hand, can be conventional
or unconventional depending on the norms within the peer
group. Unconventional connectedness to peers is likely to
develop when ?youths themselves dictate the norms, activities,
and structure that govern what youths do.?3 (p 3)
Although our measure of social belonging does not specifically
refer to peers, we believe that once the shared variance
with teacher support is removed, social belonging is
tapping primarily unconventional connectedness to peers.
Peer norms differ the most from norms adults hold for
adolescents for two behaviors: smoking and alcohol use.
For example, whereas most adults prefer that adolescents
abstain from substance use, by 12th grade, 62% of students
report having gotten drunk and 57% report having tried
cigarettes.27 If our measure of social belonging is measuring
unconventional connectedness to peers after controlling for
teacher support, then social belonging would become a risk
factor for the initiation of cigarette and alcohol use. It is
notable that connection to peers is not a risk factor for the
more serious health-risk behaviors such as regular smoking,
marijuana use, violence, and suicidality.
Teacher support is protective against the initiation of
health-risk behaviors, but has little effect on the reduction
or cessation of health-risk behaviors once initiated, with the
exception of violence. Since the violence measure reflects
participation in a single violent incident at any point in the
past year, it is possible that the respondents with higher
teacher support ceased violence long before the wave 1
measurement of teacher support. Teacher support might
have less influence on the reduction or cessation of healthrisk
behaviors than on their initiation because a student?s
involvement in risk behaviors reflects their willingness to
invest in unconventional norms, even if they continue to
feel supported by teachers and staff.28 Engagement, the
third dimension of connectedness, is the reciprocation by
the students of teacher support. It is the extent to which
students are invested in and committed to their relationships
with teachers. Engagement might be the component of
connectedness most important to the reduction of risk
behaviors. Stanton-Salazar28 describes how students who
are committed to a personal relationship with their teachers
are more likely to both seek out and respond to support
from teachers. Had we been able to measure engagement ?
a third dimension of school connectedness ? we might have
found it to be associated with the reduction or cessation of
Journal of School Health| September 2004, Vol. 74, No. 7| 291
Three limitations of this study should be noted. First, the
connectedness measures are limited in their ability to
measure additional dimensions of school connectedness,
such as engagement. A second measurement limitation
concerns our ability to accurately measure initiation of
health-risk behaviors. The transition from no participation
in these health-risk behaviors to involvement may not actually
represent initiation. Third, the analysis, although longitudinal,
is by no means causal. We could be observing a
selection effect rather than a protective effect for teacher
support. Despite these limitations, this paper has several
strengths. Distinguishing between two dimensions of
school connectedness contributes to the conceptual and
operational refinement of school connectedness. Moreover,
distinguishing between the initiation and reduction of
health-risk behaviors, which is not typically done in
research on adolescent health-risk behavior, reveals important
information about the mechanisms through which
school connectedness promotes health.
Separating school connectedness into two separate
dimensions also contributes to recommendations regarding
the translation of research into social policy. Our findings
suggest that conventional connectedness to teachers can
counterbalance negative influences of bonding to peers
with non-conventional behavioral norms. Through caring
about their students, treating them fairly and actively
engaging them in learning, teachers can delay the initiation
of health-risk behaviors. Our findings also suggest that
these same actions may not promote cessation of healthrisk
behaviors once they have been initiated. This suggests
that middle schools are a particularly important target for
promoting supportive teacher relationships, because most
middle schools students have not yet experimented with
health-risk behaviors. The transition from elementary
school to middle school has been documented as a time in
which students perceive less caring relationships with
A challenge for future research and intervention work is
to better understand the aspects of the student-teacher relationship
that promote reduction of health-risk behaviors.
Since our research suggests each dimension of school
connectedness might have a different influence on adolescent
outcomes, future research should distinguish between
dimensions of school connectedness as they relate to teachers,
peers, and learning. Additionally, an important unanswered
question is whether teacher support is equally
protective for all students as this main effects model
assumes. Another important question is whether support
from teachers can compensate for the lack of a close parentchild
relationship or whether a connection with parents is a
prerequisite to fostering connection with teachers.
1. Barber BK, Olsen JA. Socialization in context: connection, regulation,
and autonomy in the family, school and neighborhood, and with
peers. J Adolesc Res. 1997;12(2):287-315.
2. Miller B. Families Matter. Washington, DC: National Campaign to
Prevent Teen Pregnancy; 1998.
3. Karcher MJ. Connectedness and school violence: a framework for
developmental interventions. In: Gerler E, ed. Handbook of School
Violence. Haworth Press; 2003.
4. Crosnoe R. Social capital and the interplay of families and schools.
J Marriage Fam. 2004;66:267-280.
