Although several efforts have been made to develop instruments to measure quality of life and well-being of students within schools [8, 27–29], few studies have examined well-being as related to school social relationships in a Central American school context. This study explored the psychometric properties of the Student Perceptions of School Cohesion [SPSC] scale -a scale that focuses specifically on student connectedness and caring and supportive social relationships within the school context--based on a sample of public secondary school students living in the central region of El Salvador. Findings from this study indicated that the SPSC scale has three latent factors related to supportive school relationships, student-school connectedness, and student-teacher connectedness and that these subscales as well as the full SPSC scale have good internal consistency. The construct validity of the SPSC scale is further strengthened by the observed associations between school social environment and youth risk behaviors found with the full scale and all three subscales with physical fighting and illicit drug use.
The three factors identified in the rotated solution, supportive school relationships, student-school connectedness, and student-teacher connectedness, are interpretable from both a conceptual and theoretical basis. Items for the first factor stem from Battistich's and Hom's  "Students' sense of the school as a community" subscale of caring and supportive interpersonal relationships, which provides a more ecological measure of school climate. In addition to empirical findings of the importance of supportive school social relationships at an ecological level for reducing risk behaviors such as drug use and delinquency [see [8, 9]], this contextual level of social cohesion, with or without direct individual-to-school ties, may provide a sense of social belongingness for adolescents. Perceptions of social belongingness, even in the absence of direct interpersonal social ties, have been cited by sociological theorists as important for psychological well-being .
Our findings on the remaining SPSC scale items, which stem from research on school connectedness , indicate that these items represent two different latent factors: student-school connectedness and student-teacher connectedness. A student's level of connectedness to school has been found to be protective against a range of youth risk behaviors [10–12, 21]. The distinction of student-school connectedness and student-teacher connectedness may be important given findings by McNeely and Falci  that indicated a protective effect of teacher support on adolescent risk behavior but no effect of student connectedness to school. Our research also indicates that supportive social relationships at school represents a multidimensional construct that may differ in terms of school relationships at an ecological or contextual level, a student's direct connection to the school and people within the school, and a student's connection to his/her teacher.
Findings of the current study provide some evidence of construct validity for the use of the Student Perceptions of School Cohesion scale in measuring supportive school social relationships. Our findings of an inverse relation between the subscale of supportive school relationships and aggression mirror findings from Battistich and Hom  in their study of fifth and sixth graders in the United States. We also found inverse associations between student-school connectedness and student-teacher connectedness and aggression- findings supported by previous research [10, 13, 31]. The observed associations between the full SPSC scale and the supportive school relationships subscale and illicit drug use are also similar to findings by McBride et al.  regarding higher student social bonding at school and lower illicit drug use.
Several possible explanations may account for the less robust but still significant associations found between the student-school connectedness and student-teacher connectedness with illicit drug use. First, the illicit drug use measure was based on lifetime drug use; therefore, this measure may not have been specific enough to capture the influence of school connectedness on this behavior. Distinguishing between heavy drug users and experimental drug users might provide more information about the influence of school relationships on illicit drug use. Second, it is possible that the different levels and forms of social relationships at school may affect drug use differently. Feeling connected or not connected to classmates and teachers may be less important for drug use than perceiving supportive relationships within the broader school context, as suggested in this study. As such, students who perceive low support in the school environment, despite their individual connection to school, may seek out relationships outside of school to fill a social connectedness void. Lightfoot's [, p.10] qualitative research on adolescents and risk taking suggests that adolescents engage in risk behaviors with peers as a way of promoting cohesion, trust, and closeness in initiating new relationships and consolidating existing relationships. Engaging in substance use with peers may be one approach for developing connectedness with other adolescents when social connectedness is not perceived within the school environment. As evidence is mixed for the association between student-school connectedness and illicit drug use, with some studies finding a positive association  and others finding no association , further research on the dimensions of school connectedness and their association with drug use is warranted.
Limitations and Future Research Directions
Some limitations of the scale are worth noting. For example, as 72% to 92% of the responses fell within the first two response categories, the SPCS scale may be strengthened by expanding the response scale to capture more variance on the social relationship and school connectedness items. As this study was limited to measuring general social support, the scale may be enhanced by further conceptualizing and operationalizing support within the school environment, such as instrumental, motivational, and emotional support [33, 34]. Improved operationalization of social support would allow researchers to explore which types of support within the school environment are most important for preventing or reducing selected youth risk behaviors. Distinguishing between the source of social support--be it teacher, classmate, or friend--within the scale may also provide further insights into the social relationships that are most important for protecting against health risk behavior. As this study found that student-teacher connectedness may be a separate factor from student-school connectedness, the scale may be strengthened by including additional measures of the student-teacher relationship, such as fair treatment of students by teachers or personal interest of teachers in how the student is doing [see ]. Further research on the validity of the SPSC is also warranted given that our study focused only on two risk behavior outcomes that may be relatively distal to the construct under measurement. Future research on the test-retest reliability of the scale would also provide further understanding of the temporal stability of the measure.
Other limitations that should be considered include the cross-sectional study design for assessing associations with youth risk behavior, which limits our ability to assess causality. Future research is needed to assess both the causal role of school social cohesion in reducing risk behavior in Latin American contexts as well as the mediated pathways between social cohesion and youth risk behavior. Our measures were based on self-report, which may be prone to recall bias as well as social desirability bias in the reporting of risk behaviors and social connectedness. We should also note that the study was limited to those students who were present on the day of the survey. Because the study did not account for truant students, risk behavior prevalence estimates may have been underestimated and the generalizability of the findings to those students is limited. Lastly, further research is needed to expand the tests for reliability and validity of the scale with students from other age groups as well as other Latin American and U.S. Hispanic populations.