5. National Research Council. Engaging Schools: Fostering High
School Students? Motivation to Learn. Washington, DC: National
Academies of Science; 2003.
6. Rutter MB, Maughan P, Mortimore J, Ouston, Smith A. Fifteen
Thousand Hours: Secondary Schools and Their Effects on Children.
Somerset: Open Books; 1979.
7. Eccles JS, Early D, Frasier K, Belansky E, McCarthy K. The relation
of connection, regulation, and support for autonomy to adolescents?
functioning. J Adolesc Res. 1997;12:263-286.
8. Goodenow C, Grady K. The relationship of school belonging and
friends? values to academic motivation among urban adolescent students. J
Exp Educ. 1993;62:60-71.
9. Battistich VD, Solomon D, Watson KM, Schnaps E. Schools as
communities, poverty levels of student populations, and students? attitudes,
motive and performance: a multilevel analysis. Am Educ Res J.
10. Battin-Pearson S, Newcomb MD, Abbot RD, Hill KG, Catalano
RF, Hawkins JD. Predictors of early high school dropout: a test of five
theories. J Educ Psychol. 2000;92(3):568-582.
11. Connell J, Wellborn J. Competence, autonomy, and relatedness: a
motivational analysis of self-system processes. In: Gunnar M, Sroufe L,
eds. Self Processes in Development: Minnesota Symposium on Child
Development. Hillsdale, NJ: Earlbaum; 1991;23:43-77.
12. Battistich V, Hom A. The relationship between students? sense of
their school as a community and their involvement in problem behaviors.
Am J Public Health. 1997;87:1997-2001.
13. Resnick MD, Bearman PS, Blum RW, et al. Protecting adolescents
from harm: findings from the National Longitudinal Study of Adolescent
Health. JAMA. 1997;278:823-832.
14. Jenkins PH. School delinquency and the school social bond. J Res
Crime and Delinquency. 1997;34(3):337-367.
15. Roeser R, Eccles J, Strobel K. Linking the study of schooling and
mental health: selected issues and empirical illustrations at the level of the
individual. Educ Psychologist. 1998;33:153-176.
16. Hawkins JD, Guo J, Hill KG, Battin-Pearson S, Abbot RD. Longterm
effects of the Seattle Social Development Intervention on school
bonding trajectories. Appl Dev Sci. 2001;5:225-236.
17. Catalano RF, Haggerty KP, Osterle S, Fleming CB, Hawkins JD.
The importance of bonding to school for healthy development: findings
from the Social Development Research Group. 2003; unpublished manuscript.
18. Wingspread Conference. Wingspread Declaration on School
Connections. J Sch Health. 2004;74(7):233.
19. Bollen K, Hoyle RH. Perceived cohesion: a conceptual and empirical
examination. Social Forces. 1991;69:479-504.
20. Wehalge GG, Rutter RA, Smith GA, Lesko N, Fernandez RR.
Reducing the risk: schools as communities of support. New York, NY:
The Falmer Press; 1989.
21. Rosenfeld LB, Richman JM, Bowen GL. Low social support
among at-risk adolescents. Social Work Educ. 1998;20:245-260.
22. Hawkins JD, Weis JG. The Social Development Model: An integrated
approach to delinquency prevention. J Primary Prev. 1985;6:73-97.
23. Udry JR. The National Longitudinal Study of Adolescent Health
(Add Health), Waves I & II, 1994-1996 [machine-readable data file and
documentation]. Chapel Hill, NC: Carolina Population Center, University
of North Carolina at Chapel Hill; 1998.
24. McNeely CA. Connection to school as an indicator of positive
youth development. In: Lippman L, Moore K, eds. Indicators of Positive
Youth Development. Search Institute Series on Developmentally Attentive
Community and Society. Kluwer Academic/Plenum Press; (forthcoming).
25. StataCorp. Stata Statistical Software: Release 8.0. College Station,
Tex: Stata Corp; 2003.
26. Hosmer DW, Lemeshow S. Applied Logistic Regression. New
York, NY: John Wiley and Sons; 1989.
27. Johnston LD, O?Malley PM, Bachman JG. Monitoring the Future
national survey results on drug use, 1975-2002. Volume 1: Secondary
school students (NIH Publication No. 03-5375). Bethesda, Md: National
Institute on Drug Abuse; 2003.
28. Stanton-Salazar RD. Manufacturing Hope and Despair: The
School and Kin Support Networks of U.S.-Mexican Youth. New York, NY:
Teachers College Press; 2001.
29. Eccles JS, Midgefield C, Wigfield A. Development during adolescence:
the impact of stage-environment fit on young adolescents? experiences
in schools and in families. Am Psychologist. 1993;48:90-101